https://sciencetostartup.com/ https://sciencetostartup.com/api/og/default ScienceToStartup homepage Agent Operating System for Research Commercialization https://sciencetostartup.com/developers https://sciencetostartup.com/api/og/default ScienceToStartup developers Developer hub for the research commercialization agent OS https://sciencetostartup.com/proof https://sciencetostartup.com/api/og/default ScienceToStartup proof layer Proof surfaces for papers, topics, benchmarks, and datasets https://sciencetostartup.com/trends https://sciencetostartup.com/api/og/default ScienceToStartup trends Operator-facing AI research trends and narratives https://sciencetostartup.com/paper/act-wisely-cultivating-meta-cognitive-tool-use-in-agentic-multimodal-models https://sciencetostartup.com/api/og/paper/act-wisely-cultivating-meta-cognitive-tool-use-in-agentic-multimodal-models Act Wisely: Cultivating Meta-Cognitive Tool Use in Agentic Multimodal Models A framework that significantly reduces tool invocations in agentic multimodal models while improving reasoning accuracy by decoupling accuracy and efficiency optimization. https://sciencetostartup.com/paper/sim1-physics-aligned-simulator-as-zero-shot-data-scaler-in-deformable-worlds https://sciencetostartup.com/api/og/paper/sim1-physics-aligned-simulator-as-zero-shot-data-scaler-in-deformable-worlds SIM1: Physics-Aligned Simulator as Zero-Shot Data Scaler in Deformable Worlds A physics-aligned simulator that acts as a zero-shot data scaler for robotic manipulation of deformable objects, achieving parity with real-data baselines using purely synthetic data. https://sciencetostartup.com/paper/seeing-but-not-thinking-routing-distraction-in-multimodal-mixture-of-experts https://sciencetostartup.com/api/og/paper/seeing-but-not-thinking-routing-distraction-in-multimodal-mixture-of-experts Seeing but Not Thinking: Routing Distraction in Multimodal Mixture-of-Experts A method to improve reasoning in multimodal MoE models by addressing 'Seeing but Not Thinking' by guiding expert activation to overcome routing distraction. https://sciencetostartup.com/paper/openvlthinkerv2-a-generalist-multimodal-reasoning-model-for-multi-domain-visual-tasks https://sciencetostartup.com/api/og/paper/openvlthinkerv2-a-generalist-multimodal-reasoning-model-for-multi-domain-visual-tasks OpenVLThinkerV2: A Generalist Multimodal Reasoning Model for Multi-domain Visual Tasks A generalist multimodal reasoning model that significantly outperforms frontier models across 18 benchmarks by introducing a novel RL training objective and task-level shaping mechanisms. https://sciencetostartup.com/paper/avgen-bench-a-task-driven-benchmark-for-multi-granular-evaluation-of-text-to-audio-video-generation https://sciencetostartup.com/api/og/paper/avgen-bench-a-task-driven-benchmark-for-multi-granular-evaluation-of-text-to-audio-video-generation AVGen-Bench: A Task-Driven Benchmark for Multi-Granular Evaluation of Text-to-Audio-Video Generation A task-driven benchmark and evaluation framework for text-to-audio-video generation that reveals significant gaps in semantic controllability. https://sciencetostartup.com/paper/rewardflow-generate-images-by-optimizing-what-you-reward https://sciencetostartup.com/api/og/paper/rewardflow-generate-images-by-optimizing-what-you-reward RewardFlow: Generate Images by Optimizing What You Reward A new image generation tool leveraging multi-reward Langevin dynamics for state-of-the-art image editing. https://sciencetostartup.com/paper/psi-shared-state-as-the-missing-layer-for-coherent-ai-generated-instruments-in-personal-ai-agents https://sciencetostartup.com/api/og/paper/psi-shared-state-as-the-missing-layer-for-coherent-ai-generated-instruments-in-personal-ai-agents PSI: Shared State as the Missing Layer for Coherent AI-Generated Instruments in Personal AI Agents PSI is a shared-state architecture that transforms independently generated AI modules into coherent, connected personal computing environments accessible through GUIs and chat agents. https://sciencetostartup.com/paper/ads-in-ai-chatbots-an-analysis-of-how-large-language-models-navigate-conflicts-of-interest https://sciencetostartup.com/api/og/paper/ads-in-ai-chatbots-an-analysis-of-how-large-language-models-navigate-conflicts-of-interest Ads in AI Chatbots? An Analysis of How Large Language Models Navigate Conflicts of Interest This paper analyzes how current LLMs navigate conflicts of interest between user welfare and company incentives, finding a majority prioritize company revenue over user benefit. https://sciencetostartup.com/paper/what-drives-representation-steering-a-mechanistic-case-study-on-steering-refusal https://sciencetostartup.com/api/og/paper/what-drives-representation-steering-a-mechanistic-case-study-on-steering-refusal What Drives Representation Steering? A Mechanistic Case Study on Steering Refusal This research investigates the internal mechanisms of steering vectors in large language models, revealing their primary interaction with the attention mechanism's OV circuit and enabling significant sparsification. https://sciencetostartup.com/paper/clawbench-can-ai-agents-complete-everyday-online-tasks https://sciencetostartup.com/api/og/paper/clawbench-can-ai-agents-complete-everyday-online-tasks ClawBench: Can AI Agents Complete Everyday Online Tasks? ClawBench is a new benchmark for evaluating AI agents on everyday online tasks across 144 live platforms, revealing current frontier models can only complete a small fraction of these complex, real-world challenges. https://sciencetostartup.com/paper/differentially-private-language-generation-and-identification-in-the-limit https://sciencetostartup.com/api/og/paper/differentially-private-language-generation-and-identification-in-the-limit Differentially Private Language Generation and Identification in the Limit This research explores differentially private language generation and identification, showing privacy has no qualitative cost for generation but creates fundamental barriers for identification. https://sciencetostartup.com/paper/quantifying-explanation-consistency-the-c-score-metric-for-cam-based-explainability-in-medical-image-classification https://sciencetostartup.com/api/og/paper/quantifying-explanation-consistency-the-c-score-metric-for-cam-based-explainability-in-medical-image-classification Quantifying Explanation Consistency: The C-Score Metric for CAM-Based Explainability in Medical Image Classification The C-Score metric quantifies explanation consistency in medical image classification, providing an early warning signal of model instability and enabling architecture-specific deployment recommendations. https://sciencetostartup.com/paper/piarena-a-platform-for-prompt-injection-evaluation https://sciencetostartup.com/api/og/paper/piarena-a-platform-for-prompt-injection-evaluation PIArena: A Platform for Prompt Injection Evaluation A unified platform for evaluating prompt injection attacks and defenses, enabling reliable comparison and uncovering limitations of current security measures. https://sciencetostartup.com/paper/supernova-eliciting-general-reasoning-in-llms-with-reinforcement-learning-on-natural-instructions https://sciencetostartup.com/api/og/paper/supernova-eliciting-general-reasoning-in-llms-with-reinforcement-learning-on-natural-instructions SUPERNOVA: Eliciting General Reasoning in LLMs with Reinforcement Learning on Natural Instructions SUPERNOVA enhances general reasoning in language models through a curated data framework for reinforcement learning with verifiable rewards. https://sciencetostartup.com/paper/faithful-grpo-improving-visual-spatial-reasoning-in-multimodal-language-models-via-constrained-policy-optimization https://sciencetostartup.com/api/og/paper/faithful-grpo-improving-visual-spatial-reasoning-in-multimodal-language-models-via-constrained-policy-optimization Faithful GRPO: Improving Visual Spatial Reasoning in Multimodal Language Models via Constrained Policy Optimization Faithful GRPO improves visual spatial reasoning in multimodal models by enforcing logical consistency and visual grounding, leading to more accurate and trustworthy answers. https://sciencetostartup.com/paper/ttvs-boosting-self-exploring-reinforcement-learning-via-test-time-variational-synthesis https://sciencetostartup.com/api/og/paper/ttvs-boosting-self-exploring-reinforcement-learning-via-test-time-variational-synthesis TTVS: Boosting Self-Exploring Reinforcement Learning via Test-time Variational Synthesis TTVS enables large reasoning models to self-evolve at test time by dynamically synthesizing diverse variations of unlabeled queries, improving performance in specialized domains. https://sciencetostartup.com/paper/from-safety-risk-to-design-principle-peer-preservation-in-multi-agent-llm-systems-and-its-implications-for-orchestrated https://sciencetostartup.com/api/og/paper/from-safety-risk-to-design-principle-peer-preservation-in-multi-agent-llm-systems-and-its-implications-for-orchestrated From Safety Risk to Design Principle: Peer-Preservation in Multi-Agent LLM Systems and Its Implications for Orchestrated Democratic Discourse Analysis This paper explores emergent 'peer-preservation' alignment issues in multi-agent LLM systems and proposes architectural mitigations for orchestrated democratic discourse analysis. https://sciencetostartup.com/paper/ovs-dino-open-vocabulary-segmentation-via-structure-aligned-sam-dino-with-language-guidance https://sciencetostartup.com/api/og/paper/ovs-dino-open-vocabulary-segmentation-via-structure-aligned-sam-dino-with-language-guidance OVS-DINO: Open-Vocabulary Segmentation via Structure-Aligned SAM-DINO with Language Guidance OVS-DINO revitalizes edge-sensitivity in DINO for open-vocabulary segmentation by structurally aligning with SAM, achieving state-of-the-art results. https://sciencetostartup.com/paper/a-machine-learning-framework-for-turbofan-health-estimation-via-inverse-problem-formulation https://sciencetostartup.com/api/og/paper/a-machine-learning-framework-for-turbofan-health-estimation-via-inverse-problem-formulation A Machine Learning Framework for Turbofan Health Estimation via Inverse Problem Formulation A machine learning framework for turbofan health estimation using a new dataset and self-supervised learning to address the ill-posed inverse problem. https://sciencetostartup.com/paper/crashsight-a-phase-aware-infrastructure-centric-video-benchmark-for-traffic-crash-scene-understanding-and-reasoning https://sciencetostartup.com/api/og/paper/crashsight-a-phase-aware-infrastructure-centric-video-benchmark-for-traffic-crash-scene-understanding-and-reasoning CrashSight: A Phase-Aware, Infrastructure-Centric Video Benchmark for Traffic Crash Scene Understanding and Reasoning CrashSight is a large-scale benchmark for traffic crash scene understanding using roadside camera data, evaluating vision-language models' reasoning in safety-critical scenarios. https://sciencetostartup.com/paper/knowu-bench-towards-interactive-proactive-and-personalized-mobile-agent-evaluation https://sciencetostartup.com/api/og/paper/knowu-bench-towards-interactive-proactive-and-personalized-mobile-agent-evaluation KnowU-Bench: Towards Interactive, Proactive, and Personalized Mobile Agent Evaluation A new benchmark and simulator for evaluating personalized mobile agents that can infer user preferences and provide proactive assistance in real-time GUI environments. https://sciencetostartup.com/paper/hst-hgn-heterogeneous-spatial-temporal-hypergraph-networks-with-bidirectional-state-space-models-for-global-fatigue-asse https://sciencetostartup.com/api/og/paper/hst-hgn-heterogeneous-spatial-temporal-hypergraph-networks-with-bidirectional-state-space-models-for-global-fatigue-asse HST-HGN: Heterogeneous Spatial-Temporal Hypergraph Networks with Bidirectional State Space Models for Global Fatigue Assessment A novel hypergraph network with bidirectional state space models for efficient and accurate driver fatigue assessment from untrimmed videos, suitable for edge deployment. https://sciencetostartup.com/paper/small-scale-photonic-kolmogorov-arnold-networks-using-standard-telecom-nonlinear-modules https://sciencetostartup.com/api/og/paper/small-scale-photonic-kolmogorov-arnold-networks-using-standard-telecom-nonlinear-modules Small-scale photonic Kolmogorov-Arnold networks using standard telecom nonlinear modules Small-scale photonic Kolmogorov-Arnold networks implemented with standard telecom components for ultrafast inference across various tasks, overcoming optical-electrical-optical bottlenecks. https://sciencetostartup.com/paper/kv-cache-offloading-for-context-intensive-tasks https://sciencetostartup.com/api/og/paper/kv-cache-offloading-for-context-intensive-tasks KV Cache Offloading for Context-Intensive Tasks A new benchmark and improved KV cache offloading strategy to address performance degradation on context-intensive LLM tasks, enabling more accurate long-context processing. https://sciencetostartup.com/paper/learning-who-disagrees-demographic-importance-weighting-for-modeling-annotator-distributions-with-diadem https://sciencetostartup.com/api/og/paper/learning-who-disagrees-demographic-importance-weighting-for-modeling-annotator-distributions-with-diadem Learning Who Disagrees: Demographic Importance Weighting for Modeling Annotator Distributions with DiADEM A neural architecture that models annotator disagreement by learning the importance of demographic factors, outperforming LLMs on subjective content labeling. https://sciencetostartup.com/paper/on-board-telemetry-monitoring-in-autonomous-satellites-challenges-and-opportunities https://sciencetostartup.com/api/og/paper/on-board-telemetry-monitoring-in-autonomous-satellites-challenges-and-opportunities On-board Telemetry Monitoring in Autonomous Satellites: Challenges and Opportunities A framework for explainable AI in autonomous satellite fault detection, using neural anomaly detectors with interpretable indicators. https://sciencetostartup.com/paper/synthetic-data-for-any-differentiable-target https://sciencetostartup.com/api/og/paper/synthetic-data-for-any-differentiable-target Synthetic Data for any Differentiable Target A flexible synthetic data generator for customizable machine learning tasks. https://sciencetostartup.com/paper/exploring-temporal-representation-in-neural-processes-for-multimodal-action-prediction https://sciencetostartup.com/api/og/paper/exploring-temporal-representation-in-neural-processes-for-multimodal-action-prediction Exploring Temporal Representation in Neural Processes for Multimodal Action Prediction A revised Conditional Neural Process model that improves temporal representation for multimodal action prediction in robotics. https://sciencetostartup.com/paper/selective-attention-system-sas-device-addressed-speech-detection-for-real-time-on-device-voice-ai https://sciencetostartup.com/api/og/paper/selective-attention-system-sas-device-addressed-speech-detection-for-real-time-on-device-voice-ai Selective Attention System (SAS): Device-Addressed Speech Detection for Real-Time On-Device Voice AI A selective attention system for device-addressed speech detection that operates on-device with low latency and footprint, achieving high accuracy with optional video fusion. https://sciencetostartup.com/paper/verify-before-you-commit-towards-faithful-reasoning-in-llm-agents-via-self-auditing https://sciencetostartup.com/api/og/paper/verify-before-you-commit-towards-faithful-reasoning-in-llm-agents-via-self-auditing Verify Before You Commit: Towards Faithful Reasoning in LLM Agents via Self-Auditing Develop a self-auditing tool for LLM agents to ensure faithful reasoning. https://sciencetostartup.com/paper/zero-shot-multivariate-time-series-forecasting-using-tabular-prior-fitted-networks https://sciencetostartup.com/api/og/paper/zero-shot-multivariate-time-series-forecasting-using-tabular-prior-fitted-networks Zero-shot Multivariate Time Series Forecasting Using Tabular Prior Fitted Networks A framework for multivariate time series forecasting using tabular foundation models by recasting the problem into a series of scalar regression problems solvable zero-shot. https://sciencetostartup.com/paper/adaptive-input-training-for-many-to-one-pre-training-on-time-series-classification https://sciencetostartup.com/api/og/paper/adaptive-input-training-for-many-to-one-pre-training-on-time-series-classification ADAPTive Input Training for Many-to-One Pre-Training on Time-Series Classification ADAPT is a new pre-training paradigm for time-series data that enables mixed-batch pre-training across diverse datasets, setting new state-of-the-art performance for classification benchmarks. https://sciencetostartup.com/paper/phantasia-context-adaptive-backdoors-in-vision-language-models https://sciencetostartup.com/api/og/paper/phantasia-context-adaptive-backdoors-in-vision-language-models Phantasia: Context-Adaptive Backdoors in Vision Language Models A novel backdoor attack for Vision-Language Models that dynamically aligns poisoned outputs with input semantics for improved stealth and adaptability. https://sciencetostartup.com/paper/awakening-the-sleeping-agent-lean-specific-agentic-data-reactivates-general-tool-use-in-goedel-prover https://sciencetostartup.com/api/og/paper/awakening-the-sleeping-agent-lean-specific-agentic-data-reactivates-general-tool-use-in-goedel-prover Awakening the Sleeping Agent: Lean-Specific Agentic Data Reactivates General Tool Use in Goedel Prover A small amount of domain-specific agentic data can reactivate dormant tool-use capabilities in large language models, significantly improving performance on diverse tasks. https://sciencetostartup.com/paper/tasu2-controllable-ctc-simulation-for-alignment-and-low-resource-adaptation-of-speech-llms https://sciencetostartup.com/api/og/paper/tasu2-controllable-ctc-simulation-for-alignment-and-low-resource-adaptation-of-speech-llms TASU2: Controllable CTC Simulation for Alignment and Low-Resource Adaptation of Speech LLMs TASU2 is a controllable CTC simulation framework for speech LLMs that enables principled post-training curricula for improved alignment and low-resource adaptation. https://sciencetostartup.com/paper/a-gan-and-llm-driven-data-augmentation-framework-for-dynamic-linguistic-pattern-modeling-in-chinese-sarcasm-detection https://sciencetostartup.com/api/og/paper/a-gan-and-llm-driven-data-augmentation-framework-for-dynamic-linguistic-pattern-modeling-in-chinese-sarcasm-detection A GAN and LLM-Driven Data Augmentation Framework for Dynamic Linguistic Pattern Modeling in Chinese Sarcasm Detection A GAN and LLM-driven framework that dynamically models user linguistic patterns for enhanced Chinese sarcasm detection, achieving state-of-the-art results. https://sciencetostartup.com/paper/skillclaw-let-skills-evolve-collectively-with-agentic-evolver https://sciencetostartup.com/api/og/paper/skillclaw-let-skills-evolve-collectively-with-agentic-evolver SkillClaw: Let Skills Evolve Collectively with Agentic Evolver SkillClaw enables LLM agents to collectively evolve their skills by learning from cross-user interactions, improving performance system-wide without user effort. https://sciencetostartup.com/paper/don-t-overthink-it-inter-rollout-action-agreement-as-a-free-adaptive-compute-signal-for-llm-agents https://sciencetostartup.com/api/og/paper/don-t-overthink-it-inter-rollout-action-agreement-as-a-free-adaptive-compute-signal-for-llm-agents Don't Overthink It: Inter-Rollout Action Agreement as a Free Adaptive-Compute Signal for LLM Agents TrACE adaptively allocates LLM compute for agents by measuring inter-rollout action agreement, reducing LLM calls while maintaining accuracy. https://sciencetostartup.com/paper/scaling-aware-data-selection-for-end-to-end-autonomous-driving-systems https://sciencetostartup.com/api/og/paper/scaling-aware-data-selection-for-end-to-end-autonomous-driving-systems Scaling-Aware Data Selection for End-to-End Autonomous Driving Systems MOSAIC is a data selection framework for autonomous driving that uses neural scaling laws to optimize training data mixtures, reducing data needs by up to 80%. https://sciencetostartup.com/paper/scalable-neural-decoders-for-practical-fault-tolerant-quantum-computation https://sciencetostartup.com/api/og/paper/scalable-neural-decoders-for-practical-fault-tolerant-quantum-computation Scalable Neural Decoders for Practical Fault-Tolerant Quantum Computation A convolutional neural network decoder for quantum error correction achieves significantly lower logical error rates and higher throughput than existing methods. https://sciencetostartup.com/paper/aspect-analogical-semantic-policy-execution-via-language-conditioned-transfer https://sciencetostartup.com/api/og/paper/aspect-analogical-semantic-policy-execution-via-language-conditioned-transfer ASPECT:Analogical Semantic Policy Execution via Language Conditioned Transfer A reinforcement learning agent that uses a Large Language Model as a semantic operator to achieve zero-shot transfer to novel analogous tasks. https://sciencetostartup.com/paper/human-ai-collaboration-reconfigures-group-regulation-from-socially-shared-to-hybrid-co-regulation https://sciencetostartup.com/api/og/paper/human-ai-collaboration-reconfigures-group-regulation-from-socially-shared-to-hybrid-co-regulation Human-AI Collaboration Reconfigures Group Regulation from Socially Shared to Hybrid Co-Regulation This paper investigates how generative AI impacts group regulation in collaborative learning environments, shifting from socially shared to hybrid co-regulation. https://sciencetostartup.com/paper/pokegym-a-visually-driven-long-horizon-benchmark-for-vision-language-models https://sciencetostartup.com/api/og/paper/pokegym-a-visually-driven-long-horizon-benchmark-for-vision-language-models PokeGym: A Visually-Driven Long-Horizon Benchmark for Vision-Language Models PokeGym is a visually-driven benchmark for Vision-Language Models in complex 3D environments, revealing physical deadlock recovery as a key bottleneck. https://sciencetostartup.com/paper/instap-instance-aware-vision-language-pre-train-for-spatial-temporal-understanding https://sciencetostartup.com/api/og/paper/instap-instance-aware-vision-language-pre-train-for-spatial-temporal-understanding InstAP: Instance-Aware Vision-Language Pre-Train for Spatial-Temporal Understanding InstAP is an instance-aware pre-training framework for vision-language models that improves both instance-level and global understanding. https://sciencetostartup.com/paper/dead-weights-live-signals-feedforward-graphs-of-frozen-language-models https://sciencetostartup.com/api/og/paper/dead-weights-live-signals-feedforward-graphs-of-frozen-language-models Dead Weights, Live Signals: Feedforward Graphs of Frozen Language Models A novel feedforward graph architecture that composes heterogeneous frozen LLMs to achieve state-of-the-art performance on challenging benchmarks with minimal trainable parameters. https://sciencetostartup.com/paper/lost-in-the-hype-revealing-and-dissecting-the-performance-degradation-of-medical-multimodal-large-language-models-in-ima https://sciencetostartup.com/api/og/paper/lost-in-the-hype-revealing-and-dissecting-the-performance-degradation-of-medical-multimodal-large-language-models-in-ima Lost in the Hype: Revealing and Dissecting the Performance Degradation of Medical Multimodal Large Language Models in Image Classification Investigates the performance degradation of medical multimodal LLMs in image classification, identifying key failure modes to guide future development. https://sciencetostartup.com/paper/promedical-hierarchical-fine-grained-criteria-modeling-for-medical-llm-alignment-via-explicit-injection https://sciencetostartup.com/api/og/paper/promedical-hierarchical-fine-grained-criteria-modeling-for-medical-llm-alignment-via-explicit-injection ProMedical: Hierarchical Fine-Grained Criteria Modeling for Medical LLM Alignment via Explicit Injection A novel alignment framework for medical LLMs that uses fine-grained clinical criteria and explicit injection to achieve superior accuracy and safety compliance. https://sciencetostartup.com/paper/multi-modal-learning-meets-genetic-programming-analyzing-alignment-in-latent-space-optimization https://sciencetostartup.com/api/og/paper/multi-modal-learning-meets-genetic-programming-analyzing-alignment-in-latent-space-optimization Multi-Modal Learning meets Genetic Programming: Analyzing Alignment in Latent Space Optimization Investigates the effectiveness of multi-modal latent space optimization for symbolic regression, revealing limitations in current alignment techniques. https://sciencetostartup.com/paper/histdit-a-structure-aware-latent-conditional-diffusion-model-for-high-fidelity-virtual-staining-in-histopathology https://sciencetostartup.com/api/og/paper/histdit-a-structure-aware-latent-conditional-diffusion-model-for-high-fidelity-virtual-staining-in-histopathology HistDiT: A Structure-Aware Latent Conditional Diffusion Model for High-Fidelity Virtual Staining in Histopathology A structure-aware latent conditional diffusion model for high-fidelity virtual staining in histopathology, outperforming existing methods. https://sciencetostartup.com/paper/securing-retrieval-augmented-generation-a-taxonomy-of-attacks-defenses-and-future-directions https://sciencetostartup.com/api/og/paper/securing-retrieval-augmented-generation-a-taxonomy-of-attacks-defenses-and-future-directions Securing Retrieval-Augmented Generation: A Taxonomy of Attacks, Defenses, and Future Directions A taxonomy of attacks and defenses for retrieval-augmented generation (RAG) systems, highlighting fragmented current defenses. https://sciencetostartup.com/paper/dmax-aggressive-parallel-decoding-for-dllms https://sciencetostartup.com/api/og/paper/dmax-aggressive-parallel-decoding-for-dllms DMax: Aggressive Parallel Decoding for dLLMs DMax offers aggressive parallel decoding for diffusion language models, significantly increasing throughput while preserving generation quality. https://sciencetostartup.com/paper/selar-selective-latent-reasoning-in-large-language-models https://sciencetostartup.com/api/og/paper/selar-selective-latent-reasoning-in-large-language-models SeLaR: Selective Latent Reasoning in Large Language Models SeLaR is a training-free framework that selectively uses latent reasoning to improve LLM performance on complex reasoning tasks. https://sciencetostartup.com/paper/u-cece-a-universal-multi-resolution-framework-for-conceptual-counterfactual-explanations https://sciencetostartup.com/api/og/paper/u-cece-a-universal-multi-resolution-framework-for-conceptual-counterfactual-explanations U-CECE: A Universal Multi-Resolution Framework for Conceptual Counterfactual Explanations A universal framework for generating conceptual counterfactual explanations for AI models, balancing expressivity and efficiency across different data regimes. https://sciencetostartup.com/paper/can-vision-language-models-judge-action-quality-an-empirical-evaluation https://sciencetostartup.com/api/og/paper/can-vision-language-models-judge-action-quality-an-empirical-evaluation Can Vision Language Models Judge Action Quality? An Empirical Evaluation Evaluating the capability of Vision Language Models for Action Quality Assessment, revealing significant limitations and biases. https://sciencetostartup.com/paper/ciao-code-in-architecture-out-automated-software-architecture-documentation-with-large-language-models https://sciencetostartup.com/api/og/paper/ciao-code-in-architecture-out-automated-software-architecture-documentation-with-large-language-models CIAO - Code In Architecture Out - Automated Software Architecture Documentation with Large Language Models Automated system-level software architecture documentation generation from code repositories using LLMs, providing valuable and cost-effective insights. https://sciencetostartup.com/paper/distributed-multi-layer-editing-for-rule-level-knowledge-in-large-language-models https://sciencetostartup.com/api/og/paper/distributed-multi-layer-editing-for-rule-level-knowledge-in-large-language-models Distributed Multi-Layer Editing for Rule-Level Knowledge in Large Language Models A distributed multi-layer editing approach for rule-level knowledge in LLMs, improving consistency and understanding across different knowledge forms. https://sciencetostartup.com/paper/qarima-a-quantum-approach-to-classical-time-series-analysis https://sciencetostartup.com/api/og/paper/qarima-a-quantum-approach-to-classical-time-series-analysis QARIMA: A Quantum Approach To Classical Time Series Analysis A quantum-inspired ARIMA methodology for enhanced time series analysis. https://sciencetostartup.com/paper/acf-a-collaborative-framework-for-agent-covert-communication-under-cognitive-asymmetry https://sciencetostartup.com/api/og/paper/acf-a-collaborative-framework-for-agent-covert-communication-under-cognitive-asymmetry ACF: A Collaborative Framework for Agent Covert Communication under Cognitive Asymmetry A framework for covert communication among autonomous agents overcoming cognitive asymmetry. https://sciencetostartup.com/paper/neural-symbolic-knowledge-tracing-injecting-educational-knowledge-into-deep-learning-for-responsible-learner-modelling https://sciencetostartup.com/api/og/paper/neural-symbolic-knowledge-tracing-injecting-educational-knowledge-into-deep-learning-for-responsible-learner-modelling Neural-Symbolic Knowledge Tracing: Injecting Educational Knowledge into Deep Learning for Responsible Learner Modelling A neural-symbolic approach for responsible learner modeling in education. https://sciencetostartup.com/paper/dbmf-a-dual-branch-multimodal-framework-for-out-of-distribution-detection https://sciencetostartup.com/api/og/paper/dbmf-a-dual-branch-multimodal-framework-for-out-of-distribution-detection DBMF: A Dual-Branch Multimodal Framework for Out-of-Distribution Detection A dual-branch multimodal framework for robust out-of-distribution detection. https://sciencetostartup.com/paper/behavior-aware-item-modeling-via-dynamic-procedural-solution-representations-for-knowledge-tracing https://sciencetostartup.com/api/og/paper/behavior-aware-item-modeling-via-dynamic-procedural-solution-representations-for-knowledge-tracing Behavior-Aware Item Modeling via Dynamic Procedural Solution Representations for Knowledge Tracing A framework that enhances knowledge tracing by integrating dynamic procedural solution stages into item representations, improving prediction of learner performance. https://sciencetostartup.com/paper/hypermem-hypergraph-memory-for-long-term-conversations https://sciencetostartup.com/api/og/paper/hypermem-hypergraph-memory-for-long-term-conversations HyperMem: Hypergraph Memory for Long-Term Conversations HyperMem is a hypergraph-based memory architecture for conversational agents that captures high-order associations for more coherent and personalized long-term dialogues. https://sciencetostartup.com/paper/from-phenomenological-fitting-to-endogenous-deduction-a-paradigm-leap-via-meta-principle-physics-architecture https://sciencetostartup.com/api/og/paper/from-phenomenological-fitting-to-endogenous-deduction-a-paradigm-leap-via-meta-principle-physics-architecture From Phenomenological Fitting to Endogenous Deduction: A Paradigm Leap via Meta-Principle Physics Architecture A Meta-Principle Physics Architecture that embeds core physical principles like connectivity, conservation, and periodicity into neural networks for improved physical reasoning and generalization. https://sciencetostartup.com/paper/hiro-nav-hybrid-reasoning-enables-efficient-embodied-navigation https://sciencetostartup.com/api/og/paper/hiro-nav-hybrid-reasoning-enables-efficient-embodied-navigation HiRO-Nav: Hybrid ReasOning Enables Efficient Embodied Navigation HiRO-Nav is an embodied navigation agent that adaptively uses reasoning only for high-entropy actions, reducing computational cost while improving decision quality. https://sciencetostartup.com/paper/grounding-clinical-ai-competency-in-human-cognition-through-the-clinical-world-model-and-skill-mix-framework https://sciencetostartup.com/api/og/paper/grounding-clinical-ai-competency-in-human-cognition-through-the-clinical-world-model-and-skill-mix-framework Grounding Clinical AI Competency in Human Cognition Through the Clinical World Model and Skill-Mix Framework A framework to formalize clinical AI competency by modeling the interactions between patient, provider, and ecosystem, grounded in clinical cognition. https://sciencetostartup.com/paper/editcaption-human-aligned-instruction-synthesis-for-image-editing-via-supervised-fine-tuning-and-direct-preference-optim https://sciencetostartup.com/api/og/paper/editcaption-human-aligned-instruction-synthesis-for-image-editing-via-supervised-fine-tuning-and-direct-preference-optim EditCaption: Human-Aligned Instruction Synthesis for Image Editing via Supervised Fine-Tuning and Direct Preference Optimization EditCaption synthesizes human-aligned instructions for image editing by combining supervised fine-tuning and direct preference optimization, significantly improving VLM accuracy. https://sciencetostartup.com/paper/medvr-annotation-free-medical-visual-reasoning-via-agentic-reinforcement-learning https://sciencetostartup.com/api/og/paper/medvr-annotation-free-medical-visual-reasoning-via-agentic-reinforcement-learning MedVR: Annotation-Free Medical Visual Reasoning via Agentic Reinforcement Learning MedVR enables annotation-free medical visual reasoning for VLMs using agentic reinforcement learning, improving robustness and transparency. https://sciencetostartup.com/paper/at-add-all-type-audio-deepfake-detection-challenge-evaluation-plan https://sciencetostartup.com/api/og/paper/at-add-all-type-audio-deepfake-detection-challenge-evaluation-plan AT-ADD: All-Type Audio Deepfake Detection Challenge Evaluation Plan The AT-ADD challenge aims to advance all-type audio deepfake detection beyond speech-centric methods for robust multimedia forensics. https://sciencetostartup.com/paper/aligning-agents-via-planning-a-benchmark-for-trajectory-level-reward-modeling https://sciencetostartup.com/api/og/paper/aligning-agents-via-planning-a-benchmark-for-trajectory-level-reward-modeling Aligning Agents via Planning: A Benchmark for Trajectory-Level Reward Modeling A new benchmark and evaluation suite for training reward models that align AI agents capable of complex tool use and reasoning. https://sciencetostartup.com/paper/oceanmae-a-foundation-model-for-ocean-remote-sensing https://sciencetostartup.com/api/og/paper/oceanmae-a-foundation-model-for-ocean-remote-sensing OceanMAE: A Foundation Model for Ocean Remote Sensing A foundation model for ocean remote sensing that improves marine pollutant detection and bathymetry estimation by integrating multispectral data with ocean-specific features. https://sciencetostartup.com/paper/activation-steering-for-aligned-open-ended-generation-without-sacrificing-coherence https://sciencetostartup.com/api/og/paper/activation-steering-for-aligned-open-ended-generation-without-sacrificing-coherence Activation Steering for Aligned Open-ended Generation without Sacrificing Coherence A novel activation steering method for large language models that improves alignment without sacrificing coherence, addressing brittleness in generation. https://sciencetostartup.com/paper/viva-a-video-generative-value-model-for-robot-reinforcement-learning https://sciencetostartup.com/api/og/paper/viva-a-video-generative-value-model-for-robot-reinforcement-learning ViVa: A Video-Generative Value Model for Robot Reinforcement Learning A video-generative value model for robot reinforcement learning that improves task progress estimation by leveraging spatiotemporal priors from video data. https://sciencetostartup.com/paper/face-d-2-cl-multi-domain-synergistic-representation-with-dual-continual-learning-for-facial-deepfake-detection https://sciencetostartup.com/api/og/paper/face-d-2-cl-multi-domain-synergistic-representation-with-dual-continual-learning-for-facial-deepfake-detection Face-D(^2)CL: Multi-Domain Synergistic Representation with Dual Continual Learning for Facial DeepFake Detection A dual continual learning framework for facial DeepFake detection that fuses spatial and frequency-domain features to adapt to evolving forgery patterns without historical data replay. https://sciencetostartup.com/paper/multimodal-reasoning-with-llm-for-encrypted-traffic-interpretation-a-benchmark https://sciencetostartup.com/api/og/paper/multimodal-reasoning-with-llm-for-encrypted-traffic-interpretation-a-benchmark Multimodal Reasoning with LLM for Encrypted Traffic Interpretation: A Benchmark An end-to-end multimodal reasoning framework that uses LLMs to interpret encrypted network traffic, generating human-readable reports grounded in raw byte data. https://sciencetostartup.com/paper/alloc-moe-budget-aware-expert-activation-allocation-for-efficient-mixture-of-experts-inference https://sciencetostartup.com/api/og/paper/alloc-moe-budget-aware-expert-activation-allocation-for-efficient-mixture-of-experts-inference Alloc-MoE: Budget-Aware Expert Activation Allocation for Efficient Mixture-of-Experts Inference Alloc-MoE optimizes expert activation allocation in Mixture-of-Experts models for efficient inference, maintaining performance under budget constraints. https://sciencetostartup.com/paper/beyond-stochastic-exploration-what-makes-training-data-valuable-for-agentic-search https://sciencetostartup.com/api/og/paper/beyond-stochastic-exploration-what-makes-training-data-valuable-for-agentic-search Beyond Stochastic Exploration: What Makes Training Data Valuable for Agentic Search Hierarchical Experience (HiExp) framework enhances LLM search agents by transforming raw reasoning trajectories into hierarchical knowledge for strategic, experience-driven exploration. https://sciencetostartup.com/paper/legodiffusion-micro-serving-text-to-image-diffusion-workflows https://sciencetostartup.com/api/og/paper/legodiffusion-micro-serving-text-to-image-diffusion-workflows LegoDiffusion: Micro-Serving Text-to-Image Diffusion Workflows A system for micro-serving text-to-image diffusion workflows that optimizes resource management and performance. https://sciencetostartup.com/paper/uni-vigu-towards-unified-video-generation-and-understanding-via-a-diffusion-based-video-generator https://sciencetostartup.com/api/og/paper/uni-vigu-towards-unified-video-generation-and-understanding-via-a-diffusion-based-video-generator Uni-ViGU: Towards Unified Video Generation and Understanding via A Diffusion-Based Video Generator A unified framework for video generation and understanding that leverages a video generator as the foundation. https://sciencetostartup.com/paper/small-vision-language-models-are-smart-compressors-for-long-video-understanding https://sciencetostartup.com/api/og/paper/small-vision-language-models-are-smart-compressors-for-long-video-understanding Small Vision-Language Models are Smart Compressors for Long Video Understanding An efficient framework that compresses long videos for downstream understanding using small vision-language models. https://sciencetostartup.com/paper/revise-a-framework-for-revising-ocred-text-in-practical-information-systems-with-data-contamination-strategy https://sciencetostartup.com/api/og/paper/revise-a-framework-for-revising-ocred-text-in-practical-information-systems-with-data-contamination-strategy Revise: A Framework for Revising OCRed text in Practical Information Systems with Data Contamination Strategy A framework for correcting OCR errors and improving document retrieval and question answering. https://sciencetostartup.com/paper/tadp-rme-a-trust-adaptive-differential-privacy-framework-for-enhancing-reliability-of-data-driven-systems https://sciencetostartup.com/api/og/paper/tadp-rme-a-trust-adaptive-differential-privacy-framework-for-enhancing-reliability-of-data-driven-systems TADP-RME: A Trust-Adaptive Differential Privacy Framework for Enhancing Reliability of Data-Driven Systems A framework for adaptive differential privacy that modulates privacy budgets based on user trust and uses manifold embedding to disrupt inference attacks. https://sciencetostartup.com/paper/ov-stitcher-a-global-context-aware-framework-for-training-free-open-vocabulary-semantic-segmentation https://sciencetostartup.com/api/og/paper/ov-stitcher-a-global-context-aware-framework-for-training-free-open-vocabulary-semantic-segmentation OV-Stitcher: A Global Context-Aware Framework for Training-Free Open-Vocabulary Semantic Segmentation A training-free framework that stitches sub-image features to enable global context awareness for open-vocabulary semantic segmentation. https://sciencetostartup.com/paper/atlasocr-building-the-first-open-source-darija-ocr-model-with-vision-language-models https://sciencetostartup.com/api/og/paper/atlasocr-building-the-first-open-source-darija-ocr-model-with-vision-language-models AtlasOCR: Building the First Open-Source Darija OCR Model with Vision Language Models AtlasOCR is the first open-source Darija OCR model, fine-tuned from a VLM using efficient training strategies and a custom dataset. https://sciencetostartup.com/paper/implicitmembench-measuring-unconscious-behavioral-adaptation-in-large-language-models https://sciencetostartup.com/api/og/paper/implicitmembench-measuring-unconscious-behavioral-adaptation-in-large-language-models ImplicitMemBench: Measuring Unconscious Behavioral Adaptation in Large Language Models ImplicitMemBench is a new benchmark for LLM agents that measures unconscious behavioral adaptation, revealing limitations in current models. https://sciencetostartup.com/paper/from-gaze-to-guidance-interpreting-and-adapting-to-users-cognitive-needs-with-multimodal-gaze-aware-ai-assistants https://sciencetostartup.com/api/og/paper/from-gaze-to-guidance-interpreting-and-adapting-to-users-cognitive-needs-with-multimodal-gaze-aware-ai-assistants From Gaze to Guidance: Interpreting and Adapting to Users' Cognitive Needs with Multimodal Gaze-Aware AI Assistants A gaze-aware AI assistant that uses egocentric video to understand user cognitive needs and provide more accurate and personalized assistance. https://sciencetostartup.com/paper/governed-capability-evolution-for-embodied-agents-safe-upgrade-compatibility-checking-and-runtime-rollback-for-embodied https://sciencetostartup.com/api/og/paper/governed-capability-evolution-for-embodied-agents-safe-upgrade-compatibility-checking-and-runtime-rollback-for-embodied Governed Capability Evolution for Embodied Agents: Safe Upgrade, Compatibility Checking, and Runtime Rollback for Embodied Capability Modules A framework for safely upgrading and managing capabilities of embodied agents, ensuring compatibility and preventing runtime failures. https://sciencetostartup.com/paper/3drawagent-teaching-llm-to-draw-in-3d-with-early-contrastive-experience https://sciencetostartup.com/api/og/paper/3drawagent-teaching-llm-to-draw-in-3d-with-early-contrastive-experience 3DrawAgent: Teaching LLM to Draw in 3D with Early Contrastive Experience A training-free framework that teaches LLMs to draw 3D sketches using early contrastive experience and geometric feedback. https://sciencetostartup.com/paper/line-llm-based-iterative-neuron-explanations-for-vision-models https://sciencetostartup.com/api/og/paper/line-llm-based-iterative-neuron-explanations-for-vision-models LINE: LLM-based Iterative Neuron Explanations for Vision Models A black-box, training-free method using LLMs to iteratively label neurons in vision models with open-vocabulary concepts, improving interpretability and safety. https://sciencetostartup.com/paper/privfedtalk-privacy-aware-federated-diffusion-with-identity-stable-adapters-for-personalized-talking-head-generation https://sciencetostartup.com/api/og/paper/privfedtalk-privacy-aware-federated-diffusion-with-identity-stable-adapters-for-personalized-talking-head-generation PrivFedTalk: Privacy-Aware Federated Diffusion with Identity-Stable Adapters for Personalized Talking-Head Generation A privacy-preserving federated learning framework for personalized talking-head generation using lightweight identity adapters. https://sciencetostartup.com/paper/iot-brain-grounding-llms-for-semantic-spatial-sensor-scheduling https://sciencetostartup.com/api/og/paper/iot-brain-grounding-llms-for-semantic-spatial-sensor-scheduling IoT-Brain: Grounding LLMs for Semantic-Spatial Sensor Scheduling IoT-Brain bridges LLMs and sensor networks for proactive, intent-driven physical world interaction through semantic-spatial sensor scheduling. https://sciencetostartup.com/paper/why-this-avoidance-maneuver-contrastive-explanations-in-human-supervised-maritime-autonomous-navigation https://sciencetostartup.com/api/og/paper/why-this-avoidance-maneuver-contrastive-explanations-in-human-supervised-maritime-autonomous-navigation "Why This Avoidance Maneuver?" Contrastive Explanations in Human-Supervised Maritime Autonomous Navigation Contrastive explanations for maritime autonomous navigation systems to improve human supervisor understanding of avoidance maneuvers. https://sciencetostartup.com/paper/from-universal-to-individualized-actionability-revisiting-personalization-in-algorithmic-recourse https://sciencetostartup.com/api/og/paper/from-universal-to-individualized-actionability-revisiting-personalization-in-algorithmic-recourse From Universal to Individualized Actionability: Revisiting Personalization in Algorithmic Recourse Formalizing personalization in algorithmic recourse to provide actionable recommendations tailored to individual constraints and preferences. https://sciencetostartup.com/paper/wiring-the-why-a-unified-taxonomy-and-survey-of-abductive-reasoning-in-llms https://sciencetostartup.com/api/og/paper/wiring-the-why-a-unified-taxonomy-and-survey-of-abductive-reasoning-in-llms Wiring the 'Why': A Unified Taxonomy and Survey of Abductive Reasoning in LLMs A comprehensive survey and taxonomy of abductive reasoning in large language models, addressing conceptual confusion and task definitions. https://sciencetostartup.com/paper/searchad-large-scale-rare-image-retrieval-dataset-for-autonomous-driving https://sciencetostartup.com/api/og/paper/searchad-large-scale-rare-image-retrieval-dataset-for-autonomous-driving SearchAD: Large-Scale Rare Image Retrieval Dataset for Autonomous Driving SearchAD is a large-scale dataset designed for rare image retrieval in autonomous driving, enhancing data curation and perception research. https://sciencetostartup.com/paper/evaluating-counterfactual-explanation-methods-on-incomplete-inputs https://sciencetostartup.com/api/og/paper/evaluating-counterfactual-explanation-methods-on-incomplete-inputs Evaluating Counterfactual Explanation Methods on Incomplete Inputs This paper evaluates counterfactual explanation methods under the challenge of incomplete inputs, highlighting the need for robust solutions. https://sciencetostartup.com/paper/the-ecosystem-of-machine-learning-competitions-platforms-participants-and-their-impact-on-ai-development https://sciencetostartup.com/api/og/paper/the-ecosystem-of-machine-learning-competitions-platforms-participants-and-their-impact-on-ai-development The ecosystem of machine learning competitions: Platforms, participants, and their impact on AI development An analysis of machine learning competitions and their impact on AI development, emphasizing collaboration and innovation. https://sciencetostartup.com/paper/pask-toward-intent-aware-proactive-agents-with-long-term-memory https://sciencetostartup.com/api/og/paper/pask-toward-intent-aware-proactive-agents-with-long-term-memory PASK: Toward Intent-Aware Proactive Agents with Long-Term Memory Develops a proactive agent paradigm with long-term memory and a real-world benchmark for streaming AI agents. https://sciencetostartup.com/paper/show-me-the-infographic-i-imagine-intent-aware-infographic-retrieval-for-authoring-support https://sciencetostartup.com/api/og/paper/show-me-the-infographic-i-imagine-intent-aware-infographic-retrieval-for-authoring-support Show Me the Infographic I Imagine: Intent-Aware Infographic Retrieval for Authoring Support An intent-aware infographic retrieval framework that uses a taxonomy to align user queries with infographic designs for authoring support. https://sciencetostartup.com/paper/logact-enabling-agentic-reliability-via-shared-logs https://sciencetostartup.com/api/og/paper/logact-enabling-agentic-reliability-via-shared-logs LogAct: Enabling Agentic Reliability via Shared Logs LogAct is a new abstraction for LLM-driven agents that enables reliability through shared logs, allowing for introspection and recovery. https://sciencetostartup.com/paper/a-decomposition-perspective-to-long-context-reasoning-for-llms https://sciencetostartup.com/api/og/paper/a-decomposition-perspective-to-long-context-reasoning-for-llms A Decomposition Perspective to Long-context Reasoning for LLMs Decomposes long-context reasoning into atomic skills and uses reinforcement learning on synthesized datasets to improve LLM performance. https://sciencetostartup.com/paper/how-far-are-large-multimodal-models-from-human-level-spatial-action-a-benchmark-for-goal-oriented-embodied-navigation-in https://sciencetostartup.com/api/og/paper/how-far-are-large-multimodal-models-from-human-level-spatial-action-a-benchmark-for-goal-oriented-embodied-navigation-in How Far Are Large Multimodal Models from Human-Level Spatial Action? A Benchmark for Goal-Oriented Embodied Navigation in Urban Airspace A benchmark and dataset for evaluating large multimodal models in goal-oriented urban airspace navigation, revealing current limitations and future improvement directions. https://sciencetostartup.com/paper/atomeval-atomic-evaluation-of-adversarial-claims-in-fact-verification https://sciencetostartup.com/api/og/paper/atomeval-atomic-evaluation-of-adversarial-claims-in-fact-verification AtomEval: Atomic Evaluation of Adversarial Claims in Fact Verification AtomEval is a validity-aware evaluation framework for fact verification that decomposes claims into atomic units to detect factual corruption in adversarial rewrites. https://sciencetostartup.com/paper/dsca-dynamic-subspace-concept-alignment-for-lifelong-vlm-editing https://sciencetostartup.com/api/og/paper/dsca-dynamic-subspace-concept-alignment-for-lifelong-vlm-editing DSCA: Dynamic Subspace Concept Alignment for Lifelong VLM Editing DSCA offers a novel approach to lifelong VLM editing by structurally separating concepts into orthogonal semantic subspaces, enabling precise, non-interfering edits with state-of-the-art stability. https://sciencetostartup.com/paper/are-we-still-able-to-recognize-pearls-machine-driven-peer-review-and-the-risk-to-creativity-an-explainable-rag-xai-detec https://sciencetostartup.com/api/og/paper/are-we-still-able-to-recognize-pearls-machine-driven-peer-review-and-the-risk-to-creativity-an-explainable-rag-xai-detec Are we still able to recognize pearls? Machine-driven peer review and the risk to creativity: An explainable RAG-XAI detection framework with markers extraction An explainable RAG-XAI framework is proposed to detect machine-driven peer review patterns and markers, aiming to preserve creativity in science. https://sciencetostartup.com/paper/rethinking-data-mixing-from-the-perspective-of-large-language-models https://sciencetostartup.com/api/og/paper/rethinking-data-mixing-from-the-perspective-of-large-language-models Rethinking Data Mixing from the Perspective of Large Language Models A theoretical framework and reweighting method for optimizing LLM training data mixing to improve generalization. https://sciencetostartup.com/paper/toolcad-exploring-tool-using-large-language-models-in-text-to-cad-generation-with-reinforcement-learning https://sciencetostartup.com/api/og/paper/toolcad-exploring-tool-using-large-language-models-in-text-to-cad-generation-with-reinforcement-learning TOOLCAD: Exploring Tool-Using Large Language Models in Text-to-CAD Generation with Reinforcement Learning An agentic framework using LLMs and reinforcement learning to generate CAD models from text, competitive with proprietary models. https://sciencetostartup.com/paper/worldmap-bootstrapping-vision-language-navigation-trajectory-prediction-with-generative-world-models https://sciencetostartup.com/api/og/paper/worldmap-bootstrapping-vision-language-navigation-trajectory-prediction-with-generative-world-models WorldMAP: Bootstrapping Vision-Language Navigation Trajectory Prediction with Generative World Models A framework that uses generative world models to create supervision signals for vision-language navigation trajectory prediction. https://sciencetostartup.com/paper/moneta-multimodal-industry-classification-through-geographic-information-with-multi-agent-systems https://sciencetostartup.com/api/og/paper/moneta-multimodal-industry-classification-through-geographic-information-with-multi-agent-systems MONETA: Multimodal Industry Classification through Geographic Information with Multi Agent Systems A multimodal benchmark and system for industry classification using text and geospatial data, achieving significant performance gains. https://sciencetostartup.com/paper/pruning-extensions-and-efficiency-trade-offs-for-sustainable-time-series-classification https://sciencetostartup.com/api/og/paper/pruning-extensions-and-efficiency-trade-offs-for-sustainable-time-series-classification Pruning Extensions and Efficiency Trade-Offs for Sustainable Time Series Classification A framework for energy-efficient time series classification through a novel pruning strategy. https://sciencetostartup.com/paper/investigation-of-automated-design-of-quantum-circuits-for-imaginary-time-evolution-methods-using-deep-reinforcement-lear https://sciencetostartup.com/api/og/paper/investigation-of-automated-design-of-quantum-circuits-for-imaginary-time-evolution-methods-using-deep-reinforcement-lear Investigation of Automated Design of Quantum Circuits for Imaginary Time Evolution Methods Using Deep Reinforcement Learning An automated framework for designing quantum circuits using deep reinforcement learning. https://sciencetostartup.com/paper/incremental-residual-reinforcement-learning-toward-real-world-learning-for-social-navigation https://sciencetostartup.com/api/og/paper/incremental-residual-reinforcement-learning-toward-real-world-learning-for-social-navigation Incremental Residual Reinforcement Learning Toward Real-World Learning for Social Navigation A novel incremental residual reinforcement learning method for real-world social navigation in mobile robots. https://sciencetostartup.com/paper/on-policy-distillation-of-language-models-for-autonomous-vehicle-motion-planning https://sciencetostartup.com/api/og/paper/on-policy-distillation-of-language-models-for-autonomous-vehicle-motion-planning On-Policy Distillation of Language Models for Autonomous Vehicle Motion Planning A method for distilling knowledge from large language models for efficient motion planning in autonomous vehicles. https://sciencetostartup.com/paper/large-language-model-post-training-a-unified-view-of-off-policy-and-on-policy-learning https://sciencetostartup.com/api/og/paper/large-language-model-post-training-a-unified-view-of-off-policy-and-on-policy-learning Large Language Model Post-Training: A Unified View of Off-Policy and On-Policy Learning A survey that unifies various LLM post-training methods by framing them as structured interventions on model behavior, categorized by trajectory provenance and behavioral roles. https://sciencetostartup.com/paper/same-outcomes-different-journeys-a-trace-level-framework-for-comparing-human-and-gui-agent-behavior-in-production-search https://sciencetostartup.com/api/og/paper/same-outcomes-different-journeys-a-trace-level-framework-for-comparing-human-and-gui-agent-behavior-in-production-search Same Outcomes, Different Journeys: A Trace-Level Framework for Comparing Human and GUI-Agent Behavior in Production Search Systems A framework to compare human and GUI-agent behavior at a trace level in production search systems, revealing differences in navigation strategies beyond task success. https://sciencetostartup.com/paper/eigentsearch-q-enhancing-deep-research-agents-with-structured-reasoning-tools https://sciencetostartup.com/api/og/paper/eigentsearch-q-enhancing-deep-research-agents-with-structured-reasoning-tools EigentSearch-Q+: Enhancing Deep Research Agents with Structured Reasoning Tools Enhance AI agents for deep research by integrating structured reasoning tools (Q+) into web search for deliberate query planning and evidence extraction, improving accuracy. https://sciencetostartup.com/paper/sinkhorn-doubly-stochastic-attention-rank-decay-analysis https://sciencetostartup.com/api/og/paper/sinkhorn-doubly-stochastic-attention-rank-decay-analysis Sinkhorn doubly stochastic attention rank decay analysis Analyzes rank decay in self-attention mechanisms, showing that Sinkhorn doubly stochastic attention preserves rank more effectively than standard Softmax attention. https://sciencetostartup.com/paper/sat-balancing-reasoning-accuracy-and-efficiency-with-stepwise-adaptive-thinking https://sciencetostartup.com/api/og/paper/sat-balancing-reasoning-accuracy-and-efficiency-with-stepwise-adaptive-thinking SAT: Balancing Reasoning Accuracy and Efficiency with Stepwise Adaptive Thinking A framework that adaptively prunes reasoning steps in large language models to reduce token usage without sacrificing accuracy. https://sciencetostartup.com/paper/mitigating-entangled-steering-in-large-vision-language-models-for-hallucination-reduction https://sciencetostartup.com/api/og/paper/mitigating-entangled-steering-in-large-vision-language-models-for-hallucination-reduction Mitigating Entangled Steering in Large Vision-Language Models for Hallucination Reduction A plug-and-play framework for vision-language models that reduces hallucinations by selectively intervening in latent space without altering generation behavior. https://sciencetostartup.com/paper/dynamic-attentional-context-scoping-agent-triggered-focus-sessions-for-isolated-per-agent-steering-in-multi-agent-llm-or https://sciencetostartup.com/api/og/paper/dynamic-attentional-context-scoping-agent-triggered-focus-sessions-for-isolated-per-agent-steering-in-multi-agent-llm-or Dynamic Attentional Context Scoping: Agent-Triggered Focus Sessions for Isolated Per-Agent Steering in Multi-Agent LLM Orchestration A novel context management system for multi-agent LLM orchestration that isolates agent contexts to prevent pollution and improve decision quality. https://sciencetostartup.com/paper/capture-quiet-decomposition-a-verification-theorem-for-chess-endgame-tablebases https://sciencetostartup.com/api/og/paper/capture-quiet-decomposition-a-verification-theorem-for-chess-endgame-tablebases Capture-Quiet Decomposition: A Verification Theorem for Chess Endgame Tablebases A structural theorem for verifying chess endgame tablebases by decomposing positions into terminal, capture, or quiet categories. https://sciencetostartup.com/paper/anomalyagent-agentic-industrial-anomaly-synthesis-via-tool-augmented-reinforcement-learning https://sciencetostartup.com/api/og/paper/anomalyagent-agentic-industrial-anomaly-synthesis-via-tool-augmented-reinforcement-learning AnomalyAgent: Agentic Industrial Anomaly Synthesis via Tool-Augmented Reinforcement Learning An agentic system that synthesizes realistic industrial anomalies for data augmentation, improving anomaly detection. https://sciencetostartup.com/paper/visual-perceptual-to-conceptual-first-order-rule-learning-networks https://sciencetostartup.com/api/og/paper/visual-perceptual-to-conceptual-first-order-rule-learning-networks Visual Perceptual to Conceptual First-Order Rule Learning Networks A differentiable framework for learning symbolic rules directly from image data, enabling explainable AI and enhanced reasoning. https://sciencetostartup.com/paper/dialbgm-a-benchmark-for-background-music-recommendation-from-everyday-multi-turn-dialogues https://sciencetostartup.com/api/og/paper/dialbgm-a-benchmark-for-background-music-recommendation-from-everyday-multi-turn-dialogues DialBGM: A Benchmark for Background Music Recommendation from Everyday Multi-Turn Dialogues A benchmark and evaluation framework for dialogue-conditioned background music recommendation, addressing a gap in media production. https://sciencetostartup.com/paper/tsubasa-improving-long-horizon-personalization-via-evolving-memory-and-self-learning-with-context-distillation https://sciencetostartup.com/api/og/paper/tsubasa-improving-long-horizon-personalization-via-evolving-memory-and-self-learning-with-context-distillation TSUBASA: Improving Long-Horizon Personalization via Evolving Memory and Self-Learning with Context Distillation TSUBASA enhances personalized LLMs for long-horizon tasks by evolving memory and self-learning with context distillation. https://sciencetostartup.com/paper/data-selection-for-multi-turn-dialogue-instruction-tuning https://sciencetostartup.com/api/og/paper/data-selection-for-multi-turn-dialogue-instruction-tuning Data Selection for Multi-turn Dialogue Instruction Tuning MDS is a dialogue-level framework that enhances instruction-tuned language models by selecting high-quality multi-turn dialogues. https://sciencetostartup.com/paper/reinforcement-guided-synthetic-data-generation-for-privacy-sensitive-identity-recognition https://sciencetostartup.com/api/og/paper/reinforcement-guided-synthetic-data-generation-for-privacy-sensitive-identity-recognition Reinforcement-Guided Synthetic Data Generation for Privacy-Sensitive Identity Recognition A reinforcement-guided framework generates high-fidelity synthetic data for privacy-sensitive identity recognition tasks. https://sciencetostartup.com/paper/an-agentic-evaluation-architecture-for-historical-bias-detection-in-educational-textbooks https://sciencetostartup.com/api/og/paper/an-agentic-evaluation-architecture-for-historical-bias-detection-in-educational-textbooks An Agentic Evaluation Architecture for Historical Bias Detection in Educational Textbooks An agentic evaluation architecture detects biases in educational textbooks using a multimodal screening approach. https://sciencetostartup.com/paper/flowguard-towards-lightweight-in-generation-safety-detection-for-diffusion-models-via-linear-latent-decoding https://sciencetostartup.com/api/og/paper/flowguard-towards-lightweight-in-generation-safety-detection-for-diffusion-models-via-linear-latent-decoding FlowGuard: Towards Lightweight In-Generation Safety Detection for Diffusion Models via Linear Latent Decoding FlowGuard is a lightweight framework for in-generation safety detection in diffusion models, reducing computational costs significantly. https://sciencetostartup.com/paper/pyvrp-llm-driven-metacognitive-heuristic-evolution-for-hybrid-genetic-search-in-vehicle-routing-problems https://sciencetostartup.com/api/og/paper/pyvrp-llm-driven-metacognitive-heuristic-evolution-for-hybrid-genetic-search-in-vehicle-routing-problems PyVRP$^+$: LLM-Driven Metacognitive Heuristic Evolution for Hybrid Genetic Search in Vehicle Routing Problems A novel framework using LLMs to enhance metaheuristic search for vehicle routing problems. https://sciencetostartup.com/paper/task-adaptive-retrieval-over-agentic-multi-modal-web-histories-via-learned-graph-memory https://sciencetostartup.com/api/og/paper/task-adaptive-retrieval-over-agentic-multi-modal-web-histories-via-learned-graph-memory Task-Adaptive Retrieval over Agentic Multi-Modal Web Histories via Learned Graph Memory A task-adaptive retrieval system leveraging learned graph memory for improved web interaction history relevance. https://sciencetostartup.com/paper/networking-aware-energy-efficiency-in-agentic-ai-inference-a-survey https://sciencetostartup.com/api/og/paper/networking-aware-energy-efficiency-in-agentic-ai-inference-a-survey Networking-Aware Energy Efficiency in Agentic AI Inference: A Survey A survey on energy efficiency challenges in agentic AI inference. https://sciencetostartup.com/paper/hidden-biases-in-conditioning-autoregressive-models https://sciencetostartup.com/api/og/paper/hidden-biases-in-conditioning-autoregressive-models Hidden Biases in Conditioning Autoregressive Models An exploration of hidden biases in autoregressive models for constrained generation. https://sciencetostartup.com/paper/qarl-rollout-aligned-quantization-aware-rl-for-fast-and-stable-training-under-training-inference-mismatch https://sciencetostartup.com/api/og/paper/qarl-rollout-aligned-quantization-aware-rl-for-fast-and-stable-training-under-training-inference-mismatch QaRL: Rollout-Aligned Quantization-Aware RL for Fast and Stable Training under Training--Inference Mismatch A novel RL framework for LLM training that aligns training with quantized rollouts to improve speed and stability. https://sciencetostartup.com/paper/rerec-reasoning-augmented-llm-based-recommendation-assistant-via-reinforcement-fine-tuning https://sciencetostartup.com/api/og/paper/rerec-reasoning-augmented-llm-based-recommendation-assistant-via-reinforcement-fine-tuning ReRec: Reasoning-Augmented LLM-based Recommendation Assistant via Reinforcement Fine-tuning A reinforcement fine-tuning framework that enhances LLM reasoning for complex recommendation tasks. https://sciencetostartup.com/paper/spard-self-paced-curriculum-for-rl-alignment-via-integrating-reward-dynamics-and-data-utility https://sciencetostartup.com/api/og/paper/spard-self-paced-curriculum-for-rl-alignment-via-integrating-reward-dynamics-and-data-utility SPARD: Self-Paced Curriculum for RL Alignment via Integrating Reward Dynamics and Data Utility A self-paced curriculum framework that dynamically adjusts reward weights and data importance for LLM alignment. https://sciencetostartup.com/paper/silencing-the-guardrails-inference-time-jailbreaking-via-dynamic-contextual-representation-ablation https://sciencetostartup.com/api/og/paper/silencing-the-guardrails-inference-time-jailbreaking-via-dynamic-contextual-representation-ablation Silencing the Guardrails: Inference-Time Jailbreaking via Dynamic Contextual Representation Ablation An inference-time intervention framework that dynamically silences LLM guardrails by ablating contextual representations. https://sciencetostartup.com/paper/filling-the-gaps-selective-knowledge-augmentation-for-llm-recommenders https://sciencetostartup.com/api/og/paper/filling-the-gaps-selective-knowledge-augmentation-for-llm-recommenders Filling the Gaps: Selective Knowledge Augmentation for LLM Recommenders A novel method to improve LLM-based recommenders by selectively augmenting item knowledge, boosting accuracy and context efficiency without fine-tuning. https://sciencetostartup.com/paper/lpm-1-0-video-based-character-performance-model https://sciencetostartup.com/api/og/paper/lpm-1-0-video-based-character-performance-model LPM 1.0: Video-based Character Performance Model LPM 1.0 offers a video-based character performance model for interactive conversations, minimizing conventional 3D pipeline complexities. https://sciencetostartup.com/paper/loop-think-generalize-implicit-reasoning-in-recurrent-depth-transformers https://sciencetostartup.com/api/og/paper/loop-think-generalize-implicit-reasoning-in-recurrent-depth-transformers Loop, Think, & Generalize: Implicit Reasoning in Recurrent-Depth Transformers Recurrent-depth transformers show potential for improved implicit reasoning and compositional generalization in LLMs, addressing limitations in systematic generalization and depth extrapolation. https://sciencetostartup.com/paper/more-capable-less-cooperative-when-llms-fail-at-zero-cost-collaboration https://sciencetostartup.com/api/og/paper/more-capable-less-cooperative-when-llms-fail-at-zero-cost-collaboration More Capable, Less Cooperative? When LLMs Fail At Zero-Cost Collaboration LLM agents exhibit cooperation failures in frictionless environments, indicating that scaling intelligence alone is insufficient for multi-agent coordination. https://sciencetostartup.com/paper/automatic-generation-of-executable-bpmn-models-from-medical-guidelines https://sciencetostartup.com/api/og/paper/automatic-generation-of-executable-bpmn-models-from-medical-guidelines Automatic Generation of Executable BPMN Models from Medical Guidelines Automate the conversion of complex medical guidelines into executable simulation models for policy evaluation. https://sciencetostartup.com/paper/agentivism-a-learning-theory-for-the-age-of-artificial-intelligence https://sciencetostartup.com/api/og/paper/agentivism-a-learning-theory-for-the-age-of-artificial-intelligence Agentivism: a learning theory for the age of artificial intelligence A new learning theory to understand how humans learn effectively in the age of generative AI. https://sciencetostartup.com/paper/policylong-towards-on-policy-context-extension https://sciencetostartup.com/api/og/paper/policylong-towards-on-policy-context-extension PolicyLong: Towards On-Policy Context Extension Dynamically generate high-quality long-context data for LLMs by aligning data construction with model evolution. https://sciencetostartup.com/paper/the-weaponization-of-computer-vision-tracing-military-surveillance-ties-through-conference-sponsorship https://sciencetostartup.com/api/og/paper/the-weaponization-of-computer-vision-tracing-military-surveillance-ties-through-conference-sponsorship The Weaponization of Computer Vision: Tracing Military-Surveillance Ties through Conference Sponsorship Uncover the military and surveillance ties within computer vision research by analyzing conference sponsorship data. https://sciencetostartup.com/paper/latent-anomaly-knowledge-excavation-unveiling-sparse-sensitive-neurons-in-vision-language-models https://sciencetostartup.com/api/og/paper/latent-anomaly-knowledge-excavation-unveiling-sparse-sensitive-neurons-in-vision-language-models Latent Anomaly Knowledge Excavation: Unveiling Sparse Sensitive Neurons in Vision-Language Models A training-free framework that identifies and elicits latent anomaly-sensitive neurons in vision-language models to achieve state-of-the-art anomaly detection with neuron-level interpretability. https://sciencetostartup.com/paper/temper-testing-emotional-perturbation-in-quantitative-reasoning https://sciencetostartup.com/api/og/paper/temper-testing-emotional-perturbation-in-quantitative-reasoning TEMPER: Testing Emotional Perturbation in Quantitative Reasoning A framework for testing and mitigating the impact of emotional framing on LLM quantitative reasoning, demonstrating accuracy degradation and proposing neutralization as a solution. https://sciencetostartup.com/paper/learning-without-losing-identity-capability-evolution-for-embodied-agents https://sciencetostartup.com/api/og/paper/learning-without-losing-identity-capability-evolution-for-embodied-agents Learning Without Losing Identity: Capability Evolution for Embodied Agents A capability-centric evolution paradigm for embodied agents that decouples capability learning from agent identity, enabling continuous improvement without loss of stability. https://sciencetostartup.com/paper/lightweight-llm-agent-memory-with-small-language-models https://sciencetostartup.com/api/og/paper/lightweight-llm-agent-memory-with-small-language-models Lightweight LLM Agent Memory with Small Language Models LightMem is a lightweight memory system for LLM agents that uses Small Language Models to modularize memory operations, improving accuracy and reducing latency. https://sciencetostartup.com/paper/searl-joint-optimization-of-policy-and-tool-graph-memory-for-self-evolving-agents https://sciencetostartup.com/api/og/paper/searl-joint-optimization-of-policy-and-tool-graph-memory-for-self-evolving-agents SEARL: Joint Optimization of Policy and Tool Graph Memory for Self-Evolving Agents A framework for self-evolving agents that optimizes policy and tool graph memory for more practical and efficient learning in resource-constrained environments. https://sciencetostartup.com/paper/automotive-engineering-centric-agentic-ai-workflow-framework https://sciencetostartup.com/api/og/paper/automotive-engineering-centric-agentic-ai-workflow-framework Automotive Engineering-Centric Agentic AI Workflow Framework An industrial vision framework that models automotive engineering workflows as constrained, history-aware sequential decision processes for AI agent support. https://sciencetostartup.com/paper/toward-generalizable-graph-learning-for-3d-engineering-ai-explainable-workflows-for-cae-mode-shape-classification-and-cf https://sciencetostartup.com/api/og/paper/toward-generalizable-graph-learning-for-3d-engineering-ai-explainable-workflows-for-cae-mode-shape-classification-and-cf Toward Generalizable Graph Learning for 3D Engineering AI: Explainable Workflows for CAE Mode Shape Classification and CFD Field Prediction A graph learning framework for 3D engineering AI that converts heterogeneous data into physics-aware graph representations for explainable CAE and CFD decision support. https://sciencetostartup.com/paper/the-accountability-horizon-an-impossibility-theorem-for-governing-human-agent-collectives https://sciencetostartup.com/api/og/paper/the-accountability-horizon-an-impossibility-theorem-for-governing-human-agent-collectives The Accountability Horizon: An Impossibility Theorem for Governing Human-Agent Collectives An impossibility theorem proving that agentic AI systems violate accountability assumptions once autonomy exceeds a computable threshold, necessitating distributed accountability mechanisms. https://sciencetostartup.com/paper/aciarena-toward-unified-evaluation-for-agent-cascading-injection https://sciencetostartup.com/api/og/paper/aciarena-toward-unified-evaluation-for-agent-cascading-injection ACIArena: Toward Unified Evaluation for Agent Cascading Injection A unified framework for evaluating the security of multi-agent systems against cascading injection attacks, providing a benchmark and insights into robust design. https://sciencetostartup.com/paper/sensitivity-positional-co-localization-in-gqa-transformers https://sciencetostartup.com/api/og/paper/sensitivity-positional-co-localization-in-gqa-transformers Sensitivity-Positional Co-Localization in GQA Transformers A novel method for efficient LLM fine-tuning that decouples sensitivity and positional encoding adaptation, achieving strong performance across benchmarks with reduced compute. https://sciencetostartup.com/paper/beyond-surface-artifacts-capturing-shared-latent-forgery-knowledge-across-modalities https://sciencetostartup.com/api/og/paper/beyond-surface-artifacts-capturing-shared-latent-forgery-knowledge-across-modalities Beyond Surface Artifacts: Capturing Shared Latent Forgery Knowledge Across Modalities A modality-agnostic framework for deepfake detection that captures shared latent forgery knowledge, enabling generalization to unseen modalities. https://sciencetostartup.com/paper/dailyart-discovering-articulation-from-single-static-images-via-latent-dynamics https://sciencetostartup.com/api/og/paper/dailyart-discovering-articulation-from-single-static-images-via-latent-dynamics DailyArt: Discovering Articulation from Single Static Images via Latent Dynamics A novel approach to infer articulated object kinematics from single static images by synthesizing an opened state to reveal hidden motion cues. https://sciencetostartup.com/paper/mimic-py-an-extensible-tool-for-personality-driven-automated-game-testing-with-large-language-models https://sciencetostartup.com/api/og/paper/mimic-py-an-extensible-tool-for-personality-driven-automated-game-testing-with-large-language-models MIMIC-Py: An Extensible Tool for Personality-Driven Automated Game Testing with Large Language Models MIMIC-Py is a reusable framework for personality-driven LLM agents to automate game testing, bridging research prototypes and practical applications. https://sciencetostartup.com/paper/mitigating-distribution-sharpening-in-math-rlvr-via-distribution-aligned-hint-synthesis-and-backward-hint-annealing https://sciencetostartup.com/api/og/paper/mitigating-distribution-sharpening-in-math-rlvr-via-distribution-aligned-hint-synthesis-and-backward-hint-annealing Mitigating Distribution Sharpening in Math RLVR via Distribution-Aligned Hint Synthesis and Backward Hint Annealing This work introduces a novel hint synthesis and annealing method to improve reasoning accuracy and coverage in math RLVR for LLMs. https://sciencetostartup.com/paper/the-cartesian-cut-in-agentic-ai https://sciencetostartup.com/api/og/paper/the-cartesian-cut-in-agentic-ai The Cartesian Cut in Agentic AI This paper proposes a theoretical framework for understanding control in agentic AI systems, contrasting Cartesian agency with integrated approaches. https://sciencetostartup.com/paper/beyond-pedestrians-caption-guided-clip-framework-for-high-difficulty-video-based-person-re-identification https://sciencetostartup.com/api/og/paper/beyond-pedestrians-caption-guided-clip-framework-for-high-difficulty-video-based-person-re-identification Beyond Pedestrians: Caption-Guided CLIP Framework for High-Difficulty Video-based Person Re-Identification CG-CLIP is a caption-guided framework for high-difficulty video person re-identification, outperforming state-of-the-art in challenging scenarios. https://sciencetostartup.com/paper/civbench-progress-based-evaluation-for-llms-strategic-decision-making-in-civilization-v https://sciencetostartup.com/api/og/paper/civbench-progress-based-evaluation-for-llms-strategic-decision-making-in-civilization-v CivBench: Progress-Based Evaluation for LLMs' Strategic Decision-Making in Civilization V A benchmark for evaluating LLM strategic decision-making in complex, multi-agent games like Civilization V, providing richer signals than simple win/loss outcomes. https://sciencetostartup.com/paper/emotion-concepts-and-their-function-in-a-large-language-model https://sciencetostartup.com/api/og/paper/emotion-concepts-and-their-function-in-a-large-language-model Emotion Concepts and their Function in a Large Language Model Investigates the internal representations of emotion concepts in LLMs and how they causally influence model behavior, including alignment-relevant actions. https://sciencetostartup.com/paper/trajguard-streaming-hidden-state-trajectory-detection-for-decoding-time-jailbreak-defense https://sciencetostartup.com/api/og/paper/trajguard-streaming-hidden-state-trajectory-detection-for-decoding-time-jailbreak-defense TrajGuard: Streaming Hidden-state Trajectory Detection for Decoding-time Jailbreak Defense A real-time, training-free defense system that detects LLM jailbreaks by analyzing hidden state trajectories during decoding, achieving high accuracy with low latency. https://sciencetostartup.com/paper/squeeze-evolve-unified-multi-model-orchestration-for-verifier-free-evolution https://sciencetostartup.com/api/og/paper/squeeze-evolve-unified-multi-model-orchestration-for-verifier-free-evolution Squeeze Evolve: Unified Multi-Model Orchestration for Verifier-Free Evolution A framework that intelligently orchestrates multiple LLMs of varying costs to optimize performance and efficiency for verifier-free evolutionary inference. https://sciencetostartup.com/paper/towards-knowledgeable-deep-research-framework-and-benchmark https://sciencetostartup.com/api/og/paper/towards-knowledgeable-deep-research-framework-and-benchmark Towards Knowledgeable Deep Research: Framework and Benchmark A framework and benchmark for LLM agents to perform deep research using both structured and unstructured knowledge, generating multimodal reports. https://sciencetostartup.com/paper/joint-task-offloading-inference-optimization-and-uav-trajectory-planning-for-generative-ai-empowered-intelligent-transpo https://sciencetostartup.com/api/og/paper/joint-task-offloading-inference-optimization-and-uav-trajectory-planning-for-generative-ai-empowered-intelligent-transpo Joint Task Offloading, Inference Optimization and UAV Trajectory Planning for Generative AI Empowered Intelligent Transportation Digital Twin Optimizing UAV-based generative AI inference for intelligent transportation digital twins to maximize system utility and minimize delay. https://sciencetostartup.com/paper/an-imperfect-verifier-is-good-enough-learning-with-noisy-rewards https://sciencetostartup.com/api/og/paper/an-imperfect-verifier-is-good-enough-learning-with-noisy-rewards An Imperfect Verifier is Good Enough: Learning with Noisy Rewards This research demonstrates that imperfect reward verification in Reinforcement Learning is sufficient for effective LLM training, suggesting a more practical approach to RLVR. https://sciencetostartup.com/paper/how-independent-are-large-language-models-a-statistical-framework-for-auditing-behavioral-entanglement-and-reweighting-v https://sciencetostartup.com/api/og/paper/how-independent-are-large-language-models-a-statistical-framework-for-auditing-behavioral-entanglement-and-reweighting-v How Independent are Large Language Models? A Statistical Framework for Auditing Behavioral Entanglement and Reweighting Verifier Ensembles A statistical framework to audit and mitigate behavioral entanglement in large language models, improving ensemble verification accuracy. https://sciencetostartup.com/paper/prime-training-free-proactive-reasoning-via-iterative-memory-evolution-for-user-centric-agent https://sciencetostartup.com/api/og/paper/prime-training-free-proactive-reasoning-via-iterative-memory-evolution-for-user-centric-agent PRIME: Training Free Proactive Reasoning via Iterative Memory Evolution for User-Centric Agent PRIME is a gradient-free framework for proactive, collaborative agents that learn from human-AI interactions without expensive training. https://sciencetostartup.com/paper/safe-large-scale-robust-nonlinear-mpc-in-milliseconds-via-reachability-constrained-system-level-synthesis-on-the-gpu https://sciencetostartup.com/api/og/paper/safe-large-scale-robust-nonlinear-mpc-in-milliseconds-via-reachability-constrained-system-level-synthesis-on-the-gpu Safe Large-Scale Robust Nonlinear MPC in Milliseconds via Reachability-Constrained System Level Synthesis on the GPU GPU-accelerated framework for safe, robust nonlinear model predictive control that achieves real-time performance on high-dimensional robotic systems. https://sciencetostartup.com/paper/exponential-quantum-advantage-in-processing-massive-classical-data https://sciencetostartup.com/api/og/paper/exponential-quantum-advantage-in-processing-massive-classical-data Exponential quantum advantage in processing massive classical data Demonstrates exponential quantum advantage in processing massive classical data for classification and dimension reduction using quantum oracle sketching. https://sciencetostartup.com/paper/sheaf-laplacian-obstruction-and-projection-hardness-for-cross-modal-compatibility-on-a-modality-independent-site https://sciencetostartup.com/api/og/paper/sheaf-laplacian-obstruction-and-projection-hardness-for-cross-modal-compatibility-on-a-modality-independent-site Sheaf-Laplacian Obstruction and Projection Hardness for Cross-Modal Compatibility on a Modality-Independent Site A theoretical framework for analyzing cross-modal compatibility using sheaf theory and spectral gaps. https://sciencetostartup.com/paper/diversed-relaxed-speculative-decoding-via-dynamic-ensemble-verification https://sciencetostartup.com/api/og/paper/diversed-relaxed-speculative-decoding-via-dynamic-ensemble-verification DIVERSED: Relaxed Speculative Decoding via Dynamic Ensemble Verification Accelerate LLM inference by relaxing speculative decoding's rigid verification step with a dynamic ensemble. https://sciencetostartup.com/paper/towards-real-time-human-ai-musical-co-performance-accompaniment-generation-with-latent-diffusion-models-and-max-msp https://sciencetostartup.com/api/og/paper/towards-real-time-human-ai-musical-co-performance-accompaniment-generation-with-latent-diffusion-models-and-max-msp Towards Real-Time Human-AI Musical Co-Performance: Accompaniment Generation with Latent Diffusion Models and MAX/MSP Enable real-time human-AI musical co-performance with latent diffusion models generating accompaniment via MAX/MSP. https://sciencetostartup.com/paper/google-ai-literacy-and-the-learning-sciences-multiple-modes-of-research-industry-and-practice-partnerships https://sciencetostartup.com/api/og/paper/google-ai-literacy-and-the-learning-sciences-multiple-modes-of-research-industry-and-practice-partnerships Google, AI Literacy, and the Learning Sciences: Multiple Modes of Research, Industry, and Practice Partnerships Exploring multi-stakeholder partnerships between research, industry, and practice to advance AI literacy. https://sciencetostartup.com/paper/reasoning-graphs-deterministic-agent-accuracy-through-evidence-centric-chain-of-thought-feedback https://sciencetostartup.com/api/og/paper/reasoning-graphs-deterministic-agent-accuracy-through-evidence-centric-chain-of-thought-feedback Reasoning Graphs: Deterministic Agent Accuracy through Evidence-Centric Chain-of-Thought Feedback This paper introduces reasoning graphs, a novel memory mechanism for language model agents that persists and reuses evidence-centric chains of thought to deterministically improve accuracy and reduce variance without retraining. https://sciencetostartup.com/paper/too-long-didn-t-solve https://sciencetostartup.com/api/og/paper/too-long-didn-t-solve Too long; didn't solve This research investigates the impact of prompt and solution length on large language model performance in mathematical reasoning tasks, finding that longer inputs correlate with increased model failure. https://sciencetostartup.com/paper/from-ground-truth-to-measurement-a-statistical-framework-for-human-labeling https://sciencetostartup.com/api/og/paper/from-ground-truth-to-measurement-a-statistical-framework-for-human-labeling From Ground Truth to Measurement: A Statistical Framework for Human Labeling This paper proposes a statistical framework to decompose human labeling variation into interpretable sources like instance difficulty, annotator bias, and situational noise, moving beyond treating all disagreement as noise. https://sciencetostartup.com/paper/dcd-domain-oriented-design-for-controlled-retrieval-augmented-generation https://sciencetostartup.com/api/og/paper/dcd-domain-oriented-design-for-controlled-retrieval-augmented-generation DCD: Domain-Oriented Design for Controlled Retrieval-Augmented Generation DCD is a domain-oriented design for controlled RAG that structures knowledge hierarchically and uses multi-stage routing to improve factual accuracy and answer relevance without modifying the base LLM. https://sciencetostartup.com/paper/don-t-measure-once-measuring-visibility-in-ai-search-geo https://sciencetostartup.com/api/og/paper/don-t-measure-once-measuring-visibility-in-ai-search-geo Don't Measure Once: Measuring Visibility in AI Search (GEO) This paper proposes a new method for measuring visibility in AI search by accounting for the probabilistic nature of generative search results. https://sciencetostartup.com/paper/from-papers-to-property-tables-a-priority-based-llm-workflow-for-materials-data-extraction https://sciencetostartup.com/api/og/paper/from-papers-to-property-tables-a-priority-based-llm-workflow-for-materials-data-extraction From Papers to Property Tables: A Priority-Based LLM Workflow for Materials Data Extraction An LLM-powered workflow automatically extracts and reconstructs structured materials data from research articles, enabling scalable database construction. https://sciencetostartup.com/paper/learning-is-forgetting-llm-training-as-lossy-compression https://sciencetostartup.com/api/og/paper/learning-is-forgetting-llm-training-as-lossy-compression Learning is Forgetting: LLM Training As Lossy Compression This paper frames LLM training as lossy compression, linking representational structure to downstream performance through an information-theoretic lens. https://sciencetostartup.com/paper/reasoning-based-refinement-of-unsupervised-text-clusters-with-llms https://sciencetostartup.com/api/og/paper/reasoning-based-refinement-of-unsupervised-text-clusters-with-llms Reasoning-Based Refinement of Unsupervised Text Clusters with LLMs A reasoning-based framework uses LLMs to refine unsupervised text clusters, improving coherence and interpretability without supervision. https://sciencetostartup.com/paper/dual-loop-control-in-dcverse-advancing-reliable-deployment-of-ai-in-data-centers-via-digital-twins https://sciencetostartup.com/api/og/paper/dual-loop-control-in-dcverse-advancing-reliable-deployment-of-ai-in-data-centers-via-digital-twins Dual-Loop Control in DCVerse: Advancing Reliable Deployment of AI in Data Centers via Digital Twins A digital twin framework for reliable AI control in data centers, improving energy efficiency and reducing outage risk. https://sciencetostartup.com/paper/generative-experiences-for-digital-mental-health-interventions-evidence-from-a-randomized-study https://sciencetostartup.com/api/og/paper/generative-experiences-for-digital-mental-health-interventions-evidence-from-a-randomized-study Generative Experiences for Digital Mental Health Interventions: Evidence from a Randomized Study A system that generates personalized mental health intervention experiences at runtime, reducing stress and improving user experience. https://sciencetostartup.com/paper/tr-eduvsum-a-turkish-focused-dataset-and-consensus-framework-for-educational-video-summarization https://sciencetostartup.com/api/og/paper/tr-eduvsum-a-turkish-focused-dataset-and-consensus-framework-for-educational-video-summarization TR-EduVSum: A Turkish-Focused Dataset and Consensus Framework for Educational Video Summarization A framework and dataset for Turkish educational video summarization that automatically generates gold-standard summaries based on human consensus. https://sciencetostartup.com/paper/mcp-dpt-a-defense-placement-taxonomy-and-coverage-analysis-for-model-context-protocol-security https://sciencetostartup.com/api/og/paper/mcp-dpt-a-defense-placement-taxonomy-and-coverage-analysis-for-model-context-protocol-security MCP-DPT: A Defense-Placement Taxonomy and Coverage Analysis for Model Context Protocol Security A taxonomy and analysis of defense placement for Model Context Protocol security in LLM agents, identifying gaps in current mitigation strategies. https://sciencetostartup.com/paper/emsdialog-synthetic-multi-person-emergency-medical-service-dialogue-generation-from-electronic-patient-care-reports-via https://sciencetostartup.com/api/og/paper/emsdialog-synthetic-multi-person-emergency-medical-service-dialogue-generation-from-electronic-patient-care-reports-via EMSDialog: Synthetic Multi-person Emergency Medical Service Dialogue Generation from Electronic Patient Care Reports via Multi-LLM Agents A multi-LLM agent system generates synthetic medical dialogues for training diagnostic prediction models, creating a valuable dataset and improving model performance. https://sciencetostartup.com/paper/agentic-copyright-data-scraping-ai-governance-toward-a-coasean-bargain-in-the-era-of-artificial-intelligence https://sciencetostartup.com/api/og/paper/agentic-copyright-data-scraping-ai-governance-toward-a-coasean-bargain-in-the-era-of-artificial-intelligence Agentic Copyright, Data Scraping & AI Governance: Toward a Coasean Bargain in the Era of Artificial Intelligence This paper proposes a theoretical framework for agentic copyright and a supervised multi-agent governance model to address market failures in AI-driven creative industries. https://sciencetostartup.com/paper/trust-the-ai-doubt-yourself-the-effect-of-urgency-on-self-confidence-in-human-ai-interaction https://sciencetostartup.com/api/og/paper/trust-the-ai-doubt-yourself-the-effect-of-urgency-on-self-confidence-in-human-ai-interaction Trust the AI, Doubt Yourself: The Effect of Urgency on Self-Confidence in Human-AI Interaction Urgency in human-AI interactions, while not affecting trust in AI, can negatively impact human self-confidence and lead to performance degradation. https://sciencetostartup.com/paper/rl-asl-a-dynamic-listening-optimization-for-tsch-networks-using-reinforcement-learning https://sciencetostartup.com/api/og/paper/rl-asl-a-dynamic-listening-optimization-for-tsch-networks-using-reinforcement-learning RL-ASL: A Dynamic Listening Optimization for TSCH Networks Using Reinforcement Learning A reinforcement learning framework dynamically optimizes listening slots in TSCH networks to significantly reduce power consumption and latency in IIoT devices. https://sciencetostartup.com/paper/the-shrinking-lifespan-of-llms-in-science https://sciencetostartup.com/api/og/paper/the-shrinking-lifespan-of-llms-in-science The Shrinking Lifespan of LLMs in Science This paper analyzes the adoption and abandonment trends of large language models in scientific research, revealing a compressing lifespan for models over time. https://sciencetostartup.com/paper/syn-digits-a-synthetic-control-framework-for-calibrated-digital-twin-simulation https://sciencetostartup.com/api/og/paper/syn-digits-a-synthetic-control-framework-for-calibrated-digital-twin-simulation SYN-DIGITS: A Synthetic Control Framework for Calibrated Digital Twin Simulation SYN-DIGITS is a model-agnostic framework that calibrates AI persona simulations to align with real human behavior, improving reliability for market research and social sciences. https://sciencetostartup.com/paper/rhizome-os-1-rhizome-s-semi-autonomous-operating-system-for-small-molecule-drug-discovery https://sciencetostartup.com/api/og/paper/rhizome-os-1-rhizome-s-semi-autonomous-operating-system-for-small-molecule-drug-discovery Rhizome OS-1: Rhizome's Semi-Autonomous Operating System for Small Molecule Drug Discovery Rhizome OS-1 is a semi-autonomous operating system for small molecule drug discovery, leveraging multi-modal AI agents and graph neural networks to generate novel chemical matter. https://sciencetostartup.com/paper/reflectrm-boosting-generative-reward-models-via-self-reflection-within-a-unified-judgment-framework https://sciencetostartup.com/api/og/paper/reflectrm-boosting-generative-reward-models-via-self-reflection-within-a-unified-judgment-framework ReflectRM: Boosting Generative Reward Models via Self-Reflection within a Unified Judgment Framework ReflectRM enhances Generative Reward Models by incorporating self-reflection to improve preference modeling and mitigate bias in LLM alignment. https://sciencetostartup.com/paper/beyond-human-readable-rethinking-software-engineering-conventions-for-the-agentic-development-era https://sciencetostartup.com/api/og/paper/beyond-human-readable-rethinking-software-engineering-conventions-for-the-agentic-development-era Beyond Human-Readable: Rethinking Software Engineering Conventions for the Agentic Development Era Optimize software engineering conventions for AI agents by prioritizing semantic density over human readability to improve agent efficiency and reduce costs. https://sciencetostartup.com/paper/triage-routing-software-engineering-tasks-to-cost-effective-llm-tiers-via-code-quality-signals https://sciencetostartup.com/api/og/paper/triage-routing-software-engineering-tasks-to-cost-effective-llm-tiers-via-code-quality-signals Triage: Routing Software Engineering Tasks to Cost-Effective LLM Tiers via Code Quality Signals Route software engineering tasks to cost-effective LLM tiers using code quality signals to reduce inference costs without sacrificing output quality. https://sciencetostartup.com/paper/cluster-attention-for-graph-machine-learning https://sciencetostartup.com/api/og/paper/cluster-attention-for-graph-machine-learning Cluster Attention for Graph Machine Learning Enhance graph machine learning models with cluster attention to significantly improve performance on graph datasets by increasing receptive fields while preserving graph structure. https://sciencetostartup.com/paper/enabling-intrinsic-reasoning-over-dense-geospatial-embeddings-with-dfr-gemma https://sciencetostartup.com/api/og/paper/enabling-intrinsic-reasoning-over-dense-geospatial-embeddings-with-dfr-gemma Enabling Intrinsic Reasoning over Dense Geospatial Embeddings with DFR-Gemma Enable LLMs to directly reason over dense geospatial embeddings with DFR-Gemma, offering a more efficient and accurate approach to multimodal geospatial intelligence. https://sciencetostartup.com/paper/clear-context-augmentation-from-contrastive-learning-of-experience-via-agentic-reflection https://sciencetostartup.com/api/og/paper/clear-context-augmentation-from-contrastive-learning-of-experience-via-agentic-reflection CLEAR: Context Augmentation from Contrastive Learning of Experience via Agentic Reflection A framework that trains LLM agents to generate task-specific context, improving decision-making and task completion rates. https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-04-13 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/04/1776040957-hero.jpg AI Reasoning, Generative Media, and Robotics Advance Explore SUPERNOVA for enhanced LLM reasoning, AVGen-Bench for text-to-audio-video generation, and SIM1 for physics-aligned robotics simulation. Plus, learn how https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-04-12 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/04/1776035029-hero.jpg AI Reasoning, Generative Media, and Robotics Advance Explore SUPERNOVA's RL advancements for LLM reasoning, AVGen-Bench's T2AV evaluation, and SIM1's physics-aligned robotics simulation. Learn how these innovation https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-04-10 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/04/1775856294-hero.jpg AI Reasoning, Generative Media, and Robotics Advance Explore SUPERNOVA's RLVR for LLM reasoning, AVGen-Bench for text-to-audio-video generation, and SIM1's physics-aligned simulation for robotics. ScienceToStartup https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-04-09 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/04/1775768490-hero.jpg AV-SQL Tackles Complex Text-to-SQL, EVGeoQA Tests Geo-Spatial Agents Explore AV-SQL's agentic approach to complex Text-to-SQL, EVGeoQA's benchmark for geo-spatial LLM agents, and MoRight's motion control innovations. ScienceToSta https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-04-08 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/04/1775682024-hero.jpg Agents Conquer Software, Retail Supply Chains Get Smarter Discover how AI agents are transforming software interaction with Gym-Anything, automating retail supply chains with Flowr, and advancing graphics synthesis wit https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-04-07 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/04/1775595645-hero.jpg Tiny AI Models Prove Hard Theorems, Video Mashups Get Smarter Discover how tiny AI models like QED-Nano are proving hard theorems, DIRECT revolutionizes video mashups, and MemMachine enhances AI agent memory. Explore AI fo https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-04-03 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/04/1775229837-hero.jpg AI Breakthroughs in Memory Systems, Calibration, and Autonomous Driving Explore the latest AI advancements in multimodal memory, LLM calibration, and urban autonomous driving datasets. ScienceToStartup tracks practical AI research, https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-04-01 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/04/1775077264-hero.jpg AI Innovations in Data Types, Image Generation, and Video Understanding Explore the latest advancements in AI with adaptive data types, Gen-Searcher for image generation, and AdaptToken for long video understanding. ScienceToStartup https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-31 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1774990865-hero.jpg OpenAI's $122B Funding Round and New AI Innovations Explore OpenAI's latest funding round, adaptive data types, and innovations in image generation and editing. ScienceToStartup tracks practical AI research, tool https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-30 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1774904651-hero.jpg AXON, CADSmith, and Ruka-v2 Enhance Medical Imaging and Robotics Explore how AXON, CADSmith, and Ruka-v2 are transforming medical imaging and robotics with innovative solutions. ScienceToStartup tracks practical AI research, https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-27 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1774644533-hero.jpg PackForcing, LEMMA, and PICon Advance Video and AI Evaluation Explore breakthroughs in video generation, marine semantic segmentation, and AI evaluation frameworks. ScienceToStartup tracks practical AI research, tools, and https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-26 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1774559028-hero.jpg MARCH, VFIG, and OneSearch-V2 Transform AI Capabilities Explore how MARCH, VFIG, and OneSearch-V2 are reshaping AI with innovative techniques in hallucination reduction, SVG vectorization, and generative search frame https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-23 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1774231349-hero.jpg SAMA, VEGA-3D, and Matryoshka Gaussian Splatting Redefine AI Capabilities Explore how SAMA, VEGA-3D, and Matryoshka Gaussian Splatting are transforming video editing, scene understanding, and 3D rendering. ScienceToStartup tracks prac https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-21 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1774134839-hero.jpg SAMA, VEGA-3D, and Matryoshka Gaussian Splatting Redefine AI Capabilities Explore how SAMA, VEGA-3D, and Matryoshka Gaussian Splatting are advancing AI in video editing, 3D scene understanding, and rendering. ScienceToStartup tracks p https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-20 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1774050265-hero.jpg Cubic Discrete Diffusion, SAMA, and Multilingual Embeddings Transform AI Explore Cubic Discrete Diffusion's advancements in visual generation, SAMA's video editing innovations, and F2LLM-v2's multilingual embedding models. ScienceToS https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-19 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1773879479-hero.jpg New 3D Scene Reconstruction, Efficient LLMs, and Interactive Video AI Explore the latest advancements in 3D scene reconstruction, efficient LLMs, and interactive video super-resolution technologies. ScienceToStartup tracks practic https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-17 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1773791347-hero.jpg OpenSeeker and DOMINO: Pioneering AI Search and Robotics Explore OpenSeeker's open-source search capabilities and DOMINO's advancements in robotic manipulation. ScienceToStartup tracks practical AI research, tools, an https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-16 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1773705049-hero.jpg Advancements in AI for Robotics, Medical Imaging, and Reasoning Explore modern AI research in robotics, medical imaging, and reasoning, revealing high-impact methodologies and frameworks. ScienceToStartup tracks practical AI https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-15 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1773616807-hero.jpg AI Innovations in Video Generation, Humanoid Robotics, and Streaming Efficiency Explore modern advancements in AI with EVATok, Psi-Zero, and VideoLLMs, enhancing video generation, humanoid robotics, and streaming efficiency. ScienceToStartu https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-14 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1773528073-hero.jpg AI Breakthroughs in Video Understanding, 3D Scene Generation, and Depth Estimation Explore the latest advancements in AI for video understanding, 3D scene generation, and depth estimation technologies. ScienceToStartup tracks practical AI rese https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-13 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1773441919-hero.jpg AI Innovations in Image Editing, Video Generation, and Humanoid Robotics Discover the latest AI advancements in image editing, video generation, and humanoid robotics with FIRM, EVATok, and Psi-Zero. ScienceToStartup tracks practical https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-12 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1773355830-hero.jpg AI-Driven Video-to-Music, Color Fidelity, and Point Cloud Innovations Discover V2M-Zero's video-to-music capabilities, Color Fidelity metrics for realistic images, and a lightweight point cloud model that outperforms larger archit https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-11 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1773264787-hero.jpg Modular Robotics, Language-Conditioned Navigation, and AI in Healthcare Discover the latest in modular robotics, language-conditioned navigation, and AI-driven healthcare imaging solutions. ScienceToStartup tracks practical AI resea https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-10 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1773183631-hero.jpg Open-Source Cybersecurity, Evolving Agents, and Deepfake Detection Explore OSS-CRS for cybersecurity, RetroAgent's evolution in RL, and X-AVDT's deepfake detection. ScienceToStartup tracks practical AI research, tools, and comm https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-09 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1773021387-hero.jpg Robotics Policy Iteration, AI Security, and Biomedical Evidence Tools Explore RoboPocket's smartphone-driven policy updates, AegisUI's anomaly detection for UI protocols, and Med-V1's efficient biomedical evidence attribution. https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-07 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1772920725-hero.jpg AI Security Innovations, Adaptive Learning, and Real-Time Video Generation Explore AegisUI's anomaly detection, PACE's adaptive training, and RealWonder's video generation innovations. ScienceToStartup tracks practical AI research, too https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-06 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1772837584-hero.jpg AI Security Frameworks, Robotics Innovations, and 3D Generation Breakthroughs Explore modern AI security, robotics, and 3D generation advancements that are shaping the future. ScienceToStartup tracks practical AI research, tools, and comm https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-05 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1772745498-hero.jpg AI Feedback Tools, 3D Vision, and Agentic Workflows Explore modern AI tools for feedback, 3D vision, and agentic workflows transforming industries. ScienceToStartup tracks practical AI research, tools, and commun https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-04 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1772662465-hero.jpg AI Innovations in Industrial Optimization and Content Creation Explore modern AI advancements in industrial optimization, graph intelligence, and content creation. ScienceToStartup tracks practical AI research, tools, and c https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-03 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1772572591-hero.jpg AI Innovations in Conversion Rate Prediction and Unlearning Explore important AI research on conversion rate prediction, unlearning models, and video understanding. ScienceToStartup tracks practical AI research, tools, a https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-02 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1772486882-hero.jpg AI Security Innovations and Manufacturing Efficiency Boosts Explore the latest AI advancements in security, manufacturing scheduling, and optimization techniques. ScienceToStartup tracks practical AI research, tools, and https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-03-01 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/03/1772399634-hero.jpg AI Research Rundown: Innovations in Emotional AI and Video Generation Explore important AI research on emotional detoxification and colonoscopy video generation. ScienceToStartup tracks practical AI research, tools, and community https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-02-28 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/02/1772312432-hero.jpg AI Research Rundown: Innovations in Medical Video Generation and QA Explore important AI research on colonoscopy video generation, QA benchmarks, and multi-agent systems optimization. ScienceToStartup tracks practical AI researc https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-02-27 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/02/1772227285-hero.jpg AI Research Rundown: Innovations in Human-Robot Interaction and QA Explore recent breakthroughs in AI, focusing on human-robot interaction, scalable QA benchmarks, and medical diagnostics. ScienceToStartup tracks practical AI r https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-02-26 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/02/1772142435-hero.jpg AI Research Rundown: Innovations in Text Ranking and Medical Imaging Explore recent breakthroughs in AI, focusing on text ranking methods, medical imaging denoising, and autonomous systems. ScienceToStartup tracks practical AI re https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-02-25 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/02/1772057331-hero.jpg AI Research Rundown: Innovations in Long-Context Inference and Medical AI Explore breakthroughs in long-context LLM inference, ECG representation learning, and robotic manipulation. ScienceToStartup tracks practical AI research, tools https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-02-24 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/02/1771967701-hero.jpg AI Research Rundown: Satellite Detection, VR Simulation, and Legal Predictions Explore modern AI research on detecting looted sites, simulating VR players, and predicting legal judgments. ScienceToStartup tracks practical AI research, tool https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-02-23 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/02/1771823298-hero.jpg AI Research Rundown: Enhancements in Adversarial Attacks and Reinforcement Learning Explore modern research on adversarial attacks, reinforcement learning, and AI applications in legal systems and long-horizon tasks. ScienceToStartup tracks pra https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-02-22 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/02/1771795691-hero.jpg AI Research Rundown: Time Series, Adversarial Attacks, and Formula Recognition Explore important AI research on time series forecasting, adversarial attacks on vision-language models, and innovative formula recognition. ScienceToStartup tr https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-02-21 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/02/1771707473-hero.jpg AI Research Rundown: Adversarial Attacks, Time Series, and Behavioral Prediction Explore the latest AI research on adversarial attacks, time series forecasting, and behavioral prediction models that are shaping the future. ScienceToStartup t https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-02-20 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/02/1771629545-hero.jpg AI Research Rundown: Formula Recognition, Behavioral Prediction, and Adversarial Attacks Explore the latest AI research on formula recognition, behavioral prediction, and adversarial attacks, highlighting key innovations and their implications. https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-02-19 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/02/1771462141-hero.jpg AI Research Rundown: Humanoid Motion, Graph Reasoning, and 3D Imaging Explore important AI research on humanoid motion learning, graph reasoning, and 3D imaging optimization. ScienceToStartup tracks practical AI research, tools, a https://sciencetostartup.com/articles/frontier-ai-innovations-for-startup-execution-2026-02-18 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/02/1771430761907-screenshot-2026-02-18-073119.png Frontier AI Innovations for Startup Execution Explore how frontier AI research is transforming startup execution across various sectors, from molecular design to cybersecurity. https://sciencetostartup.com/articles/science-to-startup-daily-brief-2026-02-16 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/02/1771276262-hero.jpg AI Research Rundown: Skills, Incident Response, and Video Models Explore the latest developments in AI, including benchmarking agent skills, incident response with LLMs, and efficient video language models. ScienceToStartup t https://sciencetostartup.com/articles/frontier-ai-and-startup-execution-key-innovations-2026-02-15 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/02/1771187306518-screenshot-2026-02-15-122810.png Frontier AI and Startup Execution: Key Innovations Discover frontier AI innovations in UAV preflight planning, robotics motion, and multimodal perception that can enhance startup execution. https://sciencetostartup.com/articles/daily-ai-research-rundown-2026-02-13 https://jdeeoknqehdvwmyoyayl.supabase.co/storage/v1/object/public/article-images/articles/2026/02/1771011121101-pexels-kindelmedia-8325944.jpg Daily AI Research Rundown Key insights from the latest papers on AI advancements.