ARXIV:2605.00814 · LVLMS · SUBMITTED 04 MAY · 20:21 UTC · FRESHNESS STALE

VerifiedSource: PDF linkedVerifiedPaperPack: citation fields availablePartialProof: unverified proof status

Persistent Visual Memory: Sustaining Perception for Deep Generation in LVLMs

Siyuan Huang · Xiaoye Qu · Yafu Li · Tong Zhu · Zefeng He · Muxin Fu · +3 at arXiv

A lightweight module enhances Large Vision-Language Models by preventing visual signal decay during long text generation, improving performance on complex reasoning tasks.

Ship in 2-4 weeks›Score7.0Evidence unverified

Opportunity summary

Pain A lightweight module enhances Large Vision-Language Models by preventing visual signal decay during long text generation, improving performance on complex reasoning tasks.

Evidence 0 refs | 4 sources | 67% coverage

Blocker Evidence unverified

Open Build Read PDF Signal Canvas Track

PROBLEM

A lightweight module enhances Large Vision-Language Models by preventing visual signal decay during long text generation, improving performance on complex reasoning tasks. To counteract this, we propose Persistent Visual Memory (PVM), a lightweight learnable…

METHOD

Full abstract

While autoregressive Large Vision-Language Models (LVLMs) demonstrate remarkable proficiency in multimodal tasks, they face a "Visual Signal Dilution" phenomenon, where the accumulation of textual history expands the attention partition function, causing visual attention to decay inversely with generated sequence length. To counteract this, we propose Persistent Visual Memory (PVM), a lightweight learnable module designed to ensure sustained, on-demand visual perception. Integrated as a parallel branch alongside the Feed-Forward Network (FFN) in LVLMs, PVM establishes a distance-agnostic retrieval pathway that directly provides visual embeddings for precise visual perception, thereby structurally mitigating the signal suppression inherent to deep generation. Extensive experiments on Qwen3-VL models demonstrate that PVM brings notable improvements with negligible parameter overhead, delivering consistent average accuracy gains across both 4B and 8B scales, particularly in complex reasoning tasks that demand persistent visual perception. Furthermore, in-depth analysis reveals that PVM can resist length-induced signal decay and accelerate internal prediction convergence.

RESULT

ScienceToStartup currently rates this 7.0/10 on the public viability pass. While autoregressive Large Vision-Language Models (LVLMs) demonstrate remarkable proficiency in multimodal tasks, they face a "Visual Signal Dilution" phenomenon, where the accumulation of textual…

WHY NOW

LVLMs moved forward this cycle; last verified May 2026. Public score 7.0/10. Implementation evidence is present through a linked repository.

Continue into Read for claims, analysis, references, and neighboring papers.

Opportunity summary

Score7.0

PainA lightweight module enhances Large Vision-Language Models by preventing visual signal decay during long text generation, improving performance on complex reasoning tasks.

Evidence0 refs | 4 sources | 67% coverage

Blockerno shell-level blocker reported

Analysis summary

A lightweight module enhances Large Vision-Language Models by preventing visual signal decay during long text generation, improving performance on complex reasoning tasks.

VerifiedSource: PDF linkedVerifiedPaperPack: citation fields availablePartialProof: unverified proof status

ARXIV:2605.00814 · LVLMS · SUBMITTED 04 MAY · 20:21 UTC · FRESHNESS STALE

VerifiedSource: PDF linkedVerifiedPaperPack: citation fields availablePartialProof: unverified proof status

Persistent Visual Memory: Sustaining Perception for Deep Generation in LVLMs

Siyuan Huang · Xiaoye Qu · Yafu Li · Tong Zhu · Zefeng He · Muxin Fu · +3 at arXiv

A lightweight module enhances Large Vision-Language Models by preventing visual signal decay during long text generation, improving performance on complex reasoning tasks.

Ship in 2-4 weeks›Score7.0Evidence unverified

Opportunity summary

Pain A lightweight module enhances Large Vision-Language Models by preventing visual signal decay during long text generation, improving performance on complex reasoning tasks.

Evidence 0 refs | 4 sources | 67% coverage

Blocker Evidence unverified

Open Build Read PDF Signal Canvas Track

PROBLEM

METHOD

Full abstract

RESULT

WHY NOW

LVLMs moved forward this cycle; last verified May 2026. Public score 7.0/10. Implementation evidence is present through a linked repository.

Continue into Read for claims, analysis, references, and neighboring papers.

Opportunity summary

Score7.0

PainA lightweight module enhances Large Vision-Language Models by preventing visual signal decay during long text generation, improving performance on complex reasoning tasks.

Evidence0 refs | 4 sources | 67% coverage

Blockerno shell-level blocker reported

Analysis summary

A lightweight module enhances Large Vision-Language Models by preventing visual signal decay during long text generation, improving performance on complex reasoning tasks.

VerifiedSource: PDF linkedVerifiedPaperPack: citation fields availablePartialProof: unverified proof status

Paper Pack

10.48550/arXiv.2605.00814

Persistent Visual Memory: Sustaining Perception for Deep Generation in LVLMs

A lightweight module enhances Large Vision-Language Models by preventing visual signal decay during long text generation, improving performance on complex reasoning tasks.

Abstract

Source availability

PDF linked

The paper record includes a public PDF URL.

Extraction status

Parse run linked

A document parse run is attached to this paper.

Proof status

unverified

0 refs; 4 sources; 67% coverage.

What was readable

linkedon file20 anchorsderived fallback91 indexednot indexed

Derived fallback: Estimated from adjacent evidence; not verified from source.

Viability

7.0

Time to MVP

MVP estimate missing

Commercial

coderepo url

Export

Preparing verified analysis

lens / founder

PROBLEM

METHOD

RESULT

WHY NOW

LVLMs moved forward this cycle; last verified May 2026. Public score 7.0/10. Implementation evidence is present through a linked repository.

Claim map

Abstract-backed public claims while anchored extraction refreshes.

Strong 0Mixed 0Weak 4

Evidencepartial
A lightweight module enhances Large Vision-Language Models by preventing visual signal decay during long text generation, improving performance on complex reasoning tasks. To counteract this, we propose Persistent Visual Memory (PVM), a lightweight learnable module designed to ensure sustained, on-demand visual perception.
Implicationpartial
Abstract-backed fallback claim; anchored extraction has not materialized a public claim row yet.
Verificationpartial
partial
Evidencepartial
While autoregressive Large Vision-Language Models (LVLMs) demonstrate remarkable proficiency in multimodal tasks, they face a "Visual Signal Dilution" phenomenon, where the accumulation of textual history expands the attention partition function, causing visual attention to decay inversely with generated sequence length. To counteract this, we propose Persistent Visual Memory (PVM), a lightweight learnable module designed to ensure sustained, on-demand visual perception.
Implicationpartial
Abstract-backed fallback claim; anchored extraction has not materialized a public claim row yet.
Verificationpartial
partial
Evidencepartial
ScienceToStartup currently rates this 7.0/10 on the public viability pass. While autoregressive Large Vision-Language Models (LVLMs) demonstrate remarkable proficiency in multimodal tasks, they face a "Visual Signal Dilution" phenomenon, where the accumulation of textual history expands the attention partition function, causing visual attention to decay inversely with generated sequence length. A public repository is linked, so build verification can inspect implementation evidence instead of treating the paper as PDF-only.
Implicationpartial
Abstract-backed fallback claim; anchored extraction has not materialized a public claim row yet.
Verificationpartial
partial
Evidencepartial
LVLMs moved forward this cycle; last verified May 2026. Public score 7.0/10. Implementation evidence is present through a linked repository.
Implicationpartial
Abstract-backed fallback claim; anchored extraction has not materialized a public claim row yet.
Verificationpartial
partial

Constellation map

Paper-native neighborhood for concepts, methods, materials, markets, and competitors. Missing lanes stay labeled instead of disappearing behind commercialization gates.

Open full Signal Canvas

Concepts

not indexed

Methods

Materials

PDF linkedDocument parse run

Markets

LVLMs

Competitors

not indexed

Competitive landscape

A lightweight module enhances Large Vision-Language Models by preventing visual signal decay during long text generation, improving performance on complex reasoning tasks.

Segment

LVLMs

Adoption evidence

Public code linked for build inspection

Commercial read

7.0/10 public viability

Direct

not classified

Adjacent

not classified

Substitute

not classified

Unknown

not classified

Buzz

No indexed public discussion is attached to 2605.00814 yet. That is a visibility signal, not a blank module: the monitor is watching the public channels below.

Hacker News

Not indexed yet

Bluesky

Not indexed yet

PDF

Preview the source document here, or use the hero PDF action for a new tab.

References(91)

GEMS: Agent-Native Multimodal Generation with Memory and Skills

2026Zefeng He, Siyuan Huang et al.

LatentMem: Customizing Latent Memory for Multi-Agent Systems

2026Muxin Fu, Guibin Zhang et al.

Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models

2026Xin Cheng, Wangding Zeng et al.

Vision-Language Introspection: Mitigating Overconfident Hallucinations in MLLMs via Interpretable Bi-Causal Steering

2026Shuliang Liu, Songbo Yang et al.

DiffThinker: Towards Generative Multimodal Reasoning with Diffusion Models

2025Zefeng He, Xiaoye Qu et al.

VPTracker: Global Vision-Language Tracking via Visual Prompt and MLLM

2025Jingchao Wang, Kaiwen Zhou et al.

MemVerse: Multimodal Memory for Lifelong Learning Agents

2025Junming Liu, Yifei Sun et al.

OneThinker: All-in-one Reasoning Model for Image and Video

2025Kaituo Feng, Manyuan Zhang et al.

Boosting Reasoning in Large Multimodal Models via Activation Replay

2025Yun Xing, Xiaobin Hu et al.

Perceptual-Evidence Anchored Reinforced Learning for Multimodal Reasoning

2025Chi Zhang, Haibo Qiu et al.

OpenMMReasoner: Pushing the Frontiers for Multimodal Reasoning with an Open and General Recipe

2025Kaichen Zhang, Keming Wu et al.

VisMem: Latent Vision Memory Unlocks Potential of Vision-Language Models

2025Xinlei Yu, Chengming Xu et al.

Remember Me: Bridging the Long-Range Gap in LVLMs with Three-Step Inference-Only Decay Resilience Strategies

2025Peng Gao, Yujian Lee et al.

VideoSSR: Video Self-Supervised Reinforcement Learning

2025Zefeng He, Xiaoye Qu et al.

Spotlight on Token Perception for Multimodal Reinforcement Learning

2025Siyuan Huang, Xiaoye Qu et al.

ChainMPQ: Interleaved Text-Image Reasoning Chains for Mitigating Relation Hallucinations

2025Yike Wu, Yiwei Wang et al.

FrameThinker: Learning to Think with Long Videos via Multi-Turn Frame Spotlighting

2025Zefeng He, Xiaoye Qu et al.

Euclid's Gift: Enhancing Spatial Perception and Reasoning in Vision-Language Models via Geometric Surrogate Tasks

2025Shijie Lian, Changti Wu et al.

LLaVA-OneVision-1.5: Fully Open Framework for Democratized Multimodal Training

2025Xiang An, Yin Xie et al.

Context-Aware Multi-Turn Visual-Textual Reasoning in LVLMs via Dynamic Memory and Adaptive Visual Guidance

2025Weijie Shen, Xinrui Wang et al.

Showing 20 of 91 references

CITED BY

No citing papers are indexed in the public S2S graph yet. This is an explicit zero-signal state, not a hidden lookup.

Foundation

Prior WorkPerceptio: Perception Enhanced Vision Language Models via Spatial Token Generation

7.0

Prior WorkFinding the Correct Visual Evidence Without Forgetting: Mitigating Hallucination in LVLMs via Inter-Layer Visual Attention Discrepancy

7.0

Extension

Builds On ThisiGVLM: Dynamic Instruction-Guided Vision Encoding for Question-Aware Multimodal Understanding

6.0

Builds On ThisVLMs Need Words: Vision Language Models Ignore Visual Detail In Favor of Semantic Anchors

5.0

Builds On ThisCMMR-VLN: Vision-and-Language Navigation via Continual Multimodal Memory Retrieval

3.0

Builds On ThisVLM3: Vision Language Models Are Native 3D Learners

4.0

Builds On ThisGazeVLM: Active Vision via Internal Attention Control for Multimodal Reasoning

6.0

Commercially relevant

Higher ViabilityMemory-Augmented Vision-Language Agents for Persistent and Semantically Consistent Object Captioning

9.0

Higher ViabilityLLMind: Bio-inspired Training-free Adaptive Visual Representations for Vision-Language Models

8.0

Higher ViabilityPenguin-VL: Exploring the Efficiency Limits of VLM with LLM-based Vision Encoders

8.0

Conflicting

none indexed

Owned Distribution

Subscribe to the weekly brief

Get the weekly shortlist of commercializable papers, benchmark movers, and proof receipts that matter for product execution.

Agent drawer

5 surfaces preserved for agents. Humans can ignore.

Developer contracts, payload previews, evidence maps, and run controls stay here instead of the Read, Build, and Track workspace.

Run context

Paper: 2605.00814
Route: /paper/persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms
Active tab: read
Artifact: persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms

Available agents

Read extractor
Build planner
Track monitor
Competitive mapper
Related-paper scout

API/MCP endpoints

REST paper pack API/api/v1/paper/persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms/paper-pack
REST build passport API/api/v1/paper/persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms/build-passport
REST OpenAPI/api/openapi.json
MCP descriptor/api/mcp
MCP resourcesciencetostartup://surfaces/paper-workspace

Tool contracts

paper_packbuild_passportopportunity_kernelforesightsource_proofevidence_state

Payload preview

Inspect payload

{
  "contract_version": "paper-r2",
  "paper_id": "5e7ad01a-30e6-47b4-aa83-2965d0b432ce",
  "arxiv_id": "2605.00814",
  "canonical_route": "/paper/persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms",
  "active_tab": "synced from current hash by the drawer client",
  "selected_artifact": "persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms",
  "endpoints": {
    "paper_pack": "/api/v1/paper/persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms/paper-pack",
    "build_passport": "/api/v1/paper/persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms/build-passport",
    "mcp_resource": "sciencetostartup://surfaces/paper-workspace"
  }
}

Schema validation

paper-r2 contract: present
JSON-LD twin: SSR emitted
OpenAPI path parity: /api/openapi.json
MCP resource parity: paper-workspace

Job trace

queued: drawer opened by user action
running: inspect or copy payload
succeeded: payload available in SSR
failed: route errors appear in evidence cards

Evidence map

sources used: page freshness, source proof anchors, JSON-LD
missing sources: exposed by PaperPack and EvidenceState chips
derived fallbacks: marked unverified before handoff

Page Freshness

Canonical route, proof status, last verified, refs, sources, and coverage.

Page Freshness

Paper proof surface

Canonical route: /paper/persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms

stale

Proof freshness: stale
Proof status: unverified
Display score: 7/10
Last proof check: 2026-05-04
Score updated: 2026-05-04
Score fresh until: 2026-06-03
References: 0
Source count: 4
Coverage: 67%

This page is showing the last landed evidence receipt and score bundle because the latest proof data is outside the freshness window.

OpenAlex: pending — this preprint is not yet indexed by OpenAlex.

Agent Handoff

Endpoint list, payload shape, route context, and copyable handoff data.

Agent Handoff

Persistent Visual Memory: Sustaining Perception for Deep Generation in LVLMs

Canonical ID persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms | Route /paper/persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms

REST example

curl https://sciencetostartup.com/api/v1/agent-handoff/paper/persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms

MCP example

{
  "tool": "get_paper",
  "arguments": {
    "arxiv_id": "2605.00814"
  }
}

source_context

{
  "surface": "paper",
  "mode": "paper",
  "query": "Persistent Visual Memory: Sustaining Perception for Deep Generation in LVLMs",
  "normalized_query": "2605.00814",
  "route": "/paper/persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms",
  "paper_ref": "persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms",
  "topic_slug": null,
  "benchmark_ref": null,
  "dataset_ref": null
}

Buildability Receipt

Verdict, compute envelope, blockers, signature state, and receipt links.

Paper proof page receipt window

Ready for execution: Persistent Visual Memory: Sustaining Perception for Deep Generation in LVLMs

/buildability/persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms

Build Nowready

Subject: Persistent Visual Memory: Sustaining Perception for Deep Generation in LVLMs

Verdict

Build Now

Verdict is Build Now because viability and implementation proof cleared the Wave 1 scaffold thresholds.

Time to first demo

Insufficient data

No first-demo timestamp, owner estimate, or elapsed demo receipt is attached to this surface.

Compute envelope

Structured compute envelope

Insufficient data

No data, compute, hardware, memory, latency, dependency, or serving requirement receipt is attached.

Evidence ids

Receipt path

/buildability/persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms

Paper ref

persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms

arXiv id

2605.00814

Freshness

Generated at

2026-05-04T20:21:55.060Z

Evidence freshness

stale

Last verification

2026-05-04T20:21:55.060Z

Sources

References

Coverage

67%

Hash state

Lineage hash

408f6c8fbc90b27c5d7be43b8059bda199f146c2674e5d337048ae843a866532

Canonical opportunity-kernel lineage hash.

Signature state

External signature

unsigned_external

No founder, registry, pilot, or production-adoption signature is attached to this receipt.

Verification

not_verified

Verification is blocked until an external signature is provided.

Blockers

Missing: references
Missing: proof_status
Unknown: proof verification has not been recorded yet

Pending verification refs / 4 sources / Verification pending

references

proof_status

Missing proof, requirement, signature, approval, adoption, or telemetry fields are blockers and must not be inferred.

Open receipt API receipt Build Loop Signal Canvas Proof divergence Divergence API Brier outcomes API

Source Proof anchors

Visual citations from the paper document graph.

Source proof

Visual citation anchors from the paper document graph.

20 anchors

proof blockPage 368%

This equation captures one of the core mathematical components of the system. fixed set of visual tokens V (|V| = M) and a dynamically growing set of textual history tokens Tt

Page and bbox are available; crop image is pending.

proof blockPage 368%

This equation captures one of the core mathematical components of the system. (|Tt| = t). For the current query vector qt in a Transformer block [70], let sk(qt) = (q⊤

Page and bbox are available; crop image is pending.

proof blockPage 468%

This equation captures one of the core mathematical components of the system. Proof. Substituting the visual upper bound ZV ≤β and the textual lower bound ZT ≥µt into Eq.

Page and bbox are available; crop image is pending.

JSON-LD twin

The application/ld+json payload rendered for agents.

{
  "@context": "https://schema.org",
  "@graph": [
    {
      "@type": "WebPage",
      "@id": "https://sciencetostartup.com/paper/persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms#webpage",
      "url": "https://sciencetostartup.com/paper/persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms",
      "name": "Persistent Visual Memory: Sustaining Perception for Deep Generation in LVLMs",
      "description": "A lightweight module enhances Large Vision-Language Models by preventing visual signal decay during long text generation, improving performance on complex reasoning tasks.",
      "isPartOf": {
        "@id": "https://sciencetostartup.com/#website"
      }
    },
    {
      "@type": "ScholarlyArticle",
      "@id": "https://sciencetostartup.com/paper/persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms#scholarlyArticle",
      "headline": "Persistent Visual Memory: Sustaining Perception for Deep Generation in LVLMs",
      "description": "A lightweight module enhances Large Vision-Language Models by preventing visual signal decay during long text generation, improving performance on complex reasoning tasks.",
      "url": "https://sciencetostartup.com/paper/persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms",
      "sameAs": "https://arxiv.org/abs/2605.00814",
      "identifier": {
        "@type": "PropertyValue",
        "propertyID": "arXiv",
        "value": "2605.00814"
      },
      "isAccessibleForFree": true,
      "isPartOf": {
        "@id": "https://sciencetostartup.com/#website"
      },
      "datePublished": "2026-05-01T17:54:37.000Z",
      "author": [
        {
          "@type": "Person",
          "name": "Siyuan Huang"
        },
        {
          "@type": "Person",
          "name": "Xiaoye Qu"
        },
        {
          "@type": "Person",
          "name": "Yafu Li"
        },
        {
          "@type": "Person",
          "name": "Tong Zhu"
        },
        {
          "@type": "Person",
          "name": "Zefeng He"
        },
        {
          "@type": "Person",
          "name": "Muxin Fu"
        },
        {
          "@type": "Person",
          "name": "Daizong Liu"
        },
        {
          "@type": "Person",
          "name": "Wei-Long Zheng"
        },
        {
          "@type": "Person",
          "name": "Yu Cheng"
        }
      ],
      "citation": [
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "9f41bd74c19bf47418c2448c37d81d7a9c410057"
          },
          "url": "https://www.semanticscholar.org/paper/9f41bd74c19bf47418c2448c37d81d7a9c410057"
        },
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "dfd9cf2191e91dbd6ceb683831d0d554916b0b56"
          },
          "url": "https://www.semanticscholar.org/paper/dfd9cf2191e91dbd6ceb683831d0d554916b0b56"
        },
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "c283116f911bb7457e595656a28d3b500744a495"
          },
          "url": "https://www.semanticscholar.org/paper/c283116f911bb7457e595656a28d3b500744a495"
        },
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "350846151ba44c7d67c5c74735e76c2444fb3ddc"
          },
          "url": "https://www.semanticscholar.org/paper/350846151ba44c7d67c5c74735e76c2444fb3ddc"
        },
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "c101957e12836a66934d5966c6901781c7055ebf"
          },
          "url": "https://www.semanticscholar.org/paper/c101957e12836a66934d5966c6901781c7055ebf"
        },
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "1c4a2138cd1085145b274ae6ae411b653fef31dd"
          },
          "url": "https://www.semanticscholar.org/paper/1c4a2138cd1085145b274ae6ae411b653fef31dd"
        },
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "bd0e16fe2f26e000491632a1155e19ad7c15a1e0"
          },
          "url": "https://www.semanticscholar.org/paper/bd0e16fe2f26e000491632a1155e19ad7c15a1e0"
        },
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "9f4b49a82befbc701b9d6e1bdf18bf2909c88802"
          },
          "url": "https://www.semanticscholar.org/paper/9f4b49a82befbc701b9d6e1bdf18bf2909c88802"
        },
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "e6c47beea2d8df421ac7dd3ae6e064d6f77adcbd"
          },
          "url": "https://www.semanticscholar.org/paper/e6c47beea2d8df421ac7dd3ae6e064d6f77adcbd"
        },
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "e89b4f7211414dc9f3d3d5f855a0296801bcab4c"
          },
          "url": "https://www.semanticscholar.org/paper/e89b4f7211414dc9f3d3d5f855a0296801bcab4c"
        },
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "a6fa28d3912660eed29714aacde394b87bf1c315"
          },
          "url": "https://www.semanticscholar.org/paper/a6fa28d3912660eed29714aacde394b87bf1c315"
        },
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "a2276e8466c1adbdd08c6f0a94553afdc38407ee"
          },
          "url": "https://www.semanticscholar.org/paper/a2276e8466c1adbdd08c6f0a94553afdc38407ee"
        },
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "d7fbad02ea972001cd1408f5eea4947bd1ae7ab2"
          },
          "url": "https://www.semanticscholar.org/paper/d7fbad02ea972001cd1408f5eea4947bd1ae7ab2"
        },
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "13b0e844d6cb3ed13e0c410e52ca4d265b6cd921"
          },
          "url": "https://www.semanticscholar.org/paper/13b0e844d6cb3ed13e0c410e52ca4d265b6cd921"
        },
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "134fde5291ceb00da3869db8040fe8ce136bf5b0"
          },
          "url": "https://www.semanticscholar.org/paper/134fde5291ceb00da3869db8040fe8ce136bf5b0"
        },
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "22a758fc41ab606ec5f9a6f62f75bfcc325efff7"
          },
          "url": "https://www.semanticscholar.org/paper/22a758fc41ab606ec5f9a6f62f75bfcc325efff7"
        },
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "c75199c6dbbcc15fa7c63315fea4ba16f65e4846"
          },
          "url": "https://www.semanticscholar.org/paper/c75199c6dbbcc15fa7c63315fea4ba16f65e4846"
        },
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "07ba7629ff0e2c9edb2eef8faaaf8478e156b347"
          },
          "url": "https://www.semanticscholar.org/paper/07ba7629ff0e2c9edb2eef8faaaf8478e156b347"
        },
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "4da3b03dd7b24bfbe7a818b9f00300bb349dcc38"
          },
          "url": "https://www.semanticscholar.org/paper/4da3b03dd7b24bfbe7a818b9f00300bb349dcc38"
        },
        {
          "@type": "ScholarlyArticle",
          "identifier": {
            "@type": "PropertyValue",
            "propertyID": "SemanticScholar",
            "value": "c2cf8cd04cea3b9dd85927f7573c950ef078b4e5"
          },
          "url": "https://www.semanticscholar.org/paper/c2cf8cd04cea3b9dd85927f7573c950ef078b4e5"
        }
      ],
      "codeRepository": "https://github.com/huaixuheqing/PVM",
      "additionalProperty": [
        {
          "@type": "PropertyValue",
          "propertyID": "viabilityScore",
          "value": 7
        },
        {
          "@type": "PropertyValue",
          "propertyID": "researchDomain",
          "value": "LVLMs"
        },
        {
          "@type": "PropertyValue",
          "propertyID": "commercialReadiness",
          "value": "code, repo url"
        }
      ]
    },
    {
      "@type": "SoftwareSourceCode",
      "@id": "https://sciencetostartup.com/paper/persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms#software",
      "name": "Persistent Visual Memory: Sustaining Perception for Deep Generation in LVLMs - Source Code",
      "description": "A lightweight module enhances Large Vision-Language Models by preventing visual signal decay during long text generation, improving performance on complex reasoning tasks.",
      "codeRepository": "https://github.com/huaixuheqing/PVM",
      "url": "https://github.com/huaixuheqing/PVM"
    },
    {
      "@type": "BreadcrumbList",
      "itemListElement": [
        {
          "@type": "ListItem",
          "position": 1,
          "name": "Home",
          "item": "https://sciencetostartup.com"
        },
        {
          "@type": "ListItem",
          "position": 2,
          "name": "LVLMs",
          "item": "https://sciencetostartup.com/topics"
        },
        {
          "@type": "ListItem",
          "position": 3,
          "name": "Persistent Visual Memory: Sustaining Perception for Deep Gen",
          "item": "https://sciencetostartup.com/paper/persistent-visual-memory-sustaining-perception-for-deep-generation-in-lvlms"
        }
      ]
    }
  ]
}

Persistent Visual Memory: Sustaining Perception for Deep Generation in LVLMs

Persistent Visual Memory: Sustaining Perception for Deep Generation in LVLMs

Claim map

Constellation map

Competitive landscape

Buzz

PDF

References(91)

Related Papers

Subscribe to the weekly brief

Build artifacts

Brief

Experiment plan

Validation checklist

Scientific founder

Translational engineer

Domain operator

GTM lead

Regulatory/clinical advisor

Timeline

Claim map

Constellation map

Competitive landscape

Buzz

PDF

References(91)

Related Papers

Subscribe to the weekly brief

Build artifacts

Brief

Experiment plan

Validation checklist

Scientific founder

Translational engineer

Domain operator

GTM lead

Regulatory/clinical advisor

Timeline