ARXIV:2603.17811 · TRANSFORMER ROBUSTNESS · SUBMITTED 02 APR · 02:30 UTC · FRESHNESS STALE

VerifiedSource: PDF linkedPartialPaperPack: 3 of 4 citation fields filledMissingMissing fields: authorsPartialProof: unverified proof status

Dropout Robustness and Cognitive Profiling of Transformer Models via Stochastic Inference

arXiv

This research benchmarks dropout robustness in transformer models, providing insights for selecting models in uncertainty-aware applications.

Blocked on Code›Score5.0Evidence unverified

Opportunity summary

Pain This research benchmarks dropout robustness in transformer models, providing insights for selecting models in uncertainty-aware applications.

Evidence 0 refs | 0 sources | 17% coverage

Blocker Evidence unverified

Open Build Read PDF Signal Canvas Track

PROBLEM

This research benchmarks dropout robustness in transformer models, providing insights for selecting models in uncertainty-aware applications. While dropout is common during training, its inference-time effects via Monte Carlo sampling lack systematic evaluation across architectures,…

METHOD

Full abstract

Transformer-based language models are widely deployed for reasoning, yet their behavior under inference-time stochasticity remains underexplored. While dropout is common during training, its inference-time effects via Monte Carlo sampling lack systematic evaluation across architectures, limiting understanding of model reliability in uncertainty-aware applications. This work analyzes dropout-induced variability across 19 transformer models using MC Dropout with 100 stochastic forward passes per sample. Dropout robustness is defined as maintaining high accuracy and stable predictions under stochastic inference, measured by standard deviation of per-run accuracies. A cognitive decomposition framework disentangles performance into memory and reasoning components. Experiments span five dropout configurations yielding 95 unique evaluations on 1,000 samples. Results reveal substantial architectural variation. Smaller models demonstrate perfect prediction stability while medium-sized models exhibit notable volatility. Mid-sized models achieve the best overall performance; larger models excel at memory tasks. Critically, 53% of models suffer severe accuracy degradation under baseline MC Dropout, with task-specialized models losing up to 24 percentage points, indicating unsuitability for uncertainty quantification in these architectures. Asymmetric effects emerge: high dropout reduces memory accuracy by 27 percentage points while reasoning degrades only 1 point, suggesting memory tasks rely on stable representations that dropout disrupts. 84% of models demonstrate memory-biased performance. This provides the first comprehensive MC Dropout benchmark for transformers, revealing dropout robustness is architecture-dependent and uncorrelated with scale. The cognitive profiling framework offers actionable guidance for model selection in uncertainty-aware applications.

RESULT

ScienceToStartup currently rates this 5.0/10 on the public viability pass. Results reveal substantial architectural variation.

WHY NOW

Transformer Robustness moved forward this cycle; last verified April 2026. Public score 5.0/10.

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

Opportunity summary

Score5.0

PainThis research benchmarks dropout robustness in transformer models, providing insights for selecting models in uncertainty-aware applications.

Evidence0 refs | 0 sources | 17% coverage

Blockermissing authors

Analysis summary

This research benchmarks dropout robustness in transformer models, providing insights for selecting models in uncertainty-aware applications.

VerifiedSource: PDF linkedPartialPaperPack: 3 of 4 citation fields filledMissingMissing fields: authorsPartialProof: unverified proof status

Competitive landscape

This research benchmarks dropout robustness in transformer models, providing insights for selecting models in uncertainty-aware applications.

Segment

Transformer Robustness

Adoption evidence

No public code link in the paper record yet

Commercial read

5.0/10 public viability

Direct

not classified

Adjacent

not classified

Substitute

not classified

Unknown

not classified

{ "contract_version": "paper-r2", "paper_id": "b1205904-9114-4ef4-b0ac-3d2d2c9911a2", "arxiv_id": "2603.17811", "canonical_route": "/paper/dropout-robustness-and-cognitive-profiling-of-transformer-models-via-stochastic-inference", "active_tab": "synced from current hash by the drawer client", "selected_artifact": "dropout-robustness-and-cognitive-profiling-of-transformer-models-via-stochastic-inference", "endpoints": { "paper_pack": "/api/v1/paper/dropout-robustness-and-cognitive-profiling-of-transformer-models-via-stochastic-inference/paper-pack", "build_passport": "/api/v1/paper/dropout-robustness-and-cognitive-profiling-of-transformer-models-via-stochastic-inference/build-passport", "mcp_resource": "sciencetostartup://surfaces/paper-workspace" } }

{ "surface": "paper", "mode": "paper", "query": "Dropout Robustness and Cognitive Profiling of Transformer Models via Stochastic Inference", "normalized_query": "2603.17811", "route": "/paper/dropout-robustness-and-cognitive-profiling-of-transformer-models-via-stochastic-inference", "paper_ref": "dropout-robustness-and-cognitive-profiling-of-transformer-models-via-stochastic-inference", "topic_slug": null, "benchmark_ref": null, "dataset_ref": null }

{ "@context": "https://schema.org", "@graph": [ { "@type": "WebPage", "@id": "https://sciencetostartup.com/paper/dropout-robustness-and-cognitive-profiling-of-transformer-models-via-stochastic-inference#webpage", "url": "https://sciencetostartup.com/paper/dropout-robustness-and-cognitive-profiling-of-transformer-models-via-stochastic-inference", "name": "Dropout Robustness and Cognitive Profiling of Transformer Models via Stochastic Inference", "description": "This research benchmarks dropout robustness in transformer models, providing insights for selecting models in uncertainty-aware applications.", "isPartOf": { "@id": "https://sciencetostartup.com/#website" } }, { "@type": "ScholarlyArticle", "@id": "https://sciencetostartup.com/paper/dropout-robustness-and-cognitive-profiling-of-transformer-models-via-stochastic-inference#scholarlyArticle", "headline": "Dropout Robustness and Cognitive Profiling of Transformer Models via Stochastic Inference", "description": "This research benchmarks dropout robustness in transformer models, providing insights for selecting models in uncertainty-aware applications.", "url": "https://sciencetostartup.com/paper/dropout-robustness-and-cognitive-profiling-of-transformer-models-via-stochastic-inference", "sameAs": "https://arxiv.org/abs/2603.17811", "identifier": { "@type": "PropertyValue", "propertyID": "arXiv", "value": "2603.17811" }, "isAccessibleForFree": true, "isPartOf": { "@id": "https://sciencetostartup.com/#website" }, "datePublished": "2026-03-18T15:04:26.000Z", "additionalProperty": [ { "@type": "PropertyValue", "propertyID": "viabilityScore", "value": 5 }, { "@type": "PropertyValue", "propertyID": "researchDomain", "value": "Transformer Robustness" } ] }, { "@type": "BreadcrumbList", "itemListElement": [ { "@type": "ListItem", "position": 1, "name": "Home", "item": "https://sciencetostartup.com" }, { "@type": "ListItem", "position": 2, "name": "Transformer Robustness", "item": "https://sciencetostartup.com/topics" }, { "@type": "ListItem", "position": 3, "name": "Dropout Robustness and Cognitive Profiling of Transformer Mo", "item": "https://sciencetostartup.com/paper/dropout-robustness-and-cognitive-profiling-of-transformer-models-via-stochastic-inference" } ] } ] }

Competitive landscape

This research benchmarks dropout robustness in transformer models, providing insights for selecting models in uncertainty-aware applications.

Segment

Transformer Robustness

Adoption evidence

No public code link in the paper record yet

Commercial read

5.0/10 public viability

Direct

not classified

Adjacent

not classified

Substitute

not classified

Unknown

not classified

Dropout Robustness and Cognitive Profiling of Transformer Models via Stochastic Inference

Dropout Robustness and Cognitive Profiling of Transformer Models via Stochastic Inference

Claim map

Constellation map

Competitive landscape

Buzz

PDF

REFERENCES

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

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