Search: “validation-debt”

30 stories

DomainAll Cyber AIWindow24h 3d 7d 30d AllIn the wild Important only Show duplicates

CategoryAll

SeverityAll critical 0 high 0 medium 0 low 0 info 0

Smart search ranks by meaning as well as keywords (one row per story, last 45 days).

The inconvenient truth about AI pentesting: someone has to check all the work

Survey of 158 practitioners shows AI pentesting floods teams with findings, creating 'validation debt' most teams cannot process.

The article argues AI pentesting creates 'validation debt': discovery scales far faster than teams' ability to verify AI-generated findings. In a survey of 158 practitioners, only 20.3% had workflows to triage more than 500 AI-generated candidates per engagement, while 29.7% called such volume unmanageable. One respondent spent two days validating 300 AI findings, of which 250 were duplicates, non-exploitable, or nonexistent. The author recommends capacity planning, ruthless deduplication, and risk-based prioritization before adopting AI pentesting tools.

Security AffairsAug 11, 16:20 UTC · Aug 11, 2026Industry

Approval Integrity and Recovery in LLM Answer Publication

Study measures approval integrity in Lightcap LLM answer publication, finding the 14B response-act checker accepts 291 of 302 unsupported answers.

The study evaluates exact-content binding, authorization freshness, and checkpoint recovery in Lightcap's publication enforcement using 3,600 assessments over 900 human-annotated RAGTruth responses from three Ministral models. The production 14B response-act checker accepts 291 of 302 unsupported answers versus 41 for a direct-grounding baseline, with supported-answer retention of 95.2% versus 66.9%. A stateful recheck-recovery policy increases exact-match error by 9.23 percentage points relative to initial checkpoints, and controlled evidence-fingerprint changes expose asymmetric freshness enforcement between publication and recovery. A separate BIPIA prompt-injection experiment records zero target insertions among 266 valid editor outputs.

arXiv cs.CRSep 14, 13:48 UTC · 2d agoAI safety & security

A Cautionary Tale About Data Breach Claims, Verification and Carhartt

Troy Hunt cautions that claimed Carhartt breach data requires verification, warning that criminals' breach claims are not always accurate.

Troy Hunt published a cautionary tale about data breach claims, verification, and apparel brand Carhartt. He argues that claimed breaches from cybercriminals should not be taken at face value and may stem from errors by the criminals themselves. The piece underscores the need to verify breach data before treating it as authentic, in the vein of Have I Been Pwned's validation practices.

Troy HuntAug 25, 21:51 UTC · 21d agoData breach

E2A-Bench: Benchmarking Evidence-to-Action Reliability in Financial Chart Reasoning

E2A-Bench, a 969-query financial chart reasoning benchmark, finds VLMs fail evidence-to-action consistency, with fine-tuning amplifying BUY:SELL bias 4-6x.

E2A-Bench is a 969-query benchmark built from 323 HS300 constituents across three input modalities with deterministic OHLCV-derived evidence anchors, evaluating grounding, reasoning-action consistency, evidence-confidence calibration, and directional coverage via UCR, RCI, ECI, and NDR metrics. Testing 20 VLMs showed the lowest-hallucination model ranked near the bottom on coverage with only 6.4% directional coverage, and oracle-aided verification reduced unsupported claims but could collapse coverage. Financial fine-tuning amplified the BUY:SELL ratio by factors of 4.21 to 4.68 across base-fine-tuned pairs.

Hugging Face daily papersSep 12, 20:00 UTC · 3d agoAI research

Engineered Persuasion: Evaluating Personalized Pretexts in LLM-Generated Spear Phishing

A study of 180 US workers found each LLM phishing personalization level raised click-intention odds by 28%, but credibility depends on context fit.

The arXiv paper evaluates how personalized pretexts in LLM-generated spear phishing affect perceived credibility, using 180 US working adults across 1,436 evaluations of emails with four cumulative personalization levels, from workplace context to shared-project details. Convincingness rose 2.40 points per level in sensitivity analysis and click-intention odds increased 28% per level, while non-clickers shifted toward deleting rather than reporting. Qualitative coding showed details matching the recipient's role and routines supported credibility, whereas incorrect, vague, or channel-inappropriate details raised suspicion. The authors argue personalization effectiveness depends on pretext fit, with implications for workplace security training.

arXiv cs.CRSep 3, 19:21 UTC · 12d agoResearch

Necessary or Sufficient? Evaluating LLM Explanations With Behavioural Evidence

New framework tests whether LLM-cited explanation factors are necessary or sufficient, finding weak correlation across Claude, GPT, and Gemini models.

An arXiv paper introduces black-box intervention tests measuring whether factors LLMs cite in their explanations are necessary or sufficient for their outputs in agent oversight workflows. Across eight models from the Claude, GPT, and Gemini families, Spearman correlations between cited rankings and measured influence ranged from 0.349-0.354 (advisor recommendation) to 0.431-0.580 (prompt monitoring). Uncited factors scored above the lowest cited factor in up to 57.6% of advisor responses, showing cited top-three factors do not reliably identify the most influential inputs.

arXiv cs.AI / cs.LG / cs.CLSep 4, 17:37 UTC · 12d agoAI safety & security1

Fresh-Challenge VDF Attestations for Model-Relative Response Latency

Fresh-Challenge VDF Attestations bind verifiable delay functions to unpredictable public challenges, yielding succinct evidence of model-relative response latency.

The paper specifies FCLA, a protocol composition that binds a VDF to an unpredictable public challenge, a message, and independently auditable release/receipt records. Under explicit assumptions about VDF sequentiality and a calibrated bound on an adversary's sequential evaluation rate, an accepted transcript is inconsistent with post-challenge generation. A benchmark of the public reference implementation confirms the expected evaluation-versus-verification separation on one documented machine. The contribution is a protocol design analysis, not a new VDF construction.

arXiv cs.CRSep 11, 11:25 UTC · 5d agoResearch

Lightning Weave: Improving the Accuracy-Efficiency Frontier of Reasoning Models through Capability Composition

Lightning Weave composes capabilities from independently post-trained models via on-policy distillation, improving Qwen3.5-4B reasoning accuracy while cutting tokens.

Lightning Weave is a post-training framework that merges accuracy and efficiency capabilities from independently post-trained specialist models into a single student via on-policy distillation. Each capability is represented as a policy shift, combined via aligned log-ratio shifts and Tilted-Target DOPD, enabling training without serving multiple live anchor models concurrently. On Qwen3.5-4B, it raises HMMT 2025 accuracy from 59.2% to 64.0% with 10.7% fewer response tokens, and LiveCodeBench v5 accuracy from 41.7% to 54.2% with 9.6% fewer tokens. The authors report a state-of-the-art accuracy-efficiency Pareto frontier across diverse students and math/code benchmarks, with code planned for release.

Hugging Face daily papersSep 12, 20:00 UTC · 3d agoAI research

When Should LLMs Abstain? Chain-of-Self-Questioning for Selective Risk Control

Chain-of-Self-Questioning prompting cuts LLM wrong-answer commitments 32% relative while raising answered accuracy, holding across eleven model families.

The paper introduces Chain-of-Self-Questioning (CoSQ), a prompt-only framework that makes LLM answer commitment conditional on an explicit assessment of the information required to answer. On an 817-item TruthfulQA multiple-choice set, Grounded-CoSQ at τ=0.90 reduced mean unconditional wrong-commitment rate from 13.1% under chain-of-thought to 8.9% (a 32.1% relative reduction), while raising answered accuracy from 86.9% to 89.7% at 87.6% coverage. Improvements held across eleven open-weight and hosted model families and at every evaluated threshold, with convergent evidence from a Natural Questions short-answer evaluation.

arXiv cs.AI / cs.LG / cs.CLSep 15, 17:52 UTC · 23h agoAI research

Can Foundation Models Moderate Online Content? Evaluating Instruction- vs. Example-Driven Policy Operationalization

ModerationBench shows foundation models can nearly triple Bluesky's moderation F1 (0.60 vs 0.22), with instruction- and example-driven guidance performing comparably.

Researchers built ModerationBench, a new benchmark of 4,000 manually annotated in-the-wild posts from Bluesky, to test whether foundation models can reliably operationalize content moderation policies. They systematically compare instruction-driven guidance (reasoning from policy precepts) with example-driven guidance (generalizing from precedents) for Vision-Language Models. Both paradigms achieve comparable peak effectiveness, and foundation models nearly triple the F1 of Bluesky's deployed moderation system on Random Posts (0.60 vs 0.22).

arXiv cs.AI / cs.LG / cs.CLSep 9, 16:30 UTC · 7d agoAI research1

Safety for Whom? Boundary-Aware Self-Distillation for Controlled LLM Safety Refusal

A self-distillation safety framework tunes narrow-boundary refusals in Qwen3-8B, raising target-domain refusal to 84.75% while cutting over-refusal from 15.20% to 5.20%.

The paper formulates narrow-boundary safety, where deployments need refusals within specific topics rather than whole subjects, and proposes an offline self-generated framework with controlled topic generation, escalating retries, and harmful-benign boundary pairs. On political persuasion with Qwen3-8B, the method raised target-domain refusal from 9.47% to 84.75% and cut the mean unsafe-response rate across three broader benchmarks from 26.26% to 0.14%. Verified target-model responses reduced over-refusal from 15.20% to 5.20%, and boundary-pair data cut comply-side over-refusal on held-out pairs from 32.94% to 4.16%. Results show data composition controls the safety-usability trade-off and alignment should be evaluated on both sides of the refusal boundary.

Hugging Face daily papersSep 2, 20:00 UTC · 13d agoAI safety & security1

When LLM judges agree, should we believe them?

Amazon ICML paper uses Ising models to correct correlated LLM-judge votes, beating accuracy-weighted panels by 9-14%.

Amazon Science describes an ICML paper, "Dependence-aware label aggregation for LLM-as-a-judge via Ising models," addressing how correlated judge outputs inflate majority-vote confidence. The unsupervised method models pairwise dependence between judges, learning both reliability and similarity without human reference labels. Tested on relevance, toxicity, and summarization tasks with 10 judge models at temperature zero, it outperformed accuracy-weighted voting by 9% to 14%.

Hacker News · AISep 14, 16:29 UTC · 2d agoAI research HN 41↑ · 23 comments

My business partner sent a 5K vibe-coded PR that he didn't even test

A developer's business partner shipped a 5,236-line untested vibe-coded payments backend PR whose endpoints failed basic testing.

The author describes reviewing a pull request with 5,236 additions for a payments backend that a business partner generated largely with AI in a single day without testing. The PR's AI-written documentation included redundant boilerplate (e.g., 'returns 400 on error') but omitted operational details like where to obtain API keys, and the endpoints failed when tested. The post is a critical opinion piece on vibe coding and perceived skill atrophy among developers who rely on AI for everything.

Hacker News · securitySep 8, 01:48 UTC · 8d agoAI industry HN 46↑ · 75 comments1

Cross-Model Agreement as a Deployment-Time Reliability Signal for Automatic Polyp Segmentation

Referee-Based Quality Estimation flags unreliable polyp segmentations at inference without ground truth, reaching ROC-AUC 0.960 with SegFormer-B0 referees.

RBQE measures agreement between a primary segmentation model and an independently trained referee on a 1,223-image external benchmark drawn from four public datasets. A cross-architecture SegFormer-B0 referee achieves the strongest signal (ROC-AUC 0.960), beating a Test-Time Augmentation baseline by 0.055 ROC-AUC under an identical protocol. Excluding trivially separable empty-mask cases, ROC-AUC falls to 0.876 (SegFormer-B0) and 0.783 (same-architecture control), but RBQE's margin over baselines widens. Progressive rejection of low-agreement predictions increases mean Dice of retained outputs, supporting selective prediction at the cost of one extra forward pass.

arXiv cs.AI / cs.LG / cs.CLSep 9, 17:32 UTC · 7d agoAI research

SNAP3D: Physically Grounded 3D Parts for Assembly from a Single Image

SNAP3D uses physics simulation feedback to make single-image 3D part generation produce valid, stable assemblies, validated through 3D printing.

The framework improves part-aware 3D generation by resolving inter-part penetration, recovering contact graphs between neighboring parts, and placing parameterized connectors at contact surfaces. Physical simulation feedback refines connector placement, orientation, and dimensions to improve assembly stability while preserving geometry. A physics-based evaluation protocol tests assembly validity and stability under gravity, and results are validated through 3D printing and real-world assembly.

Hugging Face daily papersSep 10, 20:00 UTC · 5d agoAI research

Target leakage, not model class, explains reported accuracy in survey-based cardiovascular screening: a leakage-tiered audit of glass-box and tabular foundation models

Audits of 10 classifiers on BRFSS show target leakage, not model class, drives the reported 0.89 AUROC in survey-based cardiovascular screening.

The study benchmarks ten model classes, including glass-box and tabular foundation models, for prevalent myocardial infarction on 442,067 respondents of the 2022 BRFSS across five feature tiers of decreasing leakage risk. Removing two post-diagnostic features costs every model 0.049-0.051 AUROC and collapses performance into a 0.0045-wide band, and the explainable boosting machine matches all alternatives within 0.005 while scoring roughly 104x faster than the strongest foundation model. Frozen models transport within 0.002 AUROC to 2023 data; the authors conclude evaluation practice and feature sets, not model capacity, are the binding constraint.

arXiv cs.AI / cs.LG / cs.CLSep 10, 17:29 UTC · 6d agoAI research1

Verifiable by Construction: Claim-Level Evaluation of Verbatim Citation in Clinical Question Answering

Evaluation of twelve LLMs on 222 clinical questions shows verbatim quotes rarely substantiate claims; claude-opus-5 fully substantiates only 37.1%.

The authors build a standardized harness over four clinical practice guidelines and evaluate twelve LLMs on 222 synthetic clinical questions, measuring citation attachment, verbatim quote production, and claim substantiation. Most models attach verbatim quotes to over 90% of claims from prompting alone, though lightweight models like claude-haiku-4.5 struggle. Quotes frequently fail to substantiate claims: claude-opus-5 quotes 98.0% of claims but fully substantiates only 37.1%, exposing a capability gap for verifiable clinical QA.

arXiv cs.AI / cs.LG / cs.CLSep 14, 17:53 UTC · 1d agoAI research

The Model Proposes, the Code Disposes: A Pre-Registered Ablation of a Verifier-and-Acceptance Stage in an LLM-Orchestrated Offensive-Security Agent

Pre-registered ablation finds a model verifier stage in an LLM offensive-security agent suppresses findings; removing it eliminated suppression with precision tradeoff.

The paper evaluates a verifier-and-acceptance stage in an LLM-orchestrated offensive-security agent via a pre-registered 20-run confirmatory ablation and a 2x2 factorial study with 40 runs on vulnerable lab targets. Removing the stage eliminated pre-report suppression (median 2 vs 0 findings, p = 0.00003) but reduced model-blinded shipped precision (0.471 vs 0.353, p = 0.0087). Suppression was attributed to the model verifier rather than deterministic acceptance rules, and an instrumented canary recorded zero external contacts in all 60 runs. The full design retained 93.8% of model-adjudicated true candidates but failed its pre-registered non-inferiority floor of 0.90.

arXiv cs.CRSep 14, 17:13 UTC · 2d agoResearch

Technical Manual for a Toolkit for Measuring Contextual Individuation in Transformer Language Models

An open methodology toolkit measures whether transformer language models contextualize fixed word forms across domains using bridge forms and layer-wise silhouette analysis.

The manual documents an open toolkit built around 'bridge forms' - identical written words recurring across two or more subject domains with a different sense in each - to test whether transformer language models individuate word occurrences by context beyond the embedding layer. It covers declarative specification of bridge forms, Wikipedia corpus acquisition, occurrence localization, layer-wise representation extraction, domain-pairwise silhouette measurement, and visualization, justifying each choice against failure modes such as sense contamination and subword-tokenization misalignment. It is a methodological and implementation reference and reports no empirical results.

arXiv cs.AI / cs.LG / cs.CLSep 4, 16:38 UTC · 12d agoAI research

Why AI raises the stakes for exposure validation

Fal.Con 2026 commentary argues AI accelerates vulnerability discovery and exploitation, making evidence-based exposure validation essential for defender prioritization.

CSO Online reports on the exposure-validation theme at CrowdStrike's Fal.Con 2026 conference, where CEO George Kurtz described AI as the new cyber battlefield and emphasized AI red teaming and continuous security. The piece argues that as AI speeds up vulnerability discovery and exploitability analysis on both sides, teams must determine which exposures are actually exploitable in their environments—chained weaknesses, credential abuse, lateral movement, privilege escalation—rather than chasing theoretical risk. It points readers to Horizon3's conference perspective.

CSO OnlineSep 11, 21:49 UTC · 4d agoIndustry

Cadence: Error-Bounded Lossy Compression of Demand Time Series with a Time-Series Foundation Model

Cadence pairs Google's 330M-parameter TimesFM-3 foundation model with adaptive arithmetic coding, gaining 13-28% on 2026 demand series over classical predictors.

Cadence is an error-bounded lossy compressor for numeric time series combining the 330M-parameter Google TimesFM-3 foundation model with an adaptive arithmetic coder, guaranteeing a per-sample error bound. On 49 EIA-930 balancing-authority demand series from 2026 it gains 13.3% over the best of six classical predictors and 28.3% on 50 MTA ridership series, winning all 297 series-tolerance pairs with a 21.4% median gain. The paper also reports negative results, including that foundation models add negligible value for lossless coding and that PyTorch predictions are not bit-identical across batch sizes.

Hugging Face daily papersSep 4, 20:00 UTC · 11d agoAI research1

K-Bench: a clinically calibrated benchmark for evaluating large language models in high-risk mental health conversations

Clinician-calibrated K-Bench evaluates 125 LLM configurations on 200 high-risk mental health vignettes, exposing wide variation in suicide and violence risk handling.

K-Bench is a clinician-calibrated, protected benchmark evaluating 125 model configurations from 33 base models across 14 providers on 200 multi-turn vignettes covering suicide, self-harm, domestic violence, substance misuse and no-risk presentations. A frozen GPT-4o judge achieved 94.2% exact agreement with clinician consensus across 6,751 eligible comparisons from 151 clinician-rated transcripts. Leading models combined supportive conversation with combined-risk scores above 95, while risk exploration varied substantially among weaker configurations; therapeutic prompting helped weaker models and elevated reasoning produced no average improvement. A continuously updated public leaderboard is hosted at k-bench.ai with protected test materials.

arXiv cs.AI / cs.LG / cs.CLSep 14, 16:49 UTC · 2d agoAI safety & security

Emergency Department Revisit Quality Review Screening: Exploring Human Decision-Making and Artificial Intelligence Support

A retrospective study found GPT-4 over-flagged emergency department revisit cases while an LLM knowledge-graph screener achieved 83-100% positive predictive value.

In an exploratory retrospective study of 99 emergency department diagnosis pairs from a multihospital health system, clinicians and GPT-4 independently judged whether revisit pairs warranted further assessment. GPT-4 responses correlated poorly with clinicians, flagging 94% of pairs for follow-up, 4.4-13.3 times more than clinicians, though prompt engineering was minimal. An algorithm leveraging an LLM-populated knowledge graph (KGA) achieved 83-100% positive predictive value against at least one clinician rater, suggesting LLM-based screening could broaden revisit quality review without substantially increasing reviewer workload.

arXiv cs.AI / cs.LG / cs.CLSep 9, 16:37 UTC · 7d agoAI research1

Evaluating Time-Series Foundation Models and Multimodal Dietary Context for CGM Forecasting

Study finds zero-shot time-series foundation models underperform on CGM forecasting; fine-tuned Chronos-Bolt cuts RMSE up to 18.4% and dietary context adds signal.

The paper evaluates time-series foundation models for continuous glucose monitoring forecasting across eight public datasets covering Type 1 diabetes, Type 2 diabetes, and non-diabetes populations. Under a unified protocol, zero-shot foundation models did not consistently outperform baselines like Elastic Net and PatchTST, but lightweight fine-tuning did, with fine-tuned Chronos-Bolt reducing RMSE by 6.5%-18.4% in the T1D cohort and 8.6%-18.2% in the non-diabetes/T2D cohort. A residual-based fusion framework adding dietary context from CGMacros reduced overall RMSE by about 3% and postprandial RMSE by about 15% versus CGM-only baselines.

arXiv cs.AI / cs.LG / cs.CLSep 10, 17:43 UTC · 6d agoAI research

Learning Length-Extrapolatable Recurrent Models

Researchers propose Credit Stabilization through Time, a training method letting recurrent models extrapolate up to 128x their training length.

The paper argues that length extrapolation failure in BPTT-trained recurrent models is better explained through state credit, the signal through which future losses reach earlier recurrent states. It introduces Credit Stabilization through Time (CST), which locally rescales the state-credit signal during backpropagation without rotating the corrected component or changing forward computation. Controlled experiments show improved performance beyond the training horizon, with gains at up to 128x the training length.

arXiv cs.AI / cs.LG / cs.CLSep 8, 17:59 UTC · 7d agoAI research1

CARDEA: Auditable Reasoning Grounded in Spatial Evidence for End-to-End Coronary Angiography Interpretation

CARDEA, a vision-language model trained only on public data, matches cardiologists on coronary angiography complexity assessment while exposing auditable bounding-box evidence.

CARDEA is a unified large vision-language model serving as the inference core of an end-to-end coronary angiography pipeline from multi-view videos to study-level diagnosis. It was trained on public datasets through visual alignment, self-distilled Chain-of-Box cold start, and reinforcement learning with verifiable rewards encouraging bounding-box reasoning. It reached 0.91 accuracy on dominance classification under domain shift and 0.90 on complexity assessment, comparable to two interventional cardiologists. RLVR raised zero-shot report generation vessel-severity macro-F1 from 0.513 to 0.686, while supervised imitation alone did not.

Hugging Face daily papersSep 6, 20:00 UTC · 9d agoAI research

When Models Edit Too Much: On the Fidelity of Minimal Code Edits

A 400-task BigCodeBench evaluation shows frontier LLMs widely over-edit code; a preservation instruction cuts excess edits and raises Pass@1 by 2.3 points.

Researchers built an evaluation framework from 400 BigCodeBench problems with injected AST-level corruptions, each with a known minimal patch, to measure over-editing in LLM code repair. Even strong models like GPT-5.5 produce unnecessarily large edits despite high Pass@1. Adding a preservation instruction reduced average excess Levenshtein distance from 0.195 to 0.131, cut added cognitive complexity by 26.6%, and raised Pass@1 by 2.3 points. Reinforcement learning post-training gave the best out-of-domain edit-fidelity trade-off, while supervised fine-tuning overfit to seen corruption patterns.

Hugging Face daily papersSep 2, 20:00 UTC · 13d agoAI research1

From Reweighting to Rewriting: Unlocking the Intervention Effects of Influential Samples in Training Data Attribution

Study shows rewriting responses of influence-selected training examples shifts LLM behavior more strongly than reweighting the same samples.

The paper examines training data attribution, arguing that influence functions identify high-leverage examples whose value goes unrealized under conventional weight-based reweighting interventions. It introduces influence-guided response rewriting, which replaces the responses of influence-selected examples with behavior-aligned or behavior-opposed supervision while keeping instructions fixed, tested across four open-weight LLMs using epistemic abstention as the primary testbed. Rewriting produces stronger, more persistent, and bidirectional behavioral shifts, including on safety refusal, while reweighting the same examples yields weak, inconsistent effects. The results motivate intervention-aware evaluation of TDA methods.

Hugging Face daily papersSep 1, 20:00 UTC · 14d agoAI research

AI Changed the Exposure Problem. Validation Needs to Change With It.

Picus Security argues vulnerability validation must combine exploitability, control validation, and agentic pentesting as AI accelerates disclosure volume.

Picus Security reports 35,853 CVEs were published in H1 2026, roughly 49% more than the prior year, while only 495 were catalogued as exploited in the wild and 116 were attacked on disclosure day. The vendor argues CVSS-based triage is inadequate and promotes combining exploitability validation, security control validation, and agentic pentesting into one program. The post also cites Anthropic data showing Mythos-class models surfaced 26,153 open-source vulnerability candidates with only 421 patched upstream, and promotes Picus's Validation Summit '26 on October 14-15.

The Hacker NewsSep 14, 11:58 UTC · 2d agoIndustry

xDailyBench: Benchmarking LLMs on Professional Consultation for Real-Life Problems

xDailyBench tests 11 frontier LLMs on 248 real-life consultation tasks; the best models score 75.6% and lag on implicit requirements.

The benchmark spans 51 scenarios across personal life, white-collar work, learning and research, and cross-domain activities, grounded in requests users actually completed or intended to complete with AI. Tasks are scored with fine-grained binary rubrics covering explicit and implicit requirements under standardized agentic settings. Across 11 frontier models, the best achieved a 75.6% task-level score, with all models performing at least 9 percentage points worse on implicit than explicit requirements.

arXiv cs.AI / cs.LG / cs.CLSep 7, 17:30 UTC · 9d agoAI research