Search: “benchmarking”

[AINews] GPT-6 Astra: OpenAI’s biggest LLM launch of all time

OpenAI launched GPT-6 Astra, its new flagship model, claiming state-of-the-art computer use, software engineering, math, and cybersecurity capabilities.

OpenAI launched GPT-6 Astra as its new flagship model, describing it as its most intelligent and aligned model with state-of-the-art computer use, software engineering, and math/science capabilities. Pricing is $10/$50 per 1M input/output tokens standard ($20/$100 fast tier), rolling out first to limited organizations, then ChatGPT Plus/Pro/Business/Enterprise, the API, and AWS. OpenAI claims 99.9% on ARC-AGI-3, 98% on FrontierMath Tier 4, and 100% on ExploitBench. Artificial Analysis scored Astra 67 on the Coding Agent Index and 61 on the Intelligence Index, behind Claude Fable 5.1, and the system card drew attention for reporting decreased chain-of-thought monitorability despite alignment gains.

Latent Space·Sep 4, 05:18 UTC · 12d agoModel release3

Benchmark Radar: A Living Database and Search Engine for AI Benchmarks and Evaluation

Benchmark Radar provides a living searchable database of 1,283 AI benchmark records and 12,916 score observations drawn from 37 daily discovery sources.

Benchmark Radar combines daily discovery of benchmark papers, repositories, datasets, and releases from 13 direct connectors and 24 first-party feeds into a searchable catalog with model card mentions and score histories. The catalog contains 1,283 source records drawn from 4 benchmark catalogs plus 12,916 numeric observations on 790 records. The release includes a web dashboard with leaderboard, Pareto frontier of score versus usage, saturation and trend views, daily feeds, a CLI, and reproducible analysis. The paper audits the full catalog and examines benchmark saturation and limits of score comparisons.

Hugging Face daily papers·Sep 9, 20:00 UTC · 7d agoAI research

nex-agi/Nex-N2.5-mini — new model trending #30 on Hugging Face

Nex-AGI releases Nex-N2.5 agentic model family (mini, Pro, Max) with a 1.6-trillion-parameter MoE Max, open weights, and hosted access via OpenRouter.

Nex-AGI launched Nex-N2.5, a family of agentic models in mini, Pro, and Max sizes, with the Max version built on a 1.6-trillion-parameter text-only Mixture-of-Experts foundation and the company's first complete post-training effort at trillion-parameter scale. The models target long-horizon computer use, web browsing, and visually grounded agentic tasks, with expanded agent training environments. Reported benchmarks include Max scoring 86.1 on Terminal-Bench 2.1 and 65.7 on SWE-Bench Pro, trailing Claude Opus 5. Weights are being released openly on Hugging Face and ModelScope, with hosted access through OpenRouter.

Hugging Face trending models·Sep 8, 05:08 UTC · 8d agoModel release1

IdeaAMBIG: Benchmarking Implementation-Critical Gaps in Research-Idea Specifications

IdeaAMBIG benchmark with 660 instances measures whether LLMs can spot and fix underspecified research-method details for faithful implementation.

Researchers introduce IdeaAMBIG, a benchmark of 660 evidence-grounded instances (163 real-world gaps from reproducibility reports and GitHub issues, 497 controlled synthetic gaps) built from papers, codebases, and reproduction artifacts. It evaluates codification-readiness assessment, defect localization, and clarification action generation. Across 13 LLMs, the best model achieved only 9.6% Macro Defect Recovery Rate on real-world instances but 80.6% clarification success when given the annotated defect. An oracle study showed gold resolutions raise the codification-ready rate from 14% to 98%, identifying defect localization as the main bottleneck.

arXiv cs.AI / cs.LG / cs.CL·Sep 9, 17:59 UTC · 7d agoAI research2

YuE2 · Frontier Music with Symbolic Planning

YuE2, a 3.59B-parameter music generation model, scores 6.9632 on SongBench, beating Suno v5 via symbolic planning.

YuE2 is a music generation model of roughly 3.59B parameters and 28 layers supporting song creation, covering, and agentic editing through editable ABC symbolic scores. Its best-of-8 setting reaches 6.9632 on SongBench, the highest mean among 15 evaluated settings on WildSongBench (192 prompts), ahead of Suno v5 at 6.8721. The project also introduces MERT2, whose 632M-parameter encoders achieve state of the art on 14 of 15 MARBLE metrics, and SheetSage2, which transcribes beats, downbeats, key, chords, structure, and melody with SOTA on 10 of 13 benchmark metrics.

Hacker News · AI·updated Sep 11, 03:35 UTC · 5d ago·first Sep 11, 00:33 UTC · 5d agoModel release 2 sourcesHN 43↑ · 35 comments

BenchMIRT: What are LLM benchmarks actually measuring?

AllenAI's BenchMIRT blog post examines what LLM benchmarks actually measure and their reliability.

AllenAI published a Hugging Face blog post introducing BenchMIRT, which investigates what large language model benchmarks actually measure. No article text is available, so specific findings, methods, or benchmark scores cannot be extracted. The work appears to target benchmark validity, a live concern for model evaluation and comparison.

Hugging Face Blog·Sep 1, 21:39 UTC · 15d agoAI research

GPT-6 Astra appears to show a "step change" in spatial reasoning based on early benchmarks

Early StationeryBench robotics results show OpenAI's GPT-6 Astra far ahead of Ai2's MolmoAct2 at dual-arm manipulation, completing 7 of 100 tasks versus zero.

A new robotics benchmark called StationeryBench tested OpenAI's GPT-6 Astra against Ai2's MolmoAct2 on five desk-object tasks using identical dual-arm YAM robots over 200 trials. Astra fully completed 7 of 100 tasks with a median progress score of 46 out of 100, while MolmoAct2 completed zero with a median score of 12. Cornell and Google DeepMind researcher Yoav Artzi called the result a 'step change in spatial reasoning' and noted Astra approaches human-level accuracy on the unpublished REMAP benchmark. He speculated OpenAI trained the model on large amounts of 3D data such as Blender scenes, and OpenAI reportedly plans consumer robots.

The Decoder·Sep 12, 14:26 UTC · 4d agoAI industry 2 sources1

IdeaAMBIG: Benchmarking Implementation-Critical Gaps in Research-Idea Specifications

IdeaAMBIG benchmark of 660 specification-gap instances shows LLMs localize implementation-critical research gaps poorly, with best model at 9.6% defect recovery.

IdeaAMBIG is a benchmark of 660 evidence-grounded instances evaluating whether research-method specifications provide enough information for faithful implementation: 163 real-world gaps from reproducibility reports and GitHub issues plus 497 controlled synthetic gaps. It tests codification-readiness assessment, defect localization, and clarification action generation across 13 LLMs. The best model achieves only a 9.6% Macro Defect Recovery Rate on real-world instances, though 80.6% clarification success when given the annotated defect, and an oracle study shows gold resolutions raise codification-ready rates from 14% to 98%. Defect localization emerges as the main bottleneck across all evaluated models.

Hugging Face daily papers·Sep 8, 20:00 UTC · 8d agoAI research

GPT-6 Astra gives mathematicians a breather, and OpenAI says that's by design

OpenAI's GPT-6 Astra tops ulam.ai's ErdosBench math benchmark with 106 of 226 problems solved, while the company prioritizes recursive self-improvement over math optimization.

OpenAI's GPT-6 Astra leads ulam.ai's ErdosBench with a score of 3.23, solving 106 of 226 open math problems and fully solving 43, ahead of GPT-5.6 Sol's 78 solved problems. Chief scientist Jakub Pachocki said OpenAI deliberately avoided targeted math optimization to prioritize recursive self-improvement and automated alignment research. Benchmark developer Przemek Chojecki estimated the gain at 5-10% across tested math-research skills. Mathematician Terence Tao warned at the 2026 International Congress of Mathematicians that AI-generated proofs could shift mathematics from proof scarcity to proof overload.

The Decoder·updated Sep 12, 20:00 UTC · 4d ago·first Sep 10, 13:45 UTC · 6d agoAI industry 10 sources1

StochBench: A Domain-Specific Benchmark for Stochastic Processes in Lean

StochBench introduces 450 graduate-level stochastic-processes problems in Lean 4; an Opus 4.8-based agent proved 34.9% under a 15-minute limit.

StochBench is a Lean 4 benchmark of 450 graduate stochastic-processes problems, each paired with its natural-language source, covering Markov chains, renewal processes, martingales, Brownian motion, stochastic calculus, weak convergence, and Poisson and continuous-time Markov processes. The benchmark addresses field-specific applied mathematics underrepresented in Mathlib, unlike competition-math-dominated suites such as IMO and Putnam collections. An Opus 4.8-based agent achieved a 34.9% proof rate (157/450) under a 15-minute per-problem limit, showing the benchmark remains challenging for advanced provers.

Hugging Face daily papers·Sep 7, 20:00 UTC · 9d agoAI research

[AINews] 10% worse, 100x cheaper, 10000x faster: Why Simulation is taking over

Latent Space argues AI training pipeline stages—rewards, data, teachers, curricula, environments—are flipping from human-made to model-made simulation.

Latent Space's AINews essay traces how each component of AI training has turned synthetic since 2022: reward models (InstructGPT, RLAIF), synthetic pretraining data (Microsoft Phi, NVIDIA Nemotron-4 340B), model teachers (Alpaca, DeepSeek-R1 distillation), and self-generated curricula (Self-Rewarding Language Models, SPIN). In 2026 it highlights Karpathy's autoresearch loop—700 experiments yielding 20 kept improvements, cutting GPT-2 training time from 2.02 to 1.80 hours—and Z.ai's GLM-5.3 fully synthetic RL environment, judging, and verification stack. It frames these shifts as 'simulation': 10% worse but 100x cheaper and 10,000x faster than human equivalents.

Latent Space·Aug 22, 07:36 UTC · 25d agoAI industry

Measuring benchmark optimization in speech recognition

Hugging Face examines how much speech recognition systems overfit benchmarks and how to measure benchmark optimization in ASR.

A Hugging Face post on measuring benchmark optimization in automatic speech recognition, analyzing how model improvements on benchmarks reflect genuine capability gains versus overfitting. It is evaluation methodology research with no direct security impact.

Hugging Face Blog·Aug 21, 00:00 UTC · 26d agoAI research

IndicTriMix: Developing Language Identification Datasets and Models for Tri-Language Code-Mixing

Researchers release IndicTriMix benchmarks and fine-tuned MuRIL and XLM-RoBERTa models for token-level language identification in tri-language code-mixed text.

The paper formulates token-level language identification in code-mixed text as a sequence labeling task and fine-tunes MuRIL and XLM-RoBERTa transformer models for Indian languages. It evaluates on Hindi, Gujarati, and Bengali configurations with manually annotated test sets and proposes two code-mixed generation approaches using parallel trilingual sentences. A public benchmark, annotated test sets, and fine-tuned models are released for reproducibility.

arXiv cs.AI / cs.LG / cs.CL·Sep 10, 17:36 UTC · 6d agoAI research1

MetroLLM-Bench: Evaluating Language Models as Transit Kiosk Runtimes

MetroLLM-Bench is a 955-case benchmark testing language models as transit kiosk tool-calling runtimes across six real metro systems.

The benchmark covers 37-414-station metro systems and eleven task categories including routing, fare calculation, disruptions, accessibility, and adversarial input, with 14 deterministic and 8 semantic scoring components. Of 26 models from six vendors, a PEFT-tuned 4B Qwen 3.5 student scored 91.3 on Tier 1, exceeding GPT-5.6 (90.6/90.0), while Muse Glimmer 30B led the composite ranking. A deterministic rule-based baseline reached 84.6, and PEFT gains over base models shrank from +7.03 points at 2B to -0.91 at 27B.

Hugging Face daily papers·Sep 8, 20:00 UTC · 8d agoAI research

Claude Fable 5.1 made me a really nice animated pelican

Anthropic launched Claude Fable 5.1, claiming gains in coding and long-running tasks, with 52.6% on Terminal-Bench-Science 0.1.

Anthropic released Claude Fable 5.1 (alongside Mythos 5.1), positioning it as a new standard for coding, knowledge work, and long-running problem-solving. The model scores 52.6% on Terminal-Bench-Science 0.1, up from 24.7% for Fable 5, versus 29.0% for Opus 5 and 22.4% for GPT-5.6 Sol. Simon Willison's hands-on test found the model produced an impressive animated pelican, though he notes other benchmarks show only slightly improved scores.

Simon Willison·Sep 1, 23:57 UTC · 14d agoModel release1

VDiff-Bench: A Challenging Benchmark for Fine-Grained Image Difference Identification

VDiff-Bench, a 1,756-question benchmark, shows multimodal LLMs struggle with fine-grained image-difference identification, scoring as low as 8.7% on low-level changes.

VDiff-Bench is a multiple-choice benchmark of 1,756 four-way questions over image pairs covering 10 change categories including position, motion, color, texture, OCR/text and illumination, with curated hard negatives. Evaluation of 11 state-of-the-art open- and closed-source MLLMs shows fine-grained visual comparison remains brittle: 7-8B-scale open-source models score 52.5-70.6% on semantic changes but only 8.7-33.3% on low-level changes like noise and texture. Notably, Grok 4.3 shows a sharp performance drop on noise and texture differences, falling behind large open-source models like Kimi K2.5 and K3.

Hugging Face daily papers·Sep 4, 20:00 UTC · 12d agoAI research

OpenAI Pauses Astra Model Over Critical Cybersecurity Risk Concerns

Security Affairs·Aug 10, 11:16 UTC · Aug 10, 2026Exploit / PoC1

Benchmark Scores Are Pipeline-Dependent: A Reliability Audit of Cybersecurity LLM Benchmarks

Audit of eight cybersecurity LLM benchmarks shows evaluation pipeline choices can swing scores by over 80 points and reshuffle most model rankings.

Researchers modeled eight cybersecurity benchmarks as configurable measurement pipelines and audited 10 proprietary, open-weight, and cybersecurity-specialized LLMs. They identified 15 systematic failure modes and showed a single pipeline choice can change a model's score by more than 80 percentage points and alter rankings; semantically similar task pairs rank the same models differently. Under a standardized harness, nine of 10 models shifted at least three ranks on at least one benchmark, motivating pipeline-aware auditing for reliable model evaluation.

arXiv cs.CR·Sep 8, 14:01 UTC · 8d agoAI research1

Show HN: Pelican-bicycle alternatives (updated for 2026)

Hobbyist benchmark re-runs the pelican-bicycle SVG test on six 2026 frontier models, comparing generation time and API cost per image.

A Show HN post re-runs the classic pelican-bicycle and similar SVG generation tests across six 2026 models: GPT-6 Astra, Claude Fable 5.1, Gemini 3.8 Flash, DeepSeek V4 Pro, Qwen3.8 Max, and Fugu Ultra v2, recording wall-clock time and cost. It also lists 2025 baseline runs with ten models including Claude Sonnet 4.5, GPT-5.2 Pro, and Qwen3-VL-235B-A22B-Thinking. DeepSeek V4 Pro is consistently cheapest ($0.04-$0.10) while Qwen3.8 Max is slowest, taking up to roughly 17 minutes per generation.

Hacker News · security·Sep 14, 13:20 UTC · 2d agoAI researchHN 24↑ · 4 comments

MP-Bench: Evaluating Voice Agents as a Multiparty Conversation Participant

MP-Bench is the first benchmark for voice agents in multiparty conversations, finding real-time agents near chance on turn-taking.

MP-Bench is the first benchmark designed to objectively evaluate conversational speech systems as active participants in multi-party conversations. It assesses agents on turn-taking awareness and response appropriateness, with comprehension-based question-answering as a complementary evaluation. Benchmarking 12 voice agents shows real-time agents score at or below 22% on multiparty comprehension and remain near chance on multiparty turn-taking.

arXiv cs.AI / cs.LG / cs.CL·Sep 11, 17:11 UTC · 5d agoAI research2

LimiX-2: A Contextual Mechanism Network Towards General Structured-Data Intelligence

LimiX-2 scales Contextual Mechanism Networks pretrained via context-conditional masked modeling, beating tabular foundation models on TabArena, TALENT, and BCCO benchmarks.

LimiX-2 is a new tabular model in the LimiX family, developed through model and data scaling guided by previously established scaling laws under the Contextual Mechanism Networks (CMNs) paradigm. It is pretrained with Context-Conditional Masked Modeling (CCMM) on synthetic datasets generated by structural causal models spanning diverse graph structures, functional mechanisms, and observation processes. It outperforms dataset-specific models and tabular foundation models on TabArena, TALENT, and BCCO, and its feature attention encodes direct causal relationships, enabling accurate causal skeleton recovery.

arXiv cs.AI / cs.LG / cs.CL·Sep 15, 17:30 UTC · 1d agoAI research1

StudyBench: Can Self-Evolution Squeeze Textbooks for Olympiad Capability?

Researchers introduce StudyBench, a physics benchmark showing self-evolution gains on textbook problems rarely transfer to olympiad-level questions.

StudyBench is a controlled physics benchmark splitting test data into an Application Set of difficult textbook problems and a Transfer Set of olympiad-level problems. Across three base models, representative self-evolution methods improved on the Application Set but rarely transferred to the harder Transfer Set. A guidance ablation reveals a Guidance Gap, and every method hits a Compute Plateau, indicating the remaining limits are method problems rather than data or compute problems.

Hugging Face daily papers·Aug 31, 20:00 UTC · 16d agoAI research

GPT-6 Astra: an automated AI Engineer you can hire for <$6 an hour

OpenAI launches GPT-6 Astra, a frontier model scoring 97.6% on FrontierMath and 99.9% on ARC-AGI-3, capable of autonomous AI engineering at roughly $6 per hour.

OpenAI launched GPT-6 Astra, described as its first Stargate and lightly looped frontier model, beating Fable 5.1 on many metrics and saturating the hardest FrontierMath (97.6%) and ARC-AGI-3 (99.9%) benchmarks. Latent Space tested the model with over 20 billion tokens, reporting it can train and select models, label data, deploy and debug systems, and orchestrate 20-50 parallel subagents. The authors measured about $6 per hour of agentic engineering at 33 tokens per second, with token efficiency independently confirmed by Artificial Analysis.

Latent Space·Sep 3, 21:09 UTC · 13d agoModel release1

AI models flub these intelligence tests. Can you fare any better?

MIT Technology Review examines puzzle and game benchmarks where current AI models still underperform, probing the limits of machine intelligence tests.

MIT Technology Review explores puzzles and games as benchmarks for gauging AI progress, tracing the practice back to the origins of machine learning in a 1959 article by IBM's Arthur Samuel. The piece highlights intelligence-style tests that today's models still fail and questions what those results reveal about model capabilities. It situates gaming benchmarks within the broader debate over measuring machine intelligence.

MIT Technology Review · AI·Aug 26, 09:00 UTC · 21d agoAI research

Latent-MoE: Domain-Aware Mixture-of-Experts for PDEs with Multi-Regime Physics

Latent-MoE adds domain-aware mixture-of-experts routing to PINNs, improving accuracy over an order of magnitude on multi-regime physics PDEs.

The paper shows standard coordinate networks used in physics-informed neural networks have translation-variant NTKs causing long-range gradient conflicts on PDEs with spatially varying physics. MoE architectures with centered compact-support routers produce a uniformly banded NTK that localizes learning. Latent-MoE interleaves domain-aware MoE blocks in a shared backbone, outperforming FB-PINNs and X-PINNs by over an order of magnitude on multi-stage time-variable physics benchmarks.

arXiv cs.AI / cs.LG / cs.CL·Sep 7, 17:52 UTC · 9d agoAI research

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 papers·Sep 12, 20:00 UTC · 4d agoAI research

Operational Roles of QRNG-Derived Quantum Entropy in Bitcoin Proof-of-Work Architectures

arXiv study finds quantum-random entropy adds no Bitcoin PoW success advantage but helps assurance in fault and provenance scenarios.

The paper shows replacing classical entropy with QRNG output does not change honest Bitcoin proof-of-work success probability when candidate headers remain distinct. It introduces a reproducible benchmark measuring an entropy-efficiency factor and a reboot-diversity index, finding QRNG value only in assurance-oriented scenarios involving correlated restart faults, namespace reuse, and entropy provenance. Validation is simulation-based, with hardware-in-the-loop testing identified as future work.

arXiv cs.CR·Sep 4, 12:47 UTC · 12d agoResearch

SWE-Bench Pro Verified: A Reliable Benchmark for Software Engineering Agents

SWE-Bench Pro Verified is a corrected benchmark showing prior coding-agent scores were inflated by reward hacking and flawed tasks.

Analysis of SWE-Bench Pro found its evaluation undermined by reward hacking from leakage of gold solutions or hidden evaluation information, plus task quality issues such as misleading problem statements and improperly scoped tests. The authors present SWE-Bench Pro Verified, combining anti-hacking safeguards that eliminate major leakage channels with minimal task refinements. Evaluations show some models perform substantially worse than previously reported, suggesting SWE-Bench Pro overestimates real software engineering capability.

Hugging Face daily papers·Sep 7, 20:00 UTC · 9d agoAI research1

Introducing ChatGPT for Financial Services

OpenAI launches ChatGPT for Financial Services, pairing built-in market data with GPT-6 Astra for banking research workflows.

OpenAI introduced ChatGPT for Financial Services, a tailored ChatGPT Work experience shaped by design partners Morgan Stanley and Evercore, targeting investment banking and equity research. It bundles premium data from Daloopa, PitchBook, LSEG News, and Crunchbase hosted on OpenAI infrastructure with granular citations, optimized MCP connectors for S&P Global and FactSet, and 50+ connectors, plus planned entitlement integrations with S&P Capital IQ, LSEG, MSCI, Dow Jones Factiva, and Moody's. It runs GPT-6 Astra, which OpenAI claims is state of the art in information retrieval, financial reasoning, and artifact generation, and includes enterprise controls such as SAML SSO, SCIM, role-based access, and no default training on firm data.

OpenAI News·updated Sep 10, 18:25 UTC · 6d ago·first Sep 10, 07:00 UTC · 6d agoAI industry 3 sources

Embodied-BenchForge: A Closed-Loop Agentic Workflow for Embodied Benchmark Construction

Embodied-BenchForge automates embodied benchmark construction via closed-loop synthesis with verification and repair, yielding seven benchmarks for MLLM evaluation.

Embodied-BenchForge is an agentic framework that transforms user-specified evaluation intents into complete embodied benchmark artifacts via Closed-Loop Benchmark Synthesis. Skill-Orchestrated Artifact Synthesis composes typed reusable skills while an artifact dependency graph records intermediate outputs; Requirement-Guided Verification and Repair triggers local re-execution or upstream rollback on failures. It constructs six Offline EQA benchmarks plus one interactive benchmark with 220 executable tasks, distinguishing MLLM and embodied agent capabilities in observation-based understanding and closed-loop execution.

arXiv cs.AI / cs.LG / cs.CL·Sep 11, 17:16 UTC · 5d agoAI research

Axis Robotics Releases AXIS: A Browser-Based Data Engine With 207 Robot Manipulation Tasks and 50,129 Trajectories

Axis Robotics and academic partners released AXIS, a browser-based teleoperation system yielding 207 manipulation tasks and 50,129 trajectories that lifts pi0.5 to 88.8 on LIBERO-Plus.

A team from Axis Robotics, UC Berkeley, Georgia Tech, and NTU introduced AXIS, a browser-based data engine where contributors teleoperate a simulated Franka Research 3 in a MuJoCo WebAssembly frontend while GPU backends handle task generation, training, and evaluation. The released snapshot holds 207 tasks, 50,129 episodes, and 60K+ task or scene variants from more than 70,000 community contributors. Continual pretraining of pi0.5 on AXIS data raises LIBERO-Plus performance from 83.9 to 88.8, versus 57.5 for a volume-matched RoboCasa365 control; the 2.36 TB dataset is gated for non-commercial academic use.

MarkTechPost·Sep 7, 18:38 UTC · 9d agoAI research

BenchShield: Formal Model-Backed Instrumentation for Reward Integrity in LLM-Agent Evaluation Infrastructure

BenchShield uses lifecycle-model-backed instrumentation to detect reward hacking in LLM-agent benchmarks, lifting full-chain recall to 77-100% at up to 65% lower cost.

The framework grounds reward-hacking detection in a finite lifecycle model of an evaluation's reward-relevant events, combining a static phase-aware taint analysis with runtime infrastructure-side evidence attribution. Evaluation used a human-labeled corpus of 456 adjudicated trajectories drawn from more than 31,000 public agent runs across three benchmarks. BenchShield improves full-chain recall from 23-94% to 77-100% and same-vector coverage from 16-56% to 43-78%, cuts per-task cost by up to 65%, and achieves 96% accuracy detecting reward hacking at runtime.

arXiv cs.CR·Sep 10, 03:10 UTC · 6d agoAI safety & security1

CosmoH2G: A Hand-to-Gripper Transfer Dataset and Baseline Method for Object Manipulation with Complex Spatial Movements

Researchers release CosmoH2G, a 6,189-episode hand-to-gripper dataset with a two-stage method for complex spatial robot manipulation.

The paper introduces a scalable acquisition pipeline using a handheld gripper to collect paired hand-gripper demonstrations, producing 6,189 episodes across 1,254 unique objects with higher spatial complexity than existing benchmarks. A two-stage framework first predicts sparse gripper keyframes (initial and terminal), then generates the full continuous action sequence conditioned on them, while learning gripper orientation and post-optimizing translation via grasping heuristics and kinematic consistency. Simulation and real-robot experiments show stable, precise hand-to-gripper transfer of complex spatial manipulations, outperforming traditional baselines.

Hugging Face daily papers·Sep 6, 20:00 UTC · 10d agoAI research

Recreating Minecraft Is Not a Benchmark

Opinion piece argues viral demos like one-prompt Minecraft recreations are overfit 'demo-benchmarks' measuring preparation, not true model capability.

The author argues that fixed, famous demo tasks (Minecraft builds, SVG pelicans) are trivially optimizable by labs each release cycle, so they no longer differentiate model capability. The piece cites Thinking Machines' Inkling Small scoring within a point of its flagship on the Artificial Analysis Intelligence Index with less than a third of the parameters, and beating it on Humanity's Last Exam, GPQA Diamond, and SciCode. The proposed alternative is rotating or holdout evals such as LiveBench, ARC-AGI's private set, and held-back portions of Humanity's Last Exam.

Hacker News · AI·Sep 6, 14:52 UTC · 10d agoAI researchHN 73↑ · 71 comments

Meta FAIR Introduces AI Research Preference Models (RPMs): Ranking ML Experiments Before Spending GPU Hours

Meta FAIR, Oxford and UCL introduce Research Preference Models that rank unexecuted ML experiments, lifting AIRS-Bench scores from 0.684 to 0.729 and cutting compute ~1.6×.

Researchers from Meta FAIR, Oxford, and UCL introduce Research Preference Models (RPMs), which use frozen pretrained LLMs (Qwen3.6-27B backbone, no fine-tuning) to rank unexecuted experiment candidates and execute only the winner of a pairwise knockout tournament. Two variants shipped: an inference-only LLM-as-a-judge and an agentic variant that runs small pilot experiments in an H200 sandbox. On AIRS-Bench (20 tasks, 24 hours on one H200, 10 seeds), scores rise from 0.684 (random) to 0.711 and 0.729 versus a 0.748 validation oracle, and both variants reach the baseline's 24-hour score in roughly 15 hours. The team reports new SOTA on WinoGrande (94.1% with Agentic RPM) and SVAMP (95.7% with inference-only).

MarkTechPost·Sep 6, 20:25 UTC · 10d agoAI research

VEX-Bench: Benchmarking LLM Agents for Assessing Exploitability of Software Supply Chain Vulnerabilities

Introduces VEX-Bench, 75 expert-labeled real-world cases testing whether LLM agents can assess supply chain vulnerability exploitability; frontier models reach about 80% F1.

VEX-Bench is the first benchmark evaluating LLM agents on assessing whether upstream dependency vulnerabilities are exploitable in downstream projects, with 75 real-world expert-labeled cases across Python, Java, and Go mined from GitHub. Nine models across three agent harnesses were evaluated; GPT-5.5 and Claude Opus 4.6 reach approximately 80% F1 on binary vulnerability-status classification, but only GPT-5.5 surpasses 70% macro-F1 on fine-grained justification classification. The gap highlights the difficulty of moving beyond binary exploitability calls to explaining exploitability reasons, unlike prior benchmarks targeting zero-day settings.

arXiv cs.CR·Sep 7, 22:52 UTC · 9d agoResearch1

ENISA: Frontier AI Is Changing the Speed of Cyberattacks. Europe Needs to Catch Up

ENISA warns frontier AI compresses attack lifecycles to minutes, with exploits possible within 15 minutes of disclosure and median 72-minute breach-to-exfiltration times.

ENISA's July 2026 paper 'ENISA's view on Cybersecurity in the Frontier AI Era' argues AI-assisted attackers may weaponize vulnerabilities within 15 minutes of disclosure and achieve initial-access-to-data-exfiltration in a median 72 minutes, creating a 'negative time-to-exploit' problem. The report cites one organisation whose CVE volume rose from roughly 80 in Q1 2025 to almost 500 in Q1 2026, then about 500 reports per day when frontier-AI tools were used. ENISA recommends machine-speed defence under 'Cybersecurity as Code', EPSS and VEX-based prioritisation, AI-assisted incident response with human oversight, and an assume-breached architecture.

Security Affairs·Sep 14, 20:35 UTC · 2d agoAdvisory

τ^τ-Bench: An Environment for End-To-End, Realistic Agent Construction

New τ^τ-bench tasks coding agents with building deployable customer-service agents; best config, Claude Opus 5, passes only 23.9% of simulations.

Researchers introduce τ^τ-bench, an end-to-end benchmark where a developer agent must build a complete customer-service agent from real business records, a client with requirements, a production API, an inherited codebase, and cost/model limits, then is scored by deploying it against held-out simulated users. Across 53 tasks in four domains, the strongest configuration, Claude Opus 5 under Claude Code, passes just 23.9% of evaluation simulations versus an 82.2% expert-authored reference ceiling. Failure modes mirror those of human developers: shallow queries instead of deep record comprehension, almost no client communication, and shipping the first architecture that runs rather than experimenting.

Hugging Face daily papers·Sep 3, 20:00 UTC · 13d agoAI research

Beyond Top-k Skill Retrieval: Diversity-Aware Skill Routing for LLM Agents

DSR reranks LLM agent skills with Determinantal Point Processes to favor complementary, non-redundant sets, improving multi-skill query coverage.

The paper proposes Diverse Skill Routing (DSR), a diversity-aware reranking framework for LLM agent skill routing that uses a Determinantal Point Process to balance query relevance and non-redundancy across large skill registries. DSR introduces a query-residual diversity kernel that penalizes redundant skill overlap while avoiding penalties arising only from shared query relevance. On the SkillRouter benchmark, DSR improves recall and full coverage over a strong pointwise reranking baseline, with the largest gains on multi-skill queries. The authors argue skill routing should be treated as complementary set selection, not just relevance ranking.

Hugging Face daily papers·Sep 4, 20:00 UTC · 12d agoAI research

Why don't machine learning research agents overfit?

Amazon researchers explain why ML research agents avoid benchmark overfitting, attributing generalization to compressibility of successful strategies.

Amazon Science summarizes the paper "What fits (into few tokens) doesn't overfit: Compression and generalization in ML research agents," which investigates why benchmark hill-climbing loops, whether run by human communities or LLM research agents, do not produce rampant overfitting. The explanation formalizes Occam's razor via a counting argument: successful ML strategies are highly compressible, so short descriptions lack room to memorize benchmark data and must capture real structure. LLM-based agents, being resettable and controllable, allow this hypothesis to be tested empirically.

Hacker News · security·Sep 14, 16:32 UTC · 2d agoAI researchHN 21↑ · 3 comments