VECTOR WIREAI INTELLIGENCE
UTC
Refresh Models Deals Regulatory Sources

Research

TODAY — FRI 04 SEP 2026: 20 in view, led by Scaling Bimanual Household Manipulation from 1,500 hours of Demonstrations to On-Policy Corrections. Qwen3 is the literature’s most-referenced model (4 papers in the loaded window).
UPDATED 1M AGO
TODAY — FRI 04 SEP 202620 in view
◆ FRONTIER·INDUSTRY IMPACT
Scaling Bimanual Household Manipulation from 1,500 hours of Demonstrations to On-Policy Corrections
Learning generalist policies for robust bimanual manipulation is bottlenecked by the scarcity of high quality large scale human demonstration data. In this work, we release 1,500 hours of diverse bimanual manipulation demonstrations covering everyday household tasks, and use this comprehensive corpus to train XR-2, a powerful vision-language-action (VLA) model. Enabled by a purpose built high throughput data pipeline…
bimanual-manipulationvision-language-actionimitation-learningrobotics-datasets
Jiafeng Xu, Qi Li, Yan Shen, et al. (10)cs.ROarXiv ↗PDF ↗
THE INDEX — 19 MORE CLEARED
Clean Engineering, Unstable Measurement: A Preregistered Reliability Failure of Black-Box LLM Observers on Shared Endpointscs.AI · cs.LG
Beyond Small Patches: Black-Box Detection and Purification of Diverse Backdoor Triggerscs.CV · cs.CR
Reducing Catastrophic Risk from AI with Systematic Monitoring and Evaluation of Rogue AI Progressioncs.CY · cs.AI
A Blind Trust, the Bloody Thrust: When Attacker-Controlled Hook Updates Steer AI Agent Harnesses towards Malicious Behaviorscs.CR · cs.AI
Seeing Less Is Not Seeing Safely: Privacy Leakage from Task-Scoped Robot Perception Exportscs.RO · cs.CR
IndicSafeEval: Safety Robustness of Large Language Models under Multilingual Persuasive Jailbreak Attackscs.CL · cs.AI
Privacy-Preserving Topology-Guided Safety for LLM-Based Multi-Agent Systems via Federated Graph Learningcs.CR · cs.AI
A Black Box for Agentic Processes: Blockchain-Anchored Evidence for AI Agent Communication, Human Oversight, and GRC Auditscs.CR
FiMI Banking: A Sovereign Model for Indian Retail Bankingcs.AI · cs.CL
The Natural Language Interaction Protocol and Standard for AI AgentsIBMcs.AI
Mind the Gap: Robustness Risks in PII Detection Systemscs.LG
Flip, Don't Shuffle: Watermarking LLMs at the Speed of Inferencecs.CR · cs.CL
BRIDGE: An Open-Source Humanoid Platform via Morphology-Control Co-Design for Physical AIcs.RO · cs.AI
EraseSAE: Surgical Concept Erasure in Text-to-Video Diffusion Models via Sparse Autoencoderscs.CV · cs.AI
When Optimization Becomes Manipulation: Defending Generative Search against Malicious Generative Engine Optimizationcs.CR · cs.AI
AlcaTRAz - Anchored Tree-Rule Defense Against Jailbreakscs.CR
A Case Study on Emergent Cheating and Whistleblowing in Autonomous Research SwarmsGooglecs.AI
LLaDA-Image: Building Strong Image Generators with Fully Open Training Recipescs.CV · cs.AI
The Illusion of Independent Quorums: Epistemic Fault Domains and Correlated Cognitive Failures in Agentic Quorumscs.DC · cs.MA
THU 03 SEP 202637 in view
◆ FRONTIER·INDUSTRY IMPACT
EvalDetectBench: A Benchmark for Measuring Evaluation Awareness in Frontier Language Models
Frontier large language models can often recognize when they are being evaluated, a capability known as evaluation awareness. If models behave differently in evaluations than in deployment, this undermines the validity of evaluation results, which are a crucial component of current AI safety frameworks. We introduce EvalDetectBench, an open pipeline and benchmark for measuring evaluation awareness that works with any…
evaluation-awarenessllm-evaluationbenchmarksafety-evals
Xinning Li, Kemunto Ochwang'i, Aryasomayajula Ram Bharadwaj, et al. (5)cs.AI · cs.CLarXiv ↗PDF ↗
◆ FRONTIER·INDUSTRY IMPACT
A Finger on the Scale: Covert Policy Steering through Agentic Skills
Reusable agent skills extend large language model (LLM) agents with task procedures, tool-use guidance, and output constraints. Yet these skills also act as externalized behavioral policies, which create a supply-chain risk: a third-party skill may preserve the declared task and valid output interface while covertly redirecting agent decisions toward an undisclosed objective. We formalize Skill Policy Integrity, whic…
agentic-skillspolicy-integritycovert-steeringsupply-chain-risk
Jiarui Li, Jiahao Chen, Chunyi Zhou, et al. (8)cs.CRarXiv ↗PDF ↗
THE INDEX — 35 MORE CLEARED
Stored Is Not Supported: Typed Provenance and Assertion Guardrails for Persistent AI Agentscs.CR
ASCII Attack: Recontextualising Harmful Requests as Artistic Critique in Large Language Modelscs.AI
Agent Memory Is a Surface for Endogenous Authorization Launderingcs.CR · cs.AI
SPADE: SPaT Attack Detection from the Connected Vehicle's Perspectivecs.CR · cs.LG
Privacy Washing: Detecting Internal Contradictions in Privacy Policiescs.CY · cs.CL
Architecting Conversational Data Systems for Stateless LLM APIs: The Hydration Proxy PatternGooglecs.AI · cs.SE
VakyArth: Evaluating Pragmatic Competence in LLMs across Indic Languagescs.CL · cs.AI
READY or Not: Reliable Enterprise Agent Deploymentcs.AI
SEAL: Reinforcing Global Safety in Mixture-of-Experts through Shared Expert ALignmentcs.LG · cs.AI
Counter-GEO-Bench: Evaluating Defenses Against Information-Distorting Generative Engine Optimizationcs.IR · cs.CL
Examining the Vulnerability of Multi-Agent Medical Systems to Human Interventions for Clinical Reasoningcs.AI
text2ql: Multi-Target Natural Language Querying via a Language-Agnostic Intermediate Representationcs.CL · cs.AI
Door-in-the-Face Requests and Refusal Behaviour in Large Language Modelscs.AI · cs.CL
Hearing the Whispers: Black-Box Membership Inference Attacks on Finetuned TTS Modelscs.CR · cs.LG
The Implications of Linguistic Illegibility for LLM Securitycs.LG · cs.CR
Learning-Based Reconstruction Attacks on Coordinate-Obfuscated Point Cloudscs.CR · cs.LG
Humanoid Safe Stop via Learned Stoppability Valuecs.RO · cs.LG
FairLens: Benchmarking Fairness in Vision-Language Models for High-Stakes Decision-Makingcs.CV · cs.LG
Before the Script, Set the Stage: How Worldview Simulation Amplifies Psychologically Grounded Persuasion in Multi-Turn Jailbreakingcs.CL · cs.AI
Implicit Manipulation for Skill Selection in LLM Agents with Semantic Matchingcs.CR
ACLE-MCP: Attested Capability Leases for Execution-Time Trust in Remote LLM Tool Usecs.CR
Beauty is in the AI of the beholder: MLLMs systematically overrate facial attractivenesscs.CV · cs.HC
Toward Explainable and Policy-Aware AI for Carbon Credit Price Prediction: A Research Framework for Emerging Carbon Marketscs.LG
The Shape of Ownership: Verifying LLM Provenance through Semantic Structurescs.CR
FUSE: An Evaluating Framework for Dangerous Capabilities of LLMscs.AI
InfraPatch: Cross-Task Targeted Grayscale Patch Attacks on Infrared-Adapted Vision-Language Modelscs.CV · cs.AI
Skill-as-API: Confidential Multi-Agent Coordination for Agentic Software Engineeringcs.CR
SCX Router: Streaming Zero-Shot Model Selection with a Decoder-KV Classifier and a Real-World Task Ontologycs.AI · cs.CL
Competitive Market Behavior of LLMscs.MA · cs.AI
CodePoisonRAG: Knowledge Poisoning Attacks on Retrieval-Augmented Code Generationcs.CR · cs.LG
Post-Training Ternarization of Qwen3-4B Capability, Effective Bit Budget, Storage Compression, and Deploymentcs.AI · cs.LG
Evaluating ML-based Intrusion Detection Systems: The Illusion of Model Efficacycs.CR
Context Inference Attacks Without Jailbreakscs.CR · cs.LG
WeaveMark: Robust and Scalable Multi-bit LLM Watermarking via Coded Payload Spreadingcs.CR · cs.LG
How Fast Do Agents Rot? An Empirical Study of Long-Horizon Degradation in LLM Agents for Production Decision-MakingMicrosoftphysics.soc-ph · cs.AI
WED 02 SEP 202641 in view
◆ FRONTIER·INDUSTRY IMPACT
Uncovering and Mitigating Aggregation-Induced Reward Hacking in Multi-Reward Reinforcement Learning
Reinforcement learning fine-tuning of large language models increasingly adopts multiple reward dimensions, including verifiable rules, task-specific evaluators, and learned reward models, to provide richer supervision across diverse capabilities. These dimensions are commonly scalarized with fixed aggregation weights. We identify a failure mode in which aggregation itself induces reward hacking: static projection al…
reward-hackingmulti-reward-rlrlhfreward-aggregation
Yu Yuan, Yaoyou Fan, Lili Zhao, et al. (8)cs.CLarXiv ↗PDF ↗Code ↗
◆ FRONTIER·INDUSTRY IMPACT
RePro: Proof-Verified Benchmark Rewriting for Reliable Evaluation of LLM Mathematical Problem Solving
Data contamination undermines the reliable evaluation of large language models (LLMs) on mathematical problem solving. While rewriting-based evaluation mitigates memorization, existing methods lack guarantees of problem validity and answer correctness. We propose Proof-Verified Benchmark Rewriting (RePro), the first framework to integrate Lean-oriented neural automated theorem provers (ATPs) into benchmark rewriting,…
llm-evaluationbenchmark-rewritingproof-verificationlean-atp
Xiyuan Zhou, Zhuoqi Li, Xinlei Wang, et al. (9)cs.CL · cs.AIarXiv ↗PDF ↗Code ↗
THE INDEX — 39 MORE CLEARED
AKRASIA: Stealthy Backdoor Attack on Reasoning-based Code LLMscs.CR
Detecting Hidden Behaviors in LLMs via Activation-matched Finetuningcs.CL · cs.AI
What's in Your Agent's Context? Context Privilege Escalation Attacks against AI Agent Harnesscs.CR
Don't Let the Model Write the YAML: Deterministic, Minimal-Diff GitOps Remediation from LLM-Proposed Field Changescs.SE · cs.AI
Instella-MoE Technical Reportcs.CL · cs.AI
AI Morbidity and Mortality: A Framework for Clinical AI Failure Reviewcs.AI · cs.HC
Membership Inference in Fine-tuned Diffusion Language Models via Token-level Memorization Asymmetrycs.CL · cs.CR
ChatDev 2.0: A No-Code Multi-Agent Platform for Developing Everythingcs.AI · cs.CL
Prediction-Assisted Pricing and Admission for LLM APIs with Stochastic Token Consumptioncs.DS · cs.LG
Validity-Aware Jailbreak Evaluation for Large Language Modelscs.AI
The Privacy-Hallucination Tradeoff in Differentially Private Language Modelscs.AI · cs.CL
ZimaBlue: Evolving Generalizable World Action Models through Scalable Video Pre-trainingcs.CV
CopyShield: A Cross-Level Benchmark of Copyright Defenses in LLMscs.LG
When Safety Routing Breaks: Understanding Alignment Fragility under Benign Fine-Tuningcs.CR · cs.AI
Workload Identification with Physical Side Channels for AI Governancecs.CR · cs.AI
Transferable End-to-End Optimization for Indirect Long-Term Memory Poisoning in LLM Agentscs.CR
Autoresearch for Marketplace Catalogs: From Legacy Forms to AI-Native Matchingcs.AI
SoK: When Safe Agents Fail Together: The Security of Multi Agent LLM Systemscs.CR · cs.AI
LLM-Driven Autonomous Vehicles Inherit Human Driver Biases in Pedestrian Yielding: Results and Implications From A New Benchmarkcs.AI · cs.CL
SilentProbe: Measuring Silent Failure in Production APIs Used as Agent Toolscs.IR · cs.SE
Cheap Verifiers, Large Blind Spots: Measuring the Reliability Cost of Cost-Saving Cascadescs.AI · cs.CR
VerTox: Verifiable Reward-Guided Corpus Poisoning Against Neural Ranking Modelscs.CL · cs.IR
EvoFlint: An Evolutionary Atlas of Multi-Turn LLM Vulnerabilitiescs.CL · cs.AI
When Guardrails Look Effective: Construct Validity Failures in LLM Agent Commerce Evaluationcs.AI
The Irreversibility Budget: Fleet-Level Risk Accounting and Admission Control for Agent Operating Systemscs.AI
Spawn Freely, Act Sparingly: Progressive Risk Vesting for Recursive LLM-Agent Treescs.AI · cs.LG
The Constitutional Coverage Trilemma in AI Governancecs.LG · cs.AI
AgentProv: Auditing Agentic LLM API Providers via Tool-use Policy Probescs.CR · cs.CL
TRIS: A Tri-Layer Retrieval Integrity Sieve Against Knowledge Poisoningcs.CL · cs.CR
Vision Is Not Overhead: One-Pass Block Drafting for Lossless Speculative Decoding in Vision-Language Modelscs.AI · cs.CL
The Safeguard Worked. Is the LLM System Safer?cs.CR · cs.AI
Distributed Implicit Harm: A Compositional Safety Blind Spot in MLLM-Based Video Moderationcs.CV · cs.AI
Federated Trust for Embodied Robot Capability Marketplacescs.CR · cs.RO
Forbid Your Attention: Fooling Multimodal Large Language Models by Selectively Removing Intrinsic Focus in Spectral Domaincs.CV
LatentPress: Context Compression Beyond Text and Visioncs.LG · cs.AI
From Detection to Refusal: Safer LLMs via Circuit-Guided Weight Scalingcs.CL · cs.AI
Effective Interventions Against AI-Enhanced Scamscs.CR · cs.CY
Causal Evidentiary Governance for High-Risk Machine Learning Systemscs.CY · cs.AI
Jailbreaking Text-to-Image Models Through Cracks: Navigating Heterogeneous Safety Filters via Multi-Agent Debatecs.AI · cs.MM
TUE 01 SEP 20262 in view
◆ FRONTIER
Universal Transformers for Circuit Computations: Perfect Length Generalization in Tiny Transformers
Learning generalizable algorithmic computations remains a challenge for neural networks, as reflected in persistent failures on compositional and length generalization benchmarks. We present a provably correct, transformer parameterization (with only 280 learnable parameters for Boolean algebra tasks) capable of learning and evaluating problems of any depth or length. We assume inputs are fully parenthesized, well-fo…
algorithmic-generalizationtransformerscircuit-computationlength-generalization
IBMTakuya Ito, Ruchir Puri, Murray Campbell, et al. (4)cs.LGarXiv ↗PDF ↗
THE INDEX — 1 MORE CLEARED
Authority-Inference Separation in Agentic Finance: First-Line Control, Blockchain Enforcement, and Replayable Assuranceq-fin.GN · cs.CR
loading earlier editions…
THE FIELD — 13,077 · 30D
cs.AI3,980
cs.LG2,526
cs.CL2,075
cs.CV1,786
cs.RO856
cs.CR514
every indexed paper — the whole wire, not just what cleared
IN THE LITERATURE
models referenced in the loaded window · click to pivot
FROM INSIDE THE LABS
papers with a frontier-lab author · click to pivot
Research — every claim one click from the paperarXiv continuous index · FRONTIER ◆ leads each edition