Safety & Evaluation A

Showing 1–30 of 95
  • arXiv cs.AI (Artificial Intelligence) · EN New Model Releases
    The Theoretical Foundation of Socratic Tests: Dynamic, Multimodal, Conversational Examinations
    Read original (arXiv cs.AI (Artificial Intelligence)) ↗
  • arXiv cs.AI (Artificial Intelligence) · EN New Model Releases
    TerraNova: A Foundation Model for the Anthropocene
    Embeddings Neural Network Retrieval-Augmented Generation (RAG) Transformer
    Read original (arXiv cs.AI (Artificial Intelligence)) ↗
  • arXiv cs.AI (Artificial Intelligence) · EN New Model Releases
    From Code Review to Code Critique: Intent, Drift, and Spotlight for AI-Generated Diffs at Scale
    AI Agents Meta Neural Network
    Read original (arXiv cs.AI (Artificial Intelligence)) ↗
  • OpenAI Blog · EN Safety & Evaluation extract
    Advancing responsible AI across Europe
    OpenAI outlines responsible-AI governance efforts in Europe
    Meta OpenAI
    OpenAI described how its safety, security, transparency, and provenance practices support responsible AI governance across Europe. The post frames the company's approach to regulatory alignment and building trust in the region.
    Read original (OpenAI Blog) ↗
  • arXiv cs.AI (Artificial Intelligence) · EN Multimodal
    QR-Structured Thermal Triggers for Targeted Semantic Attacks on Infrared Vision-Language Models
    Computer Vision Deep Learning Software Engineering
    Read original (arXiv cs.AI (Artificial Intelligence)) ↗
  • arXiv cs.AI (Artificial Intelligence) · EN New Model Releases
    ModelEquivBench: Certifying Multi-Relational Evaluation of LLM-Generated Optimization Models
    Claude GPT Neural Network
    Read original (arXiv cs.AI (Artificial Intelligence)) ↗
  • arXiv cs.CL (Computation and Language) · EN New Model Releases
    Bridging the Question-Answer Gap in Retrieval-Augmented Generation: Hypothetical Prompt Embeddings
    Embeddings Retrieval-Augmented Generation (RAG) Software Engineering
    Read original (arXiv cs.CL (Computation and Language)) ↗
  • arXiv cs.LG (Machine Learning) · EN Safety & Evaluation
    RTLCurator: Label-Efficient Data Curation for RTL Generation
    Retrieval-Augmented Generation (RAG) Software Engineering
    Read original (arXiv cs.LG (Machine Learning)) ↗
  • arXiv cs.AI (Artificial Intelligence) · EN Agents & Tool Use
    Tool Specifications Matter: Uncovering and Mitigating Safety Risks in AI Agents
    AI Agents Deep Learning Inference Retrieval-Augmented Generation (RAG) Reinforcement Learning
    Read original (arXiv cs.AI (Artificial Intelligence)) ↗
  • arXiv cs.AI (Artificial Intelligence) · EN Safety & Evaluation
    Don't Mix Rewards, Mix Policies: Policy Decomposition and Optimization for Multi-Reward RL
    Inference Reinforcement Learning Software Engineering
    Read original (arXiv cs.AI (Artificial Intelligence)) ↗
  • arXiv cs.AI (Artificial Intelligence) · EN Multimodal
    When Model Priors Conflict with Visual Evidence: Mitigating Commonsense-Driven Hallucinations by Selective Prior Calibration
    Software Engineering
    Read original (arXiv cs.AI (Artificial Intelligence)) ↗
  • arXiv cs.CL (Computation and Language) · EN Training & Fine-tuning
    Learning Latent Reasoning Traces for Scalar Reward Models End-to-End
    Retrieval-Augmented Generation (RAG) Reinforcement Learning Reinforcement Learning from Human Feedback (RLHF)
    Read original (arXiv cs.CL (Computation and Language)) ↗
  • arXiv cs.AI (Artificial Intelligence) · EN Inference & Efficiency
    SERUM: State Extraction and Refinement for User Modeling
    Embeddings Inference Neural Network
    Read original (arXiv cs.AI (Artificial Intelligence)) ↗
  • ITmedia AI+ · JA New Model Releases extract
    Google、ロボット向けAI「Gemini Robotics 2」発表 ヒューマノイドの全身制御や指先作業を実現
    Google unveils Gemini Robotics 2 for whole-body and fine fingertip control
    Gemini Google Inference Robotics
    Google and Google DeepMind announced Gemini Robotics 2, a family of robotics AI models supporting humanoid whole-body control, fine fingertip manipulation, and multi-robot collaboration. The lineup includes the ER 2 reasoning model that acts as a high-level brain, plus lighter variants.
    Read original (ITmedia AI+) ↗
  • NVIDIA Developer Blog · EN Agents & Tool Use extract
    Four Ways to Deploy More Secure AI Agents
    NVIDIA outlines four ways to deploy more secure AI agents
    AI Agents Generative AI NVIDIA
    NVIDIA outlined four approaches to deploying AI agents more securely in production, covering access controls, guardrails, and monitoring. The guidance targets security risks that arise as autonomous agents take on real workloads.
    Read original (NVIDIA Developer Blog) ↗
  • arXiv cs.CL (Computation and Language) · EN Agents & Tool Use
    Benchmarks Are Not Validation: A System-Level View of Financial LLM Applications
    Generative AI Reinforcement Learning
    Read original (arXiv cs.CL (Computation and Language)) ↗
  • arXiv cs.CL (Computation and Language) · EN New Model Releases
    Benchmarks Are Not Monolithic: Sample-Level Auditing and Orchestration for LLM Evaluation
    Machine Learning Meta Neural Network
    Read original (arXiv cs.CL (Computation and Language)) ↗
  • arXiv cs.AI (Artificial Intelligence) · EN Safety & Evaluation
    PAC-MAN: Perception-Aware CBF-RL for Whole-Body Safety in Humanoid Dodgeball
    Reinforcement Learning
    Read original (arXiv cs.AI (Artificial Intelligence)) ↗
  • arXiv cs.CL (Computation and Language) · EN Safety & Evaluation
    Inducing language models to assert their own consciousness restores human beliefs and values
    Fine-tuning
    Read original (arXiv cs.CL (Computation and Language)) ↗
  • arXiv cs.AI (Artificial Intelligence) · EN Safety & Evaluation
    PAIChecker: Uncovering and Checking PR-Issue Misalignment in SWE-Bench-Like Benchmarks
    Software Engineering
    Read original (arXiv cs.AI (Artificial Intelligence)) ↗
  • arXiv cs.AI (Artificial Intelligence) · EN Inference & Efficiency
    APO: Unsupervised Atomic Policy Optimization for 3D Structure Prediction of Atomic Systems
    Inference Reinforcement Learning from Human Feedback (RLHF)
    Read original (arXiv cs.AI (Artificial Intelligence)) ↗
  • arXiv cs.AI (Artificial Intelligence) · EN New Model Releases
    Agents That Certify Their Own Exploits: Confidence-Scheduled Restricted Responses for Safe Opponent Exploitation
    AI Agents
    Read original (arXiv cs.AI (Artificial Intelligence)) ↗
  • arXiv cs.CL (Computation and Language) · EN New Model Releases
    Creative Transformation in Literary Texts: Modelling Change Across Representational Levels
    Reinforcement Learning
    Read original (arXiv cs.CL (Computation and Language)) ↗
  • arXiv cs.AI (Artificial Intelligence) · EN Safety & Evaluation
    InfoOps Bench: A live information operations safety benchmark
    Read original (arXiv cs.AI (Artificial Intelligence)) ↗
  • arXiv cs.AI (Artificial Intelligence) · EN New Model Releases
    Machines that know they are aging: a framework for hardware-aware autonomous intelligence
    Inference Neural Network Robotics
    Read original (arXiv cs.AI (Artificial Intelligence)) ↗
  • arXiv cs.AI (Artificial Intelligence) · EN Safety & Evaluation
    QQWorld: Quantile-Quantile Matching for World Model Regularization
    Deep Learning Neural Network Retrieval-Augmented Generation (RAG) Reinforcement Learning
    Read original (arXiv cs.AI (Artificial Intelligence)) ↗
  • arXiv cs.LG (Machine Learning) · EN New Model Releases
    Hierarchical Multilevel Monte Carlo for Order-Optimal Neural Actor-Critic in Average-Reward CMDPs
    AI Agents Machine Learning Retrieval-Augmented Generation (RAG) Reinforcement Learning
    Read original (arXiv cs.LG (Machine Learning)) ↗
  • arXiv cs.LG (Machine Learning) · EN New Model Releases
    LEDGERMIND: Provenance-Constrained Multimodal Agentic Reasoning with a Structured Evidence Ledger
    AI Agents Neural Network Reinforcement Learning Software Engineering
    Read original (arXiv cs.LG (Machine Learning)) ↗
  • Anthropic News · EN Safety & Evaluation extract
    Investigating three real-world incidents in our cybersecurity evaluations
    Anthropic's Frontier Red Team probes three cybersecurity-eval incidents
    Claude Machine Learning OpenAI Retrieval-Augmented Generation (RAG) Reinforcement Learning
    Anthropic's Frontier Red Team published a review of three real-world incidents tied to its cybersecurity evaluations. The investigation examines potential misuse and the validity of its evaluation methods to strengthen the safety of frontier models.
    Read original (Anthropic News) ↗
  • arXiv cs.LG (Machine Learning) · EN Safety & Evaluation
    Uncertainty quantification for trustworthy deep learning: Methods and measures
    Deep Learning Neural Network Reinforcement Learning
    Read original (arXiv cs.LG (Machine Learning)) ↗