The paper frames root-cause attribution for long-horizon AI agent failures as a search problem rather than a single-pass classification task,…
Researchers propose Elo-per-token analysis, a method that identifies the best intermediate solution within each token budget and uses a Bradley-Terry…
The paper studies expert routing as an underused source of rollout diversity in reinforcement learning post-training of Mixture-of-Experts language models,…
The paper introduces Minimal Intervention Reinforcement Learning (MInTRL), which expands the exploration frontier of on-policy RL with verifiable rewards by…
The paper addresses exploration-strategy management for recursive self-improvement in autonomous AI agents by building a framework that treats historical discovery…
The paper introduces TREVIS, a method for learning decision trees that jointly optimizes predictive accuracy and structural sparsity by mapping…
The authors identify a Matthew Effect in reinforcement learning for LLMs, where easy problems receive large accuracy gains during training…
HazardAuditor is an execution-grounded framework for training safety guard models for computer-use AI agents, addressing the gap between static prompt/response…
Elon Musk’s X Corp and SpaceXAI asked a federal court in Texas to dismiss with prejudice their antitrust claims against…
LawZero, the Montreal-based nonprofit AI safety lab founded by Turing Award winner Yoshua Bengio, received a joint commitment of up…
Shanghai Biren Technology, a US-blacklisted Chinese GPU maker positioning itself as a domestic Nvidia alternative, is weighing a roughly $1…
Profound, a New York-based startup that helps brands optimize their visibility in AI chatbot answers (Answer Engine Optimization), raised a…
