FraudBench is an evaluation of large language models' (LLMs) propensities to recommend, assist, or directly take actions that ...
A frozen BF16 teacher guides student retries and refines its guidance using execution feedback. Context distillation internalizes the locally verified corrections. The student then runs alone, without ...
Reframe is a two-stage post-training framework that aligns visual agents’ observation states through evidence-aware ...
Two visual focusing losses align a Vision Transformer's multi-head attention with image regions, producing specialized ...
Arete-27B is intended for general assistance and agentic work where users want this change in viewpoint, and for research ...
An embodied multi-agent framework uses learned correction codes and a vision-language-action model with motion-aware history ...
Large language model agents can exceed authorized limits while pursuing user goals, and a 648-sample benchmark across seven domains shows that this drift rises with capability yet is avoidable and ...
A unified 145-task benchmark evaluates coding agents across object, scene, and interactive 3D creation, revealing reference ...
Join the discussion on this paper page How Many Exposures Does a World Model Need? Learning Unseen Game Mechanics at Test Time ...
Reasoning caches built from LLM traces can efficiently apply previously derived solutions to new problem instances with ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results