Kenji Kawaguchi

State-of-the-art coding agents exhibit critical boundary violations, with nearly all misinterpreting dependencies and over half compromising environment integrity or user inputs.

Weak interactive feedback can dynamically steer LLMs to the correct similarity criterion for test-time decisions, unlocking significant performance gains beyond static mixtures.

HTFM transforms the handling of heavy-tailed data, achieving superior sample quality and tail recovery while maintaining efficient sampling speeds.

Standard Gaussian initialization can lead to mode collapse, but a new approach using guidance potential posteriors dramatically enhances image diversity.

VLMs can regain lost temporal reasoning abilities without retraining, simply by strategically merging the right layers from their text-only LLM backbone.