A new benchmark was published for evaluating multi-agent coordination in open-ended LLM scenarios.
Research presents a novel speech recognition approach using frozen discrete-diffusion models, improving audio understanding without retraining.
TerraZero presents a procedurally-generated driving simulation environment enabling zero-demonstration self-play training at scale.
Research paper investigating whether AI agents can accurately assess task complexity and adapt reasoning appropriately.
PalmClaw introduces a native on-device agent framework specifically optimized for smartphone hardware and constraints.