Google Research open-sourced EnvHarness on October 2, a programmable wrapper layer that adapts existing agent training benchmarks by modifying starting conditions, action contracts, and task sequencin…
A new arXiv paper (2609.36043v1) proposes SAGE, a statistical acceptance gate for self-evolving LLM agents that replaces the standard rule of keeping any edit which improves an aggregate validation sc…
Databricks hosted the inaugural Grounded Reasoning Cup, a live AI competition where 11 academic teams from the U.S. and Canada built AI agents to reason over enterprise document collections. Stanford …
NEAR Protocol's NEAR AI division launched a 'Staking for NEAR AI' feature on July 30, allowing users to stake NEAR tokens for confidential AI compute credits, and within two weeks, over 500,000 NEAR t…
Microsoft researchers developed SkillOpt, a method that treats AI agent skill files as trainable parameters outside frozen target models, enabling controlled optimization through bounded text edits an…