inference — Running a trained model to get answers — the phase where AI is actually used, as opposed to trained.
tool use — A model's ability to call external functions — run code, search the web, edit files — instead of only generating text.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
Why it matters
Reward design over tool choice, argument quality, and JSON validity moved a compact open model from 70.8% to 81.2% on held-out tool useA model's ability to call external functions — run code, search the web, edit files — instead of only generating text.Full definition →. This provides a worked template for post-training an AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → on evidence-seeking rather than plausible-looking answers.
articleSERPO: Self-Evolving Rubric Policy Optimization for Open-Ended Test-Time Reinforcement LearningJianze Wang, Kunwang Zheng, Ying Liu, Yu Cao, Qilong Zhang, Jinlong Chen, Hua Yang, Qianglong Chen
videoRL Without Verifiable Rewards (Will Brown, Prime Intellect)AI Engineer
articlePutting Task Expertise Into RlThinking Machines editorial sitemap