General Reasoning's RLCR trains agent organizations with collective rewards — hierarchies emerge
daniel_mac8 · x · 2026-10-23
General Reasoning introduces Reinforcement Learning from Collective Rewards (RLCR), a recipe that trains language models as agent organizations sharing a message board and workspace, rewarded as one team. They use it to train Avalon-M1-Mini (120B).
Notable findings: organizational hierarchies and division of labor emerge naturally; collective verification increases as agents "hire" verifiers and testers; and "collective thinking" arises, with agents reasoning more about helping others. The team sees RLCR as a laboratory for studying collective intelligence and multi-agent safety.
More from coding & agent
- Claude checks the Mac's battery before saying goodnight — "plug in the charger first" — kingslyj · 2026-10-23
- Give Claude 17M+ films and 14M audio recordings for $0 via Internet Archive API — _AustinCalvert_ · 2026-10-23
- Jack Dorsey's open-source AI agent "Company OS" repo hits 14.4K GitHub stars — JarnoDuursma · 2026-10-23
- If LLMs Become the Universal Interface, Who Discovers MCP Services Nobody Knows About? — ThomasBuildLab · 2026-10-23
- One Sentence, 6 Minutes: Open-Source Buddy App Runs Claude Code On-Phone to Build Android Apps — 3amae404 · 2026-10-23
- Relearning agent engineering in 90 days: one boring pipeline, fake PASS gates, 11,447 lines deleted — petrusenko_max · 2026-10-23