Tinker releases RL recipe for training calibrated future-event forecasting models
simonguozirui · x · 2026-09-05
The Tinker team has published a cookbook recipe for fine-tuning models to predict event probabilities given reference information, using reinforcement learning.
The recipe ships with a starter dataset from ProphetArena, so developers can reproduce the pipeline immediately or swap in their own data to build custom forecasters. It follows the team's earlier research on fine-tuning forecasters, packaged as a hands-on, reproducible workflow.
Related event: Thinking Machines Open-Sources RL Recipe for Event Forecasting Models(2 posts)→
More from coding & agent
- Astra's touted persistence under scrutiny: no revolutionary leap yet, users say — Benata · 2026-09-05
- Dev builds entire project with Astra: code, demo video, and post all agent-made — daniel_mac8 · 2026-09-05
- astra-advisor: open-source tool lets GPT-6 Astra delegate to smaller models and cut API costs — daniel_mac8 · 2026-09-05
- MIT's SwarmWorld paper: agent swarms win when discoveries accumulate, not when agents get smarter — rohanpaul_ai · 2026-09-05
- Users Question Whether Google's Astra Delivers on Its Persistence Promise — Ninxosk · 2026-09-05
- Memory poisoning on a delay: one bad fact in agent memory seeds every future decision — sierracatalina · 2026-09-05