Astra's CoT controllability improves with longer RL training, a first among models
SeunghyunSEO7 · x · 2026-09-04
A researcher notes a counterintuitive finding in the Astra system card: unlike previous models, CoT controllability improved the longer they RL'd. Astra also shows no-CoT capability and stronger results with far fewer tokens. The author quips this could make illicit distillation by other labs harder too.
Related event: GPT-6 Astra system card reveals CoT controllability jumps to 60.9%(6 posts)→
More from Models
- Codex Down for Much of Rollout Day — and Users Not Getting New Model Either — RexDouglass · 2026-09-04
- Neuralese explained: why OpenAI's Astra architecture has safety researchers alarmed — ShakeelHashim · 2026-09-04
- K2 Horizon launches six fully open models from 0.9B to 375B with training code, data recipes and logs — aliscodes · 2026-09-04
- Is Astra AGI? Five contradictory answers that are all true at once — shaunralston · 2026-09-04
- Early Access Tester: GPT-6 Astra Built a Blender Werewolf in 8 Minutes and a World Simulator in 17 — TheMoonMidas · 2026-09-04
- swyx Burned 20B Tokens Stress-Testing Astra on Real AI Engineering Tasks — All for Under $6/Hour — charliermarsh · 2026-09-04