$T^5$: Twin-Critic Training for Token-Level Thoughts in Reinforcement Mid-Training
Reinforcement mid-training lets language models learn internal thoughts from unlabeled text, but efficient token-level credit assignment remains challenging. Existing group-relative methods require costly repeated generation. Learned critics offer single-rollout feedback, but accurate return prediction alone does not ensure reliable policy updates. Our analysis shows how training--inference mismatch and PPO clipping prevent a common offset in advantage estimates from cancelling out, introducing additional update drift. We propose \tfour{}, a twin-critic method that calibrates token-level advantages from a single generated trajectory. After warmup and held-out qualification, the critics pr...
| Authors | Nan Qiao, Yebin Yang, Weinong Wang et al. |
|---|---|
| Primary category | cs.AI |
| Published | 26 September 2026 |
| Last updated at source | 2 October 2026 |
Silytics shows public metadata from arXiv and links to the original. Information may be incomplete or out of date. Last refreshed 17 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.