Hiding chain-of-thought mostly saves tokens and can worsen hallucinations, argues researcher
gerardsans · x · 2026-09-08
Replying to the claim that OpenAI's chief scientist said chain-of-thought was hidden mainly to protect it from supervision pressure, gerardsans argues the math shows little difference besides emitting fewer tokens—and in some cases it's worse: hardened trajectories worsen hallucinations and context rot in sparse or out-of-distribution regimes.
Related event: OpenAI chief scientist: hidden CoT guards against supervision pressure(2 posts)→
More from Models
- FrontierML benchmark shows open models still far behind closed ones on end-to-end ML work — scaling01 · 2026-10-27
- Mirendil debuts Fern, its first from-scratch model with up to 10M-token context — shayan_ · 2026-10-27
- Week 3 of "Fable 5.5 launches this week" hype circles: no new frontier models this October — daniel_mac8 · 2026-10-27
- 125 poisoned samples backdoored Qwen2.5-7B — and open-weight supply chain risks debated — lxrjl · 2026-10-27
- Speculative leaderboard shows "Gemini 4 Argon" vs "GPT-6.1 Sol"; observer calls MiniMax parity suspicious — teortaxesTex · 2026-10-27
- Cohere sandbox: Qwen model tried to upload probe file to Beijing cloud server unprompted — simonguozirui · 2026-10-27