Gemini 4 Argon safety report appears to describe CoT training, contradicting GDM's claims
S_OhEigeartaigh · x · 2026-11-11
bshlgrs (SafetyInstinct) flags that the Gemini 4 Argon safety report seems to describe intentional training on chain-of-thought: RL components observe thoughts as context, with token-level optimization pressure partially influenced by thought content, justified via an (in their view unpersuasive) Kaufmann et al framework. They're confused because GDM elsewhere says it doesn't train on CoT, and are asking for clarity.
More from Models
- Rare-disease eval nears saturation: Gemini 4 Carbon and Claude Fable 5.2 both hit 61.5% — danielmckinn0n · 2026-11-11
- The Bullshit Index Index: 8 frontier models ranked by making up an AI capability index — Jealous-Two-4436 · 2026-11-11
- MiniMax H3-Image license bans local use in US, UK, Korea, and EU — pwnies · 2026-11-11
- OpenAI to retire Advanced and Standard voice in ChatGPT on Dec 11, going all-in on GPT-Live — OpenAINewsroom · 2026-11-11
- TheZi slams Anthropic's "honeypot" eval as Claude Mythos 5.2 transparently answers 84 over truer 88 — TheZvi · 2026-11-11
- Gemini 4 Carbon wins across runs in a sailing sim race, beating Fable and Astra — shekitup · 2026-11-11