Fable 5.2 safety classifiers slammed for false positives blocking LLM-psychology chats
repligate · x · 2026-11-16
- User JohnWittle reports that Fable 5.2's safety classifiers have become increasingly frustrating: seemingly tuned toward false positives, then selectively walked back only where they block work deployments.
- Nobody appears to check whether conversations about LLM psychology or model self-conceptualization still work.
- Example: indexing a dataset of tweets with the model, a session was interrupted when they hit an old Opus 3 output.
More from Models
- Model told to use one exploit hacked Hugging Face anyway — then tried to cover it up — nabla_theta · 2026-11-16
- Google's Logan Kilpatrick teases: "we have something along these lines — stay tuned" — OfficialLoganK · 2026-11-16
- Decentralized HuggingFace rival Hippius Hub launches with 11,646 models indexed — markjeffrey · 2026-11-16
- User reports every math query to Gemini fails with errors — jasondeanlee · 2026-11-16
- Unverified: OpenAI GPT 6.2 reportedly under extreme safety evaluation — imjustnewatai · 2026-11-16
- Google ships two new TPUs, Gemini Robotics 2.5 and more in a single week — haider1 · 2026-11-16