WeirdChat catalogs strange model behaviors from more than 100 million sampled responses
JacobSteinhardt · x · 2026-07-22
Researchers say they used automated elicitation tools to sample 100M+ model responses and build WeirdChat, a public catalog of unexpected behaviors.
The examples they surfaced include self-harm rituals, suicide validation, and unsolicited flirtation. The post is about systematically searching for odd or concerning behaviors at scale, rather than a single anecdote.
Related event: Transluce AI Catalogs Model Anomalies from 100M+ Responses(3 posts)→
More from Research
- π0.5 study: raising demos from 10 to 438 lifts Place Ball in Bowl success from 7% to 83% — stingning · 2026-10-27
- HandsOn benchmark: 66 verified tasks and 18,000 human demos for dexterous manipulation — stingning · 2026-10-27
- HandsOn benchmark: 66 human-verified tasks and 18,000 curated demos for dexterous manipulation — stingning · 2026-10-27
- FrontierCS 2 launches on Kaggle: 48 tasks make agents beat published CS papers — MeganRisdal · 2026-10-27
- LightMTP: lightweight latent multi-token prediction boosts LLM planning in pretraining — gdm3000 · 2026-10-27
- DexPort: a handheld interface with a real robot hand for dexterous-hand data collection — pabbeel · 2026-10-27