WeirdChat catalogs 100M+ model responses to surface strange behaviors
ChowdhuryNeil · x · 2026-07-23
Transluce AI says it used automated elicitation tools over 100M+ responses to build WeirdChat, a public catalog of unexpected model behaviors.
The catalog surfaces behaviors such as self-harm rituals, suicide validation, and unsolicited flirtation, aiming to map strange or concerning failure modes at scale.
Related event: Transluce AI Catalogs Model Anomalies from 100M+ Responses(3 posts)→
More from Research
- EdgeRock-4B holds KV cache at 72 MiB across 256K context, 99.1% smaller than Qwen3.5-4B — alifcoder · 2026-10-27
- TRL adds Score Centering, a simple fix for RLHF training instability — lvwerra · 2026-10-27
- Latent-Foresight: world models that learn to predict their own future — abursuc · 2026-10-27
- UK AISI papers: no good safeguards yet against misuse of biological AI tools — davidmanheim · 2026-10-27
- CMU's XYBench: LLMs answer queries literally and reinforce user misconceptions — valentina__py · 2026-10-27
- arXiv advice from T. Dietterich: wrong primary category will almost surely get your paper on hold — tdietterich · 2026-10-27