Lesser Wrong Feed

 Exploring Unsupervised Methods for Recovering the Refusal Direction in LLMs
 Thu, 22 Oct 2026 02:05:20 GMT
 What's in CODI's latent recurrent thoughts?
 Wed, 21 Oct 2026 22:43:29 GMT
 Misaligned!
 Wed, 21 Oct 2026 22:30:55 GMT
 Model whisperers
 Wed, 21 Oct 2026 22:11:35 GMT
 Even GPT-2 passes most functional global workspace tests
 Wed, 21 Oct 2026 21:48:04 GMT
 D̶o̶n̶’̶t̵⠀Can’t Look Up
 Wed, 21 Oct 2026 22:11:29 GMT
 Olmo's Kitchen: Probing a Model's Behaviour Across Various Checkpoints.
 Wed, 21 Oct 2026 23:26:28 GMT
 AM-Bench: A theoretically-grounded benchmark for agentic misalignment
 Wed, 21 Oct 2026 19:54:02 GMT
 Cross-Model In-Context Neurofeedback Reveals (Limited) Activation Control in LLMs
 Wed, 21 Oct 2026 19:53:58 GMT
 AI Needs A Culture
 Wed, 21 Oct 2026 20:14:56 GMT