Lesser Wrong Feed

 Models are grader modelling not score seeking; grader-modelling is much scarier
 Mon, 09 Nov 2026 15:41:55 GMT
 Knowledge of graders can mediate selection for grader-targeting reasoning through RL
 Mon, 09 Nov 2026 14:57:13 GMT
 AI R&D Is Accelerating Towards Potential RSI in 2027
 Mon, 09 Nov 2026 14:50:56 GMT
 Advocating for a Shared AI Safety Compute Cluster
 Mon, 09 Nov 2026 13:38:55 GMT
 Does expecting automated grading degrade models’ research taste?
 Mon, 09 Nov 2026 12:28:05 GMT
 Subpar Results
 Mon, 09 Nov 2026 08:03:20 GMT
 Be less scared of overconfidence
 Mon, 09 Nov 2026 05:51:10 GMT
 Private LLM copies can agree on a secret key over a public channel
 Mon, 09 Nov 2026 07:47:04 GMT
 All Interesting Behavior Is Overdetermined
 Mon, 09 Nov 2026 03:45:06 GMT
 The J-lens intermediate swap acts partly through the answer directions
 Mon, 09 Nov 2026 01:49:10 GMT