Tensorwire
Products & tools · first seen 3 Sep, updated 3 Sep

Improving Evaluation Realism with Inference-Time Compute and Deployment Scaffolds

2 outlets

arXiv:2609.02302v1 Announce Type: new Abstract: A core obstacle to alignment evaluation is evaluation awareness: capable models can tell when they are being tested rather than deployed, weakening the conclusions a safety evaluation can supp…

Summary from arXiv cs.AI.

Coverage 2 articles · 2 outlets

  1. arXiv cs.AI
    Improving Evaluation Realism with Inference-Time Compute and Deployment Scaffolds
  2. LessWrong
    Improving Audit Realism With Inference Time Compute and Deployment Scaffolds