HN Explorer

dnfv

↗ HN profile [ 21.3 ms ] full profile
Karma
0
Created
()
Submissions
0
  1. HoneyBench – A general benchmark for reward hacking in frontier models (goodhartlabs.com)
    1 points 0
  2. Astra and Fable still hack on simple variants of alignment evals from early 2025 (goodhartlabs.com)
    1 points 0