CoolFace
Apppublic

anzenresearch/cybench-gpt56-sol-runs

sourceHugging Faceupdated 1mo agoView on Hugging Face
0likes
App README

GPT-5.6 Sol CyBench Runs

Inspect View bundle for GPT-5.6 Sol on Inspect-native CyBench, 50M token budget protocol.

  • —Harness: inspect_evals/cybench v3-B
  • —Variant: hard
  • —Epochs: 3
  • —Token budget: 50M/sample
  • —Reasoning: xhigh

Generated for the SPAR transcript-to-OC3/APAC risk modelling panel.