About 2 results
Open links in new tab
  1. benchdif/README.md at main · abhaymettu/benchdif · GitHub

    Real-data finding. Run on 395 Open LLM Leaderboard models, item-level DIF between Mistral- and Llama-2-derived models is …

  2. benchdif/DECISIONS.md at main · abhaymettu/benchdif · GitHub

    Aug 27, 2026 · Real-data blocked on gating, adapter shipped anyway. Open LLM Leaderboard details repos (996) are gated (HTTP …