Commit History

  1. add the crossed Qwen3.8, Nemotron-3.5-Lightning and Inkling-Small runs (training_runs, splits)

    Verified signature
    bbcca45
  2. add the Inkling-Small runs (training_runs, splits)

    Verified signature
    dbc195a
  3. add the Qwen3.8 and Nemotron-3.5-Lightning runs (training_runs, splits, weights)

    Verified signature
    cfd2de2

  1. default = the post's models; other DeepSeek/Nemotron runs in default_other_models

    Verified signature
    dac988a
  2. Kimi/Qwen3.8 runs in their own config (viewer: max 30 splits per config); Qwen3.8 high-risk rows

    Verified signature
    7bb2fed
  3. All exp04 temptation evals: configs by prompt set x backend, one split per model; judge_input_truncated

    Verified signature
    75a4744
  4. data + card

    Verified signature
    d08e8c8
  5. data + card

    Verified signature
    c75abbe