Commit History

  1. fix: GRPO zero-gradient trap, DSL prompt seed, best_reward, teacher oracle

    0b77231
  2. training results

    ede0718

  1. discovery plot

    1edadfb
  2. reward curve

    8c686ce
  3. imp DSL, reward structure, paradigm shift

    8270f31
  4. submodule access,timeout, gro checkpoints, mdl penalty

    92e9724
  5. V1

    Ankit
    be9f416