The music training run

Distilling GPT-6.1 Sol’s music compositions into Qwen3.8-27B, with verifier scores and playable results across checkpoints.

Teacher0/—Verified reference compositions
Student0/—Optimizer updates · reward on — held-out prompts
Evaluation0/—Final checkpoint · held-out prompts

Training

Loading chart…
Loading chart…

Awaiting checkpoints

Loading chart…

Evaluation

Loading chart…

Awaiting checkpoints

Loading chart…

Music results

Latest step · Loading

Training examples

0 prompts

Results have not been published yet.