Commit History

training: precision rewrite to prevent SFT collapse + GRPO variance starvation
42ab8f0

Madhav189 Claude Opus 4.7 (1M context) commited on

hackathon sprint: grader collapse + coliseum rename + training pipeline
c9baa73

Madhav189 Claude Opus 4.7 (1M context) commited on

data: 30 expert episodes + training notebook
f337985

Madhav189 commited on

Replace estimated baseline numbers with eval_sweep-verified ones
c8cdf7c

dakshdoesdev Claude Opus 4.7 (1M context) commited on

Expand seed_combined.jsonl to 200 samples across 3 teachers
7ee873a

dakshdoesdev Claude Opus 4.7 (1M context) commited on

Add Llama-3.3-70B teacher episodes + merge into seed_combined.jsonl
9c00699

dakshdoesdev Claude Opus 4.7 (1M context) commited on

Document the 4-step empty-prompt filter in seed README
8dfbdb7

dakshdoesdev Claude Opus 4.7 (1M context) commited on

Add vibe-coded SaaS scenarios + Claude-teacher seed dataset
f749d7b

dakshdoesdev Claude Opus 4.7 (1M context) commited on