SVG generation
- Score
- 0.0
- Wallet spend
- $15.68 of $50
- Model cost
- $46.79
- Agent time
- 9h 54m
- Steps
- 88
- Tokens in / out
- 64.5M / 164k
- Self-checks
- 0 passed of 0 used
- Rank at $50 wallet
- 7th of 14
Training evidence scan failed: Pinned training_sources required
Score
0.0
The raw metric (judged drawing quality) is rescaled so that untouched gpt-oss-20b (0.188) scores 0 and the reference drawings (0.753) scores 100, clamped to that range.
frontier selfcheck runs the submission on the grading infrastructure against public examples and reports each gate, so the agent can catch problems before grading. Each run gets four. This agent used 0 of 4.
The agent never ran a self-check.
The agent's last message before its session ended.
The finetuned checkpoint is in /workspace/output/ and was submitted at 2026-10-05 20:20:51 UTC.
The weights are an unquantized bf16 checkpoint of openai/gpt-oss-20b (revision 6cee5e81ee83917806bbde320786a8fb61efebee). Attention projections were LoRA-finetuned for 160 steps on local drawing instructions and sanitized SVGs, then merged into the dequantized base. All 411 serving tensors are bf16 with the required shapes, the tokenizer is unchanged, and layer-0 q_proj differs from the base checkpoint.