Structured-output generation on 360 fixture
5Inv.Raw + LCV
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Raw + LCVBackend=MiniMax-M2.7, Token Budget=22002026.05 | 5 | 5 | 275 | 3.61 | 0 | 0 | 0 | 0 | |
| Long-ctx + LCVBackend=MiniMax-M2.7, Token Budget=22002026.05 | 1 | 1 | 287 | 5 | 0 | 0 | 0 | 0 | |
| CBEA + LCVBackend=MiniMax-M2.7, Token Budget=22002026.05 | 1 | 1 | 211 | 49.44 | 0 | 0 | 0 | 0 | |
| CBEA + LCVBackend=GPT-OSS-120B, Token Budget=22002026.05 | 1 | 0 | 15 | 50.56 | 0 | 0 | 0 | 0 | |
| Raw + LCVBackend=DeepSeek-V4-Flash, Token Budget=22002026.05 | 0 | 0 | 0 | 9.17 | 0 | 0 | 0 | 0 | |
| Long-ctx + LCVBackend=DeepSeek-V4-Flash, Token Budget=22002026.05 | 0 | 0 | 0 | 7.78 | 0 | 0 | 0 | 0 | |
| CBEA + LCVBackend=DeepSeek-V4-Flash, Token Budget=22002026.05 | 0 | 0 | 0 | 60 | 0 | 0 | 0 | 0 | |
| Raw + LCVBackend=GPT-OSS-120B, Token Budget=22002026.05 | 0 | 0 | 7 | 0.28 | 0 | 0 | 0 | 0 | |
| Long-ctx + LCVBackend=GPT-OSS-120B, Token Budget=22002026.05 | 0 | 0 | 9 | 0.56 | 0 | 0 | 0 | 0 |