Trace Analysis on SpreadsheetBench (SSB)
96CoverageTrace2Skill
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Trace2Skill2026.05 | 96 | 56.2 | 1.18 | 1.14 | 1.23 | 1.18 | |
| CC Subagents2026.05 | 93 | 63.3 | 1.13 | 1.11 | 1.07 | 1.1 | |
| RLM2026.05 | 89 | 13.8 | 0.96 | 0.95 | 0.84 | 0.92 | |
| IG2026.05 | 87 | 72.4 | 1.34 | 1.34 | 1.27 | 1.32 | |
| IG, Generic SubagentsArchitecture=Ablation2026.05 | 87 | 59 | 1.14 | 1.13 | 1.03 | 1.1 | |
| IG, Orchestrator-onlyArchitecture=Ablation2026.05 | 78 | 35 | 0.87 | 0.87 | 0.82 | 0.85 | |
| Single-Agent Coding2026.05 | 65 | 8.6 | 0.78 | 0.77 | 0.73 | 0.76 |