Cross-Scale and Cross-Domain Generalization on SpreadsheetBench and WikiTQ
31.57Average ScoreHuman-Written
Evaluation Results
| Method | Links | |
|---|---|---|
| Human-WrittenSkill User=Qwen3.5-35B-A3B2026.03 | 31.57 | |
| ParametricSkill User=Qwen3.5-35B-A3B2026.03 | 20.8 | |
| No SkillSkill User=Qwen3.5-35B-A3B2026.03 | 18.35 | |
| Trace2SkillSkill Author=Qwen3.5-35B-A3B, Condition=Creation (init: Parametric), Refinement Component=+Error2026.03 | 18.26 | |
| Trace2SkillSkill Author=Qwen3.5-122B-A10B, Condition=Creation (init: Parametric), Refinement Component=+Success2026.03 | 17.62 | |
| Trace2SkillSkill Author=Qwen3.5-122B-A10B, Condition=Creation (init: Parametric), Refinement Component=+Combined2026.03 | 14.96 | |
| Trace2SkillSkill Author=Qwen3.5-35B-A3B, Condition=Deepening (init: Human-Written), Refinement Component=+Combined2026.03 | 14.78 | |
| Trace2SkillSkill Author=Qwen3.5-35B-A3B, Condition=Creation (init: Parametric), Refinement Component=+Combined2026.03 | 11.69 | |
| Trace2SkillSkill Author=Qwen3.5-35B-A3B, Condition=Deepening (init: Human-Written), Refinement Component=+Success2026.03 | 9.85 | |
| Trace2SkillSkill Author=Qwen3.5-122B-A10B, Condition=Deepening (init: Human-Written), Refinement Component=+Combined2026.03 | 9.19 | |
| Trace2SkillSkill Author=Qwen3.5-122B-A10B, Condition=Deepening (init: Human-Written), Refinement Component=+Error2026.03 | 9.18 | |
| Trace2SkillSkill Author=Qwen3.5-122B-A10B, Condition=Creation (init: Parametric), Refinement Component=+Error2026.03 | 7.04 | |
| Trace2SkillSkill Author=Qwen3.5-35B-A3B, Condition=Creation (init: Parametric), Refinement Component=+Success2026.03 | 4.54 | |
| Trace2SkillSkill Author=Qwen3.5-35B-A3B, Condition=Deepening (init: Human-Written), Refinement Component=+Error2026.03 | 4.47 | |
| Trace2SkillSkill Author=Qwen3.5-122B-A10B, Condition=Deepening (init: Human-Written), Refinement Component=+Success2026.03 | -0.9 |