Spreadsheet Reasoning on SpreadsheetBench (test)
58.5AccuracyMAA
Evaluation Results
| Method | Links | |
|---|---|---|
| MAABackbone=Qwen3.7-Max2026.06 | 58.5 | |
| MAABackbone=GPT-5.42026.06 | 58.5 | |
| SkillOptBackbone=GPT-5.42026.06 | 56.8 | |
| SkillOptBackbone=Qwen3.7-Max2026.06 | 56.2 | |
| Trace2SkillBackbone=GPT-5.42026.06 | 52.7 | |
| Trace2SkillBackbone=Qwen3.7-Max2026.06 | 52.6 | |
| Reactive UpdateBackbone=GPT-5.42026.06 | 51.9 | |
| Reactive UpdateBackbone=Qwen3.7-Max2026.06 | 51.4 | |
| MAABackbone=DeepSeek-V4-Flash2026.06 | 51.2 | |
| SkillOptBackbone=DeepSeek-V4-Flash2026.06 | 50.6 | |
| MAABackbone=Qwen3.6-Flash2026.06 | 48.5 | |
| Single-shot (ExpeL-style)Backbone=GPT-5.42026.06 | 47.8 | |
| Single-shot (ExpeL-style)Backbone=Qwen3.7-Max2026.06 | 47.3 | |
| SkillOptBackbone=Qwen3.6-Flash2026.06 | 46.8 | |
| Trace2SkillBackbone=DeepSeek-V4-Flash2026.06 | 45.2 | |
| Reactive UpdateBackbone=DeepSeek-V4-Flash2026.06 | 44.5 | |
| Trace2SkillBackbone=Qwen3.6-Flash2026.06 | 42.8 | |
| Frozen (no memory)Backbone=GPT-5.42026.06 | 42.6 | |
| Frozen (no memory)Backbone=Qwen3.7-Max2026.06 | 41.8 | |
| Reactive UpdateBackbone=Qwen3.6-Flash2026.06 | 41.6 | |
| Single-shot (ExpeL-style)Backbone=DeepSeek-V4-Flash2026.06 | 40.8 | |
| Single-shot (ExpeL-style)Backbone=Qwen3.6-Flash2026.06 | 37.4 | |
| Frozen (no memory)Backbone=DeepSeek-V4-Flash2026.06 | 35.6 | |
| Frozen (no memory)Backbone=Qwen3.6-Flash2026.06 | 32.8 |