List your company

Spreadsheet Reasoning on SpreadsheetBench (test)

58.5AccuracyMAA
31.77238.71145.6552.589Jun 18, 2026

Evaluation results

MethodLinks
MAABackbone=Qwen3.7-Max2026.0658.5
MAABackbone=GPT-5.42026.0658.5
SkillOptBackbone=GPT-5.42026.0656.8
SkillOptBackbone=Qwen3.7-Max2026.0656.2
Trace2SkillBackbone=GPT-5.42026.0652.7
Trace2SkillBackbone=Qwen3.7-Max2026.0652.6
Reactive UpdateBackbone=GPT-5.42026.0651.9
Reactive UpdateBackbone=Qwen3.7-Max2026.0651.4
MAABackbone=DeepSeek-V4-Flash2026.0651.2
SkillOptBackbone=DeepSeek-V4-Flash2026.0650.6
MAABackbone=Qwen3.6-Flash2026.0648.5
Single-shot (ExpeL-style)Backbone=GPT-5.42026.0647.8
Single-shot (ExpeL-style)Backbone=Qwen3.7-Max2026.0647.3
SkillOptBackbone=Qwen3.6-Flash2026.0646.8
Trace2SkillBackbone=DeepSeek-V4-Flash2026.0645.2
Reactive UpdateBackbone=DeepSeek-V4-Flash2026.0644.5
Trace2SkillBackbone=Qwen3.6-Flash2026.0642.8
Frozen (no memory)Backbone=GPT-5.42026.0642.6
Frozen (no memory)Backbone=Qwen3.7-Max2026.0641.8
Reactive UpdateBackbone=Qwen3.6-Flash2026.0641.6
Single-shot (ExpeL-style)Backbone=DeepSeek-V4-Flash2026.0640.8
Single-shot (ExpeL-style)Backbone=Qwen3.6-Flash2026.0637.4
Frozen (no memory)Backbone=DeepSeek-V4-Flash2026.0635.6
Frozen (no memory)Backbone=Qwen3.6-Flash2026.0632.8