# Spreadsheet Manipulation and Reasoning on SpreadsheetBench

Source: [Spreadsheet Manipulation and Reasoning on SpreadsheetBench](<https://www.sota2.com/research/sota/spreadsheet-manipulation-and-reasoning-on-spreadsheetbench>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-11T07:26:49.366Z

Evaluation of model performance on the SpreadsheetBench dataset, focusing on combined spreadsheet manipulation and reasoning capabilities, measured by Soft Accuracy and Hard Accuracy.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Spreadsheet manipulation and reasoning |
| Dataset (including any stated split) | SpreadsheetBench |
| Metric | Soft Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 8 of 8 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Soft Accuracy](<https://www.sota2.com/research/sota/spreadsheet-manipulation-and-reasoning-on-spreadsheetbench.md?metric=Soft%20Accuracy>): Higher is better; unit: PERCENT.

- [Hard Accuracy](<https://www.sota2.com/research/sota/spreadsheet-manipulation-and-reasoning-on-spreadsheetbench.md?metric=Hard%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Soft Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| TabClaw | 28.8 | Not recorded | 6a28f31a90966ade919c15e2 |
| Binder | 24.1 | Not recorded | 6a28f31a90966ade919c15e2 |
| SpreadsheetLLM | 23.3 | Not recorded | 6a28f31a90966ade919c15e2 |
| SheetAgent | 22.3 | Not recorded | 6a28f31a90966ade919c15e2 |
| Vanilla | 18.8 | Not recorded | 6a28f31a90966ade919c15e2 |
| ReAct | 18.8 | Not recorded | 6a28f31a90966ade919c15e2 |
| Chain-of-Table | 14.8 | Not recorded | 6a28f31a90966ade919c15e2 |
| NormTab | 14.8 | Not recorded | 6a28f31a90966ade919c15e2 |

### TabClaw

Soft Accuracy: 28.8. Result ID: 6a2a632efbc373592c271a34.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f31a90966ade919c15e2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a28f31a90966ade919c15e2.md>)

Recorded evaluation configuration: {"Category":"Ours","Backbone":"DeepSeek-v3.2"}. This record does not assert independent reproduction.

### Binder

Soft Accuracy: 24.1. Result ID: 6a2a632efbc373592c271a2e.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f31a90966ade919c15e2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ea71e209131cea411d9.md>)

[Implementation](<https://github.com/hkunlp/binder>)

Recorded evaluation configuration: {"Category":"Context Augmentation","Backbone":"DeepSeek-v3.2"}. This record does not assert independent reproduction.

### SpreadsheetLLM

Soft Accuracy: 23.3. Result ID: 6a2a632efbc373592c271a31.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f31a90966ade919c15e2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c71e209131cea75447.md>)

Recorded evaluation configuration: {"Category":"Context Augmentation","Backbone":"DeepSeek-v3.2"}. This record does not assert independent reproduction.

### SheetAgent

Soft Accuracy: 22.3. Result ID: 6a2a632efbc373592c271a33.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f31a90966ade919c15e2.md>)

Recorded evaluation configuration: {"Category":"Autonomous Agent","Backbone":"DeepSeek-v3.2"}. This record does not assert independent reproduction.

### Vanilla

Soft Accuracy: 18.8. Result ID: 6a2a632efbc373592c271a2d.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f31a90966ade919c15e2.md>)

Recorded evaluation configuration: {"Category":"Direct Inference","Backbone":"DeepSeek-v3.2"}. This record does not assert independent reproduction.

### ReAct

Soft Accuracy: 18.8. Result ID: 6a2a632efbc373592c271a32.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f31a90966ade919c15e2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6eb41e209131cea41992.md>)

[Implementation](<https://react-lm.github.io/>)

Recorded evaluation configuration: {"Category":"Autonomous Agent","Backbone":"DeepSeek-v3.2"}. This record does not assert independent reproduction.

### Chain-of-Table

Soft Accuracy: 14.8. Result ID: 6a2a632efbc373592c271a2f.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f31a90966ade919c15e2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71d71e209131cea534db.md>)

Recorded evaluation configuration: {"Category":"Context Augmentation","Backbone":"DeepSeek-v3.2"}. This record does not assert independent reproduction.

### NormTab

Soft Accuracy: 14.8. Result ID: 6a2a632efbc373592c271a30.

[Reporting paper](<https://www.sota2.com/research/paper/6a28f31a90966ade919c15e2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77c81e209131cea754f7.md>)

[Implementation](<https://github.com/mahadi-nahid/NormTab>)

Recorded evaluation configuration: {"Category":"Context Augmentation","Backbone":"DeepSeek-v3.2"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.