# Chart Data Extraction on ExChart-Bench 1.0 (test)

Source: [Chart Data Extraction on ExChart-Bench 1.0 (test)](<https://www.sota2.com/research/sota/chart-data-extraction-on-exchart-bench-1-0-test>)

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

Data updated: 2026-06-30T09:05:49.669Z

Evaluation of Chart Data Extraction performance on the test split of the ExChart-Bench 1.0 dataset, measured using Format Success Rate and Adaptive MAPE across various chart types.

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 | Chart Data Extraction |
| Dataset (including any stated split) | ExChart-Bench 1.0 (test) |
| Metric | Format Success Rate (All) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 14 of 14 |
| 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

- [Format Success Rate (All)](<https://www.sota2.com/research/sota/chart-data-extraction-on-exchart-bench-1-0-test.md?metric=Format%20Success%20Rate%20(All)>): Higher is better; unit: PERCENT.

- [Format Success Rate (Bar)](<https://www.sota2.com/research/sota/chart-data-extraction-on-exchart-bench-1-0-test.md?metric=Format%20Success%20Rate%20(Bar)>): Higher is better; unit: PERCENT.

- [Format Success Rate (Line)](<https://www.sota2.com/research/sota/chart-data-extraction-on-exchart-bench-1-0-test.md?metric=Format%20Success%20Rate%20(Line)>): Higher is better; unit: PERCENT.

- [Format Success Rate (Scatter)](<https://www.sota2.com/research/sota/chart-data-extraction-on-exchart-bench-1-0-test.md?metric=Format%20Success%20Rate%20(Scatter)>): Higher is better; unit: PERCENT.

- [Format Success Rate (Pie)](<https://www.sota2.com/research/sota/chart-data-extraction-on-exchart-bench-1-0-test.md?metric=Format%20Success%20Rate%20(Pie)>): Higher is better; unit: PERCENT.

- [Format Success Rate (Radar)](<https://www.sota2.com/research/sota/chart-data-extraction-on-exchart-bench-1-0-test.md?metric=Format%20Success%20Rate%20(Radar)>): Higher is better; unit: PERCENT.

- [Adaptive MAPE (All)](<https://www.sota2.com/research/sota/chart-data-extraction-on-exchart-bench-1-0-test.md?metric=Adaptive%20MAPE%20(All)>): Lower is better; unit: PERCENT.

- [Adaptive MAPE (Bar)](<https://www.sota2.com/research/sota/chart-data-extraction-on-exchart-bench-1-0-test.md?metric=Adaptive%20MAPE%20(Bar)>): Lower is better; unit: PERCENT.

- [Adaptive MAPE (Line)](<https://www.sota2.com/research/sota/chart-data-extraction-on-exchart-bench-1-0-test.md?metric=Adaptive%20MAPE%20(Line)>): Lower is better; unit: PERCENT.

- [Adaptive MAPE (Scatter)](<https://www.sota2.com/research/sota/chart-data-extraction-on-exchart-bench-1-0-test.md?metric=Adaptive%20MAPE%20(Scatter)>): Lower is better; unit: PERCENT.

- [Adaptive MAPE (Pie)](<https://www.sota2.com/research/sota/chart-data-extraction-on-exchart-bench-1-0-test.md?metric=Adaptive%20MAPE%20(Pie)>): Lower is better; unit: PERCENT.

- [Adaptive MAPE (Radar)](<https://www.sota2.com/research/sota/chart-data-extraction-on-exchart-bench-1-0-test.md?metric=Adaptive%20MAPE%20(Radar)>): Lower is better; unit: PERCENT.

## Recorded results

| Method | Format Success Rate (All) | Result date | Report paper |
| --- | --- | --- | --- |
| GPT-4o | 99.83 | Not recorded | 6a434da0ba73165dbb743d8c |
| Gemini 2.5 Flash | 99.72 | Not recorded | 6a434da0ba73165dbb743d8c |
| GLM-4.5V (\>100B) | 99.39 | Not recorded | 6a434da0ba73165dbb743d8c |
| Qwen2.5-VL 72B | 99.28 | Not recorded | 6a434da0ba73165dbb743d8c |
| ExChart (7B) | 99.11 | Not recorded | 6a434da0ba73165dbb743d8c |
| ExChart w/o CSPE (7B) | 98.94 | Not recorded | 6a434da0ba73165dbb743d8c |
| GPT-4.1 | 98.89 | Not recorded | 6a434da0ba73165dbb743d8c |
| InternVL3.5 4B | 98.69 | Not recorded | 6a434da0ba73165dbb743d8c |
| Qwen2.5-VL 32B | 97.83 | Not recorded | 6a434da0ba73165dbb743d8c |
| ExChart w/o CTA (7B) | 95.78 | Not recorded | 6a434da0ba73165dbb743d8c |
| InternVL3.5 38B | 94.14 | Not recorded | 6a434da0ba73165dbb743d8c |
| Qwen2.5-VL 7B | 89.58 | Not recorded | 6a434da0ba73165dbb743d8c |
| InternVL3.5 14B | 74.56 | Not recorded | 6a434da0ba73165dbb743d8c |
| LLaVA 1.6 Mistral 7B | 66.92 | Not recorded | 6a434da0ba73165dbb743d8c |

### GPT-4o

Format Success Rate (All): 99.83. Result ID: 6a4386d3fbc373592c9535b2.

[Reporting paper](<https://www.sota2.com/research/paper/6a434da0ba73165dbb743d8c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75681e209131cea67a98.md>)

Recorded evaluation configuration: {"Model Category":"Commercial proprietary models"}. This record does not assert independent reproduction.

### Gemini 2.5 Flash

Format Success Rate (All): 99.72. Result ID: 6a4386d3fbc373592c9535b4.

[Reporting paper](<https://www.sota2.com/research/paper/6a434da0ba73165dbb743d8c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c311e209131cea8cef5.md>)

Recorded evaluation configuration: {"Model Category":"Commercial proprietary models"}. This record does not assert independent reproduction.

### GLM-4.5V (\>100B)

Format Success Rate (All): 99.39. Result ID: 6a4386d3fbc373592c9535bf.

[Reporting paper](<https://www.sota2.com/research/paper/6a434da0ba73165dbb743d8c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7aaf1e209131cea849ba.md>)

Recorded evaluation configuration: {"Model Category":"Open-source models","Size":"\>100B"}. This record does not assert independent reproduction.

### Qwen2.5-VL 72B

Format Success Rate (All): 99.28. Result ID: 6a4386d3fbc373592c9535be.

[Reporting paper](<https://www.sota2.com/research/paper/6a434da0ba73165dbb743d8c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa773a1e209131cea71d3c.md>)

Recorded evaluation configuration: {"Model Category":"Open-source models","Size":"72B"}. This record does not assert independent reproduction.

### ExChart (7B)

Format Success Rate (All): 99.11. Result ID: 6a4386d3fbc373592c9535ba.

[Reporting paper](<https://www.sota2.com/research/paper/6a434da0ba73165dbb743d8c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a434da0ba73165dbb743d8c.md>)

Recorded evaluation configuration: {"Model Category":"Open-source models","Size":"7B"}. This record does not assert independent reproduction.

### ExChart w/o CSPE (7B)

Format Success Rate (All): 98.94. Result ID: 6a4386d3fbc373592c9535b8.

[Reporting paper](<https://www.sota2.com/research/paper/6a434da0ba73165dbb743d8c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a434da0ba73165dbb743d8c.md>)

Recorded evaluation configuration: {"Model Category":"Open-source models","Size":"7B","Ablation":"w/o CSPE"}. This record does not assert independent reproduction.

### GPT-4.1

Format Success Rate (All): 98.89. Result ID: 6a4386d3fbc373592c9535b3.

[Reporting paper](<https://www.sota2.com/research/paper/6a434da0ba73165dbb743d8c.md>)

Recorded evaluation configuration: {"Model Category":"Commercial proprietary models"}. This record does not assert independent reproduction.

### InternVL3.5 4B

Format Success Rate (All): 98.69. Result ID: 6a4386d3fbc373592c9535b5.

[Reporting paper](<https://www.sota2.com/research/paper/6a434da0ba73165dbb743d8c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71de1e209131cea5370d.md>)

[Implementation](<https://github.com/OpenGVLab/InternVL>)

Recorded evaluation configuration: {"Model Category":"Open-source models","Size":"4B"}. This record does not assert independent reproduction.

### Qwen2.5-VL 32B

Format Success Rate (All): 97.83. Result ID: 6a4386d3fbc373592c9535bc.

[Reporting paper](<https://www.sota2.com/research/paper/6a434da0ba73165dbb743d8c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa773a1e209131cea71d3c.md>)

Recorded evaluation configuration: {"Model Category":"Open-source models","Size":"32B"}. This record does not assert independent reproduction.

### ExChart w/o CTA (7B)

Format Success Rate (All): 95.78. Result ID: 6a4386d3fbc373592c9535b9.

[Reporting paper](<https://www.sota2.com/research/paper/6a434da0ba73165dbb743d8c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a434da0ba73165dbb743d8c.md>)

Recorded evaluation configuration: {"Model Category":"Open-source models","Size":"7B","Ablation":"w/o CTA"}. This record does not assert independent reproduction.

### InternVL3.5 38B

Format Success Rate (All): 94.14. Result ID: 6a4386d3fbc373592c9535bd.

[Reporting paper](<https://www.sota2.com/research/paper/6a434da0ba73165dbb743d8c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71de1e209131cea5370d.md>)

[Implementation](<https://github.com/OpenGVLab/InternVL>)

Recorded evaluation configuration: {"Model Category":"Open-source models","Size":"38B"}. This record does not assert independent reproduction.

### Qwen2.5-VL 7B

Format Success Rate (All): 89.58. Result ID: 6a4386d3fbc373592c9535b7.

[Reporting paper](<https://www.sota2.com/research/paper/6a434da0ba73165dbb743d8c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa773a1e209131cea71d3c.md>)

Recorded evaluation configuration: {"Model Category":"Open-source models","Size":"7B"}. This record does not assert independent reproduction.

### InternVL3.5 14B

Format Success Rate (All): 74.56. Result ID: 6a4386d3fbc373592c9535bb.

[Reporting paper](<https://www.sota2.com/research/paper/6a434da0ba73165dbb743d8c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71de1e209131cea5370d.md>)

[Implementation](<https://github.com/OpenGVLab/InternVL>)

Recorded evaluation configuration: {"Model Category":"Open-source models","Size":"14B"}. This record does not assert independent reproduction.

### LLaVA 1.6 Mistral 7B

Format Success Rate (All): 66.92. Result ID: 6a4386d3fbc373592c9535b6.

[Reporting paper](<https://www.sota2.com/research/paper/6a434da0ba73165dbb743d8c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa732e1e209131cea5af59.md>)

[Implementation](<https://llava-vl.github.io/>)

Recorded evaluation configuration: {"Model Category":"Open-source models","Size":"7B"}. This record does not assert independent reproduction.

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