# STEM & Puzzle on MMMU (val)

Source: [STEM & Puzzle on MMMU (val)](<https://www.sota2.com/research/sota/stem-and-puzzle-on-mmmu-val>)

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

Data updated: 2026-07-02T08:29:22.655Z

Evaluation of model performance on the STEM & Puzzle task using the validation split of the MMMU dataset, measured by 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 | STEM & Puzzle |
| Dataset (including any stated split) | MMMU (val) |
| Metric | Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 26 of 26 |
| 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

- [Accuracy](<https://www.sota2.com/research/sota/stem-and-puzzle-on-mmmu-val.md?metric=Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| M2Note | 79.7 | Not recorded | 6a45f070dc71372d7795a090 |
| GPT5 mini | 79 | Not recorded | 697c3b66a8b01e76cc8da6e1 |
| Qwen3-VL 32B | 78.1 | Not recorded | 697c3b66a8b01e76cc8da6e1 |
| Vanilla | 77.9 | Not recorded | 6a45f070dc71372d7795a090 |
| M2Note | 77.9 | Not recorded | 6a45f070dc71372d7795a090 |
| Gemini-2.5 Flash | 77.7 | Not recorded | 697c3b66a8b01e76cc8da6e1 |
| Vanilla | 77.1 | Not recorded | 6a45f070dc71372d7795a090 |
| Qwen3-VL 30B-A3B | 76 | Not recorded | 697c3b66a8b01e76cc8da6e1 |
| Qwen3-VL 8B | 74.1 | Not recorded | 697c3b66a8b01e76cc8da6e1 |
| MMFineReason-8B | 71.3 | Not recorded | 697c3b66a8b01e76cc8da6e1 |
| MMFineReason-4B | 69.6 | Not recorded | 697c3b66a8b01e76cc8da6e1 |
| CoT+M2Note | 69.4 | Not recorded | 6a45f070dc71372d7795a090 |
| DPE | 69.1 | Not recorded | 6a45f070dc71372d7795a090 |
| M2Note | 68.9 | Not recorded | 6a45f070dc71372d7795a090 |
| CoT | 68.8 | Not recorded | 6a45f070dc71372d7795a090 |
| M2Note | 68.4 | Not recorded | 6a45f070dc71372d7795a090 |
| Vanilla | 67.1 | Not recorded | 6a45f070dc71372d7795a090 |
| Vanilla | 67.1 | Not recorded | 6a45f070dc71372d7795a090 |
| HoneyBee 8B | 63.1 | Not recorded | 697c3b66a8b01e76cc8da6e1 |
| MMR1 8B | 62.8 | Not recorded | 697c3b66a8b01e76cc8da6e1 |
| Vision-Zero | 58.8 | Not recorded | 6a45f070dc71372d7795a090 |
| OMR 7B | 57.8 | Not recorded | 697c3b66a8b01e76cc8da6e1 |
| VisPlay | 54.9 | Not recorded | 6a45f070dc71372d7795a090 |
| MMFineReason-2B | 54.8 | Not recorded | 697c3b66a8b01e76cc8da6e1 |
| iReasoner | 52.4 | Not recorded | 6a45f070dc71372d7795a090 |
| EvoLMM | 52 | Not recorded | 6a45f070dc71372d7795a090 |

### M2Note

Accuracy: 79.7. Result ID: 6a46216160b88bb339850dbe.

[Reporting paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

Recorded evaluation configuration: {"Backbone":"Qwen3-VL-Plus","Evolving Mode":"Self-Evolving","Cost":"≈ $7","Mem":"28","Len":"462"}. This record does not assert independent reproduction.

### GPT5 mini

Accuracy: 79. Result ID: 6997797c5698665ec4cdd04c.

[Reporting paper](<https://www.sota2.com/research/paper/697c3b66a8b01e76cc8da6e1.md>)

Recorded evaluation configuration: {"Model Source Category":"Closed-source VLMs","Evaluation Mode":"Thinking Mode"}. This record does not assert independent reproduction.

### Qwen3-VL 32B

Accuracy: 78.1. Result ID: 6997797c5698665ec4cdd04f.

[Reporting paper](<https://www.sota2.com/research/paper/697c3b66a8b01e76cc8da6e1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6927db54d6b0dd3342930f66.md>)

Recorded evaluation configuration: {"Model Source Category":"Open-weight VLMs","Model Parameters":"32B","Evaluation Mode":"Thinking Mode"}. This record does not assert independent reproduction.

### Vanilla

Accuracy: 77.9. Result ID: 6a46216160b88bb339850dbd.

[Reporting paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

Recorded evaluation configuration: {"Backbone":"Qwen3-VL-Plus","Evolving Mode":"Self-Evolving"}. This record does not assert independent reproduction.

### M2Note

Accuracy: 77.9. Result ID: 6a46216160b88bb339850dc0.

[Reporting paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

Recorded evaluation configuration: {"Backbone":"GPT-5.4","Evolving Mode":"Self-Evolving","Cost":"≈ $7","Mem":"28","Len":"462"}. This record does not assert independent reproduction.

### Gemini-2.5 Flash

Accuracy: 77.7. Result ID: 6997797c5698665ec4cdd04b.

[Reporting paper](<https://www.sota2.com/research/paper/697c3b66a8b01e76cc8da6e1.md>)

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

Recorded evaluation configuration: {"Model Source Category":"Closed-source VLMs","Evaluation Mode":"Thinking Mode"}. This record does not assert independent reproduction.

### Vanilla

Accuracy: 77.1. Result ID: 6a46216160b88bb339850dbf.

[Reporting paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

Recorded evaluation configuration: {"Backbone":"GPT-5.4","Evolving Mode":"Self-Evolving"}. This record does not assert independent reproduction.

### Qwen3-VL 30B-A3B

Accuracy: 76. Result ID: 6997797c5698665ec4cdd04e.

[Reporting paper](<https://www.sota2.com/research/paper/697c3b66a8b01e76cc8da6e1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6927db54d6b0dd3342930f66.md>)

Recorded evaluation configuration: {"Model Source Category":"Open-weight VLMs","Model Parameters":"30B-A3B","Evaluation Mode":"Thinking Mode"}. This record does not assert independent reproduction.

### Qwen3-VL 8B

Accuracy: 74.1. Result ID: 6997797c5698665ec4cdd04d.

[Reporting paper](<https://www.sota2.com/research/paper/697c3b66a8b01e76cc8da6e1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6927db54d6b0dd3342930f66.md>)

Recorded evaluation configuration: {"Model Source Category":"Open-weight VLMs","Model Parameters":"8B","Evaluation Mode":"Thinking Mode"}. This record does not assert independent reproduction.

### MMFineReason-8B

Accuracy: 71.3. Result ID: 6997797c5698665ec4cdd055.

[Reporting paper](<https://www.sota2.com/research/paper/697c3b66a8b01e76cc8da6e1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/697c3b66a8b01e76cc8da6e1.md>)

Recorded evaluation configuration: {"Model Source Category":"Ours","Model Parameters":"8B","Evaluation Mode":"Thinking Mode"}. This record does not assert independent reproduction.

### MMFineReason-4B

Accuracy: 69.6. Result ID: 6997797c5698665ec4cdd054.

[Reporting paper](<https://www.sota2.com/research/paper/697c3b66a8b01e76cc8da6e1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/697c3b66a8b01e76cc8da6e1.md>)

Recorded evaluation configuration: {"Model Source Category":"Ours","Model Parameters":"4B","Evaluation Mode":"Thinking Mode"}. This record does not assert independent reproduction.

### CoT+M2Note

Accuracy: 69.4. Result ID: 6a46216160b88bb339850dbc.

[Reporting paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

Recorded evaluation configuration: {"Backbone":"Qwen3-VL-8B-Instruct","Evolving Mode":"Self-Evolving","Cost":"≈ $4","Mem":"32","Len":"315"}. This record does not assert independent reproduction.

### DPE

Accuracy: 69.1. Result ID: 6a46216160b88bb339850db9.

[Reporting paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69a124f12d603e6de3e3c3ab.md>)

[Implementation](<https://github.com/hongruijia/DPE>)

Recorded evaluation configuration: {"Backbone":"Qwen3-VL-8B-Instruct","Evolving Mode":"Self-Evolving","Cost":"\> $500"}. This record does not assert independent reproduction.

### M2Note

Accuracy: 68.9. Result ID: 6a46216160b88bb339850dc2.

[Reporting paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

Recorded evaluation configuration: {"Backbone (Tuning)":"Qwen3-VL-8B-Instruct","Backbone (Tuner)":"Qwen3-VL-Plus","Evolving Mode":"Cross-Model Evolving","Cost":"≈ $5","Mem":"43","Len":"445"}. This record does not assert independent reproduction.

### CoT

Accuracy: 68.8. Result ID: 6a46216160b88bb339850dbb.

[Reporting paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e571e209131cea3f27b.md>)

Recorded evaluation configuration: {"Backbone":"Qwen3-VL-8B-Instruct","Evolving Mode":"Self-Evolving"}. This record does not assert independent reproduction.

### M2Note

Accuracy: 68.4. Result ID: 6a46216160b88bb339850dba.

[Reporting paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

Recorded evaluation configuration: {"Backbone":"Qwen3-VL-8B-Instruct","Evolving Mode":"Self-Evolving","Cost":"≈ $3","Mem":"50","Len":"261"}. This record does not assert independent reproduction.

### Vanilla

Accuracy: 67.1. Result ID: 6a46216160b88bb339850db8.

[Reporting paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

Recorded evaluation configuration: {"Backbone":"Qwen3-VL-8B-Instruct","Evolving Mode":"Self-Evolving"}. This record does not assert independent reproduction.

### Vanilla

Accuracy: 67.1. Result ID: 6a46216160b88bb339850dc1.

[Reporting paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

Recorded evaluation configuration: {"Backbone (Tuning)":"Qwen3-VL-8B-Instruct","Backbone (Tuner)":"Qwen3-VL-Plus","Evolving Mode":"Cross-Model Evolving"}. This record does not assert independent reproduction.

### HoneyBee 8B

Accuracy: 63.1. Result ID: 6997797c5698665ec4cdd051.

[Reporting paper](<https://www.sota2.com/research/paper/697c3b66a8b01e76cc8da6e1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c1a1e209131cea8c598.md>)

[Implementation](<https://github.com/facebookresearch/HoneyBee_VLM>)

Recorded evaluation configuration: {"Model Source Category":"Open-source VLMs","Model Parameters":"8B","Evaluation Mode":"Thinking Mode"}. This record does not assert independent reproduction.

### MMR1 8B

Accuracy: 62.8. Result ID: 6997797c5698665ec4cdd050.

[Reporting paper](<https://www.sota2.com/research/paper/697c3b66a8b01e76cc8da6e1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7b8f1e209131cea8920f.md>)

Recorded evaluation configuration: {"Model Source Category":"Open-source VLMs","Model Parameters":"8B","Evaluation Mode":"Thinking Mode"}. This record does not assert independent reproduction.

### Vision-Zero

Accuracy: 58.8. Result ID: 6a46216160b88bb339850db4.

[Reporting paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7bb11e209131cea89eff.md>)

Recorded evaluation configuration: {"Evolving Mode":"Self-Evolving","Cost":"Basically"}. This record does not assert independent reproduction.

### OMR 7B

Accuracy: 57.8. Result ID: 6997797c5698665ec4cdd052.

[Reporting paper](<https://www.sota2.com/research/paper/697c3b66a8b01e76cc8da6e1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/691ff1f4d6b0dd334292e589.md>)

[Implementation](<https://github.com/EvolvingLMMs-Lab/OpenMMReasoner>)

Recorded evaluation configuration: {"Model Source Category":"Open-source VLMs","Model Parameters":"7B","Evaluation Mode":"Thinking Mode"}. This record does not assert independent reproduction.

### VisPlay

Accuracy: 54.9. Result ID: 6a46216160b88bb339850db7.

[Reporting paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

[Method origin paper](<https://www.sota2.com/research/paper/691ea0bfd6b0dd334292e137.md>)

[Implementation](<https://github.com/bruno686/VisPlay>)

Recorded evaluation configuration: {"Evolving Mode":"Self-Evolving"}. This record does not assert independent reproduction.

### MMFineReason-2B

Accuracy: 54.8. Result ID: 6997797c5698665ec4cdd053.

[Reporting paper](<https://www.sota2.com/research/paper/697c3b66a8b01e76cc8da6e1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/697c3b66a8b01e76cc8da6e1.md>)

Recorded evaluation configuration: {"Model Source Category":"Ours","Model Parameters":"2B","Evaluation Mode":"Thinking Mode"}. This record does not assert independent reproduction.

### iReasoner

Accuracy: 52.4. Result ID: 6a46216160b88bb339850db6.

[Reporting paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69647fda53fdd5c3617445c3.md>)

Recorded evaluation configuration: {"Evolving Mode":"Self-Evolving"}. This record does not assert independent reproduction.

### EvoLMM

Accuracy: 52. Result ID: 6a46216160b88bb339850db5.

[Reporting paper](<https://www.sota2.com/research/paper/6a45f070dc71372d7795a090.md>)

[Method origin paper](<https://www.sota2.com/research/paper/691ff210d6b0dd334292e6db.md>)

Recorded evaluation configuration: {"Evolving Mode":"Self-Evolving","Cost":"\> $500"}. This record does not assert independent reproduction.

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