# Multi-step Reasoning on BrowseComp-Plus OOD

Source: [Multi-step Reasoning on BrowseComp-Plus OOD](<https://www.sota2.com/research/sota/multi-step-reasoning-on-browsecomp-plus-ood>)

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

Data updated: 2026-04-08T06:56:54.304Z

Evaluation of deep research agents performing multi-step reasoning on the Out-of-Distribution (OOD) subset of the BrowseComp-Plus dataset, measured by Success Rate (SR), Recall, and average steps taken.

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 | Deep research agents / Multi-step reasoning |
| Dataset (including any stated split) | BrowseComp-Plus OOD |
| Metric | Success Rate (SR) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 24 of 24 |
| 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

- [Success Rate (SR)](<https://www.sota2.com/research/sota/multi-step-reasoning-on-browsecomp-plus-ood.md?metric=Success%20Rate%20(SR)>): Higher is better; unit: PERCENT.

- [Recall](<https://www.sota2.com/research/sota/multi-step-reasoning-on-browsecomp-plus-ood.md?metric=Recall>): Higher is better; unit: PERCENT.

- [Average Steps](<https://www.sota2.com/research/sota/multi-step-reasoning-on-browsecomp-plus-ood.md?metric=Average%20Steps>): Lower is better; unit: SCALAR.

## Recorded results

| Method | Success Rate (SR) | Result date | Report paper |
| --- | --- | --- | --- |
| Qwen3-Emb + LRAT | 54.6 | Not recorded | 69d5e0de7d0172140ec758c1 |
| E5-Large + LRAT | 50.6 | Not recorded | 69d5e0de7d0172140ec758c1 |
| E5-Large + LRAT | 48.7 | Not recorded | 69d5e0de7d0172140ec758c1 |
| Qwen3-Emb + LRAT | 48.3 | Not recorded | 69d5e0de7d0172140ec758c1 |
| E5-Large | 46.4 | Not recorded | 69d5e0de7d0172140ec758c1 |
| E5-Large | 46.4 | Not recorded | 69d5e0de7d0172140ec758c1 |
| Qwen3-Emb | 43.9 | Not recorded | 69d5e0de7d0172140ec758c1 |
| Qwen3-Emb | 38.2 | Not recorded | 69d5e0de7d0172140ec758c1 |
| E5-Large + LRAT | 29 | Not recorded | 69d5e0de7d0172140ec758c1 |
| Qwen3-Emb + LRAT | 27.2 | Not recorded | 69d5e0de7d0172140ec758c1 |
| E5-Large | 25.4 | Not recorded | 69d5e0de7d0172140ec758c1 |
| E5-Large + LRAT | 23.9 | Not recorded | 69d5e0de7d0172140ec758c1 |
| Qwen3-Emb + LRAT | 23.7 | Not recorded | 69d5e0de7d0172140ec758c1 |
| Qwen3-Emb | 21 | Not recorded | 69d5e0de7d0172140ec758c1 |
| E5-Large | 20.7 | Not recorded | 69d5e0de7d0172140ec758c1 |
| Qwen3-Emb | 17.8 | Not recorded | 69d5e0de7d0172140ec758c1 |
| E5-Large | 15.9 | Not recorded | 69d5e0de7d0172140ec758c1 |
| E5-Large + LRAT | 15.9 | Not recorded | 69d5e0de7d0172140ec758c1 |
| Qwen3-Emb + LRAT | 15.8 | Not recorded | 69d5e0de7d0172140ec758c1 |
| Qwen3-Emb | 13.5 | Not recorded | 69d5e0de7d0172140ec758c1 |
| E5-Large + LRAT | 13.1 | Not recorded | 69d5e0de7d0172140ec758c1 |
| Qwen3-Emb + LRAT | 12.1 | Not recorded | 69d5e0de7d0172140ec758c1 |
| E5-Large | 10.8 | Not recorded | 69d5e0de7d0172140ec758c1 |
| Qwen3-Emb | 9 | Not recorded | 69d5e0de7d0172140ec758c1 |

### Qwen3-Emb + LRAT

Success Rate (SR): 54.6. Result ID: 69d5fbb147f34ce924be01f5.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Implementation](<https://github.com/Yuqi-Zhou/LRAT>)

Recorded evaluation configuration: {"Agent Backbone":"GLM-4.7 (358B)"}. This record does not assert independent reproduction.

### E5-Large + LRAT

Success Rate (SR): 50.6. Result ID: 69d5fbb147f34ce924be01f7.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Implementation](<https://github.com/Yuqi-Zhou/LRAT>)

Recorded evaluation configuration: {"Agent Backbone":"GLM-4.7 (358B)"}. This record does not assert independent reproduction.

### E5-Large + LRAT

Success Rate (SR): 48.7. Result ID: 69d5fbb147f34ce924be01f3.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Implementation](<https://github.com/Yuqi-Zhou/LRAT>)

Recorded evaluation configuration: {"Agent Backbone":"MiniMax-M2.1 (229B)"}. This record does not assert independent reproduction.

### Qwen3-Emb + LRAT

Success Rate (SR): 48.3. Result ID: 69d5fbb147f34ce924be01f1.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Implementation](<https://github.com/Yuqi-Zhou/LRAT>)

Recorded evaluation configuration: {"Agent Backbone":"MiniMax-M2.1 (229B)"}. This record does not assert independent reproduction.

### E5-Large

Success Rate (SR): 46.4. Result ID: 69d5fbb147f34ce924be01f2.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72251e209131cea54c85.md>)

Recorded evaluation configuration: {"Agent Backbone":"MiniMax-M2.1 (229B)"}. This record does not assert independent reproduction.

### E5-Large

Success Rate (SR): 46.4. Result ID: 69d5fbb147f34ce924be01f6.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72251e209131cea54c85.md>)

Recorded evaluation configuration: {"Agent Backbone":"GLM-4.7 (358B)"}. This record does not assert independent reproduction.

### Qwen3-Emb

Success Rate (SR): 43.9. Result ID: 69d5fbb147f34ce924be01f4.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79601e209131cea7d516.md>)

Recorded evaluation configuration: {"Agent Backbone":"GLM-4.7 (358B)"}. This record does not assert independent reproduction.

### Qwen3-Emb

Success Rate (SR): 38.2. Result ID: 69d5fbb147f34ce924be01f0.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79601e209131cea7d516.md>)

Recorded evaluation configuration: {"Agent Backbone":"MiniMax-M2.1 (229B)"}. This record does not assert independent reproduction.

### E5-Large + LRAT

Success Rate (SR): 29. Result ID: 69d5fbb147f34ce924be01e7.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Implementation](<https://github.com/Yuqi-Zhou/LRAT>)

Recorded evaluation configuration: {"Agent Backbone":"WebExplore (8B)"}. This record does not assert independent reproduction.

### Qwen3-Emb + LRAT

Success Rate (SR): 27.2. Result ID: 69d5fbb147f34ce924be01e5.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Implementation](<https://github.com/Yuqi-Zhou/LRAT>)

Recorded evaluation configuration: {"Agent Backbone":"WebExplore (8B)"}. This record does not assert independent reproduction.

### E5-Large

Success Rate (SR): 25.4. Result ID: 69d5fbb147f34ce924be01e6.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72251e209131cea54c85.md>)

Recorded evaluation configuration: {"Agent Backbone":"WebExplore (8B)"}. This record does not assert independent reproduction.

### E5-Large + LRAT

Success Rate (SR): 23.9. Result ID: 69d5fbb147f34ce924be01eb.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Implementation](<https://github.com/Yuqi-Zhou/LRAT>)

Recorded evaluation configuration: {"Agent Backbone":"Tongyi-DeepResearch (30B)"}. This record does not assert independent reproduction.

### Qwen3-Emb + LRAT

Success Rate (SR): 23.7. Result ID: 69d5fbb147f34ce924be01e9.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Implementation](<https://github.com/Yuqi-Zhou/LRAT>)

Recorded evaluation configuration: {"Agent Backbone":"Tongyi-DeepResearch (30B)"}. This record does not assert independent reproduction.

### Qwen3-Emb

Success Rate (SR): 21. Result ID: 69d5fbb147f34ce924be01e4.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79601e209131cea7d516.md>)

Recorded evaluation configuration: {"Agent Backbone":"WebExplore (8B)"}. This record does not assert independent reproduction.

### E5-Large

Success Rate (SR): 20.7. Result ID: 69d5fbb147f34ce924be01ea.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72251e209131cea54c85.md>)

Recorded evaluation configuration: {"Agent Backbone":"Tongyi-DeepResearch (30B)"}. This record does not assert independent reproduction.

### Qwen3-Emb

Success Rate (SR): 17.8. Result ID: 69d5fbb147f34ce924be01e8.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79601e209131cea7d516.md>)

Recorded evaluation configuration: {"Agent Backbone":"Tongyi-DeepResearch (30B)"}. This record does not assert independent reproduction.

### E5-Large

Success Rate (SR): 15.9. Result ID: 69d5fbb147f34ce924be01e2.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72251e209131cea54c85.md>)

Recorded evaluation configuration: {"Agent Backbone":"AgentCPM-Explore (4B)"}. This record does not assert independent reproduction.

### E5-Large + LRAT

Success Rate (SR): 15.9. Result ID: 69d5fbb147f34ce924be01e3.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Implementation](<https://github.com/Yuqi-Zhou/LRAT>)

Recorded evaluation configuration: {"Agent Backbone":"AgentCPM-Explore (4B)"}. This record does not assert independent reproduction.

### Qwen3-Emb + LRAT

Success Rate (SR): 15.8. Result ID: 69d5fbb147f34ce924be01e1.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Implementation](<https://github.com/Yuqi-Zhou/LRAT>)

Recorded evaluation configuration: {"Agent Backbone":"AgentCPM-Explore (4B)"}. This record does not assert independent reproduction.

### Qwen3-Emb

Success Rate (SR): 13.5. Result ID: 69d5fbb147f34ce924be01e0.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79601e209131cea7d516.md>)

Recorded evaluation configuration: {"Agent Backbone":"AgentCPM-Explore (4B)"}. This record does not assert independent reproduction.

### E5-Large + LRAT

Success Rate (SR): 13.1. Result ID: 69d5fbb147f34ce924be01ef.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Implementation](<https://github.com/Yuqi-Zhou/LRAT>)

Recorded evaluation configuration: {"Agent Backbone":"GPT-OSS (120B)"}. This record does not assert independent reproduction.

### Qwen3-Emb + LRAT

Success Rate (SR): 12.1. Result ID: 69d5fbb147f34ce924be01ed.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Implementation](<https://github.com/Yuqi-Zhou/LRAT>)

Recorded evaluation configuration: {"Agent Backbone":"GPT-OSS (120B)"}. This record does not assert independent reproduction.

### E5-Large

Success Rate (SR): 10.8. Result ID: 69d5fbb147f34ce924be01ee.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72251e209131cea54c85.md>)

Recorded evaluation configuration: {"Agent Backbone":"GPT-OSS (120B)"}. This record does not assert independent reproduction.

### Qwen3-Emb

Success Rate (SR): 9. Result ID: 69d5fbb147f34ce924be01ec.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e0de7d0172140ec758c1.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79601e209131cea7d516.md>)

Recorded evaluation configuration: {"Agent Backbone":"GPT-OSS (120B)"}. This record does not assert independent reproduction.

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