# Deep Search Reasoning on XBench DeepSearch2505

Source: [Deep Search Reasoning on XBench DeepSearch2505](<https://www.sota2.com/research/sota/deep-search-reasoning-on-xbench-deepsearch2505>)

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

Data updated: 2026-02-26T06:03:50.670Z

Evaluation of model performance on Deep Search Reasoning capabilities using the XBench DeepSearch2505 dataset, measured by Score.

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 Search Reasoning |
| Dataset (including any stated split) | XBench DeepSearch2505 |
| Metric | Score |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 9 of 9 |
| 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

- [Score](<https://www.sota2.com/research/sota/deep-search-reasoning-on-xbench-deepsearch2505.md?metric=Score>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Score | Result date | Report paper |
| --- | --- | --- | --- |
| Claude-3.7-Sonnet | 41 | Not recorded | 6982d2aba8b01e76cc8e957f |
| CSO | 29 | Not recorded | 6982d2aba8b01e76cc8e957f |
| GPT-4.1 | 27 | Not recorded | 6982d2aba8b01e76cc8e957f |
| Step-DPO | 25 | Not recorded | 6982d2aba8b01e76cc8e957f |
| IPR | 24 | Not recorded | 6982d2aba8b01e76cc8e957f |
| CK-Pro-8B | 23 | Not recorded | 6982d2aba8b01e76cc8e957f |
| ETO | 22 | Not recorded | 6982d2aba8b01e76cc8e957f |
| RFT | 20 | Not recorded | 6982d2aba8b01e76cc8e957f |
| Qwen3-8B | 7 | Not recorded | 6982d2aba8b01e76cc8e957f |

### Claude-3.7-Sonnet

Score: 41. Result ID: 69964fea5698665ec494aaad.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2aba8b01e76cc8e957f.md>)

Recorded evaluation configuration: {"Model Type":"Proprietary"}. This record does not assert independent reproduction.

### CSO

Score: 29. Result ID: 69964fea5698665ec494aab4.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2aba8b01e76cc8e957f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6982d2aba8b01e76cc8e957f.md>)

Recorded evaluation configuration: {"Base Model":"CK-Pro-8B","Post-Training Method":"Critical Step Optimization"}. This record does not assert independent reproduction.

### GPT-4.1

Score: 27. Result ID: 69964fea5698665ec494aaac.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2aba8b01e76cc8e957f.md>)

Recorded evaluation configuration: {"Model Type":"Proprietary"}. This record does not assert independent reproduction.

### Step-DPO

Score: 25. Result ID: 69964fea5698665ec494aab2.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2aba8b01e76cc8e957f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77041e209131cea70816.md>)

Recorded evaluation configuration: {"Base Model":"CK-Pro-8B","Post-Training Method":"Step-wise DPO"}. This record does not assert independent reproduction.

### IPR

Score: 24. Result ID: 69964fea5698665ec494aab3.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2aba8b01e76cc8e957f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74f81e209131cea64fa0.md>)

Recorded evaluation configuration: {"Base Model":"CK-Pro-8B","Post-Training Method":"Iterative Process Refinement"}. This record does not assert independent reproduction.

### CK-Pro-8B

Score: 23. Result ID: 69964fea5698665ec494aaaf.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2aba8b01e76cc8e957f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a941e209131cea84031.md>)

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

### ETO

Score: 22. Result ID: 69964fea5698665ec494aab0.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2aba8b01e76cc8e957f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74101e209131cea5f92e.md>)

Recorded evaluation configuration: {"Base Model":"CK-Pro-8B","Post-Training Method":"Exploration-based Trajectory Optimization"}. This record does not assert independent reproduction.

### RFT

Score: 20. Result ID: 69964fea5698665ec494aab1.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2aba8b01e76cc8e957f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a1a1e209131cea812fc.md>)

[Implementation](<https://github.com/iiis-ai/cumulative-reasoning>)

Recorded evaluation configuration: {"Base Model":"CK-Pro-8B","Post-Training Method":"Rejection Sampling Fine-Tuning"}. This record does not assert independent reproduction.

### Qwen3-8B

Score: 7. Result ID: 69964fea5698665ec494aaae.

[Reporting paper](<https://www.sota2.com/research/paper/6982d2aba8b01e76cc8e957f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa788e1e209131cea79116.md>)

Recorded evaluation configuration: {"Model Type":"Open-Source Baseline"}. This record does not assert independent reproduction.

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