# Multi-turn Conversation Evaluation on MT-Bench 1.0 (test)

Source: [Multi-turn Conversation Evaluation on MT-Bench 1.0 (test)](<https://www.sota2.com/research/sota/multi-turn-conversation-evaluation-on-mt-bench-1-0-test>)

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

Data updated: 2026-02-26T05:45:46.577Z

Evaluation of multi-turn conversational ability using the MT-Bench 1.0 test set, specifically reporting scores derived from GPT-4 and GPT-4-Turbo model judgments.

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 | Multi-turn Conversation Evaluation |
| Dataset (including any stated split) | MT-Bench 1.0 (test) |
| Metric | GPT-4 Score |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 5 of 5 |
| 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

- [GPT-4 Score](<https://www.sota2.com/research/sota/multi-turn-conversation-evaluation-on-mt-bench-1-0-test.md?metric=GPT-4%20Score>): Higher is better; unit: SCALAR.

- [GPT-4-Turbo Score](<https://www.sota2.com/research/sota/multi-turn-conversation-evaluation-on-mt-bench-1-0-test.md?metric=GPT-4-Turbo%20Score>): Higher is better; unit: SCALAR.

## Recorded results

| Method | GPT-4 Score | Result date | Report paper |
| --- | --- | --- | --- |
| DAR | 8.538 | Not recorded | 698eaff3c97a0b3e3b5e6617 |
| GRPO | 8.425 | Not recorded | 698eaff3c97a0b3e3b5e6617 |
| RLOO | 8.409 | Not recorded | 698eaff3c97a0b3e3b5e6617 |
| Iter-SFT | 8.378 | Not recorded | 698eaff3c97a0b3e3b5e6617 |
| Qwen2-7B-Instruct (π0) | 8.334 | Not recorded | 698eaff3c97a0b3e3b5e6617 |

### DAR

GPT-4 Score: 8.538. Result ID: 6998bfe45698665ec422fcd6.

[Reporting paper](<https://www.sota2.com/research/paper/698eaff3c97a0b3e3b5e6617.md>)

[Method origin paper](<https://www.sota2.com/research/paper/698eaff3c97a0b3e3b5e6617.md>)

Recorded evaluation configuration: {"Response Length":"1358"}. This record does not assert independent reproduction.

### GRPO

GPT-4 Score: 8.425. Result ID: 6998bfe45698665ec422fcd4.

[Reporting paper](<https://www.sota2.com/research/paper/698eaff3c97a0b3e3b5e6617.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73101e209131cea5a395.md>)

[Implementation](<https://github.com/deepseek-ai/DeepSeek-Math>)

Recorded evaluation configuration: {"Response Length":"1559"}. This record does not assert independent reproduction.

### RLOO

GPT-4 Score: 8.409. Result ID: 6998bfe45698665ec422fcd3.

[Reporting paper](<https://www.sota2.com/research/paper/698eaff3c97a0b3e3b5e6617.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa725a1e209131cea55df7.md>)

Recorded evaluation configuration: {"Response Length":"1580"}. This record does not assert independent reproduction.

### Iter-SFT

GPT-4 Score: 8.378. Result ID: 6998bfe45698665ec422fcd5.

[Reporting paper](<https://www.sota2.com/research/paper/698eaff3c97a0b3e3b5e6617.md>)

Recorded evaluation configuration: {"Response Length":"1343"}. This record does not assert independent reproduction.

### Qwen2-7B-Instruct (π0)

GPT-4 Score: 8.334. Result ID: 6998bfe45698665ec422fcd2.

[Reporting paper](<https://www.sota2.com/research/paper/698eaff3c97a0b3e3b5e6617.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa767e1e209131cea6d6a7.md>)

Recorded evaluation configuration: {"Response Length":"1340"}. This record does not assert independent reproduction.

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