# Dialogue Generation on Norm-grounded Dialogue (Chinese) (test)

Source: [Dialogue Generation on Norm-grounded Dialogue (Chinese) (test)](<https://www.sota2.com/research/sota/dialogue-generation-on-norm-grounded-dialogue-chinese-test>)

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

Data updated: 2026-03-13T07:08:02.312Z

Evaluation of dialogue generation performance on the test split of the Norm-grounded Dialogue (Chinese) dataset, using human preference metrics against baseline models NormDial and SODA.

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 | Dialogue Generation |
| Dataset (including any stated split) | Norm-grounded Dialogue (Chinese) (test) |
| Metric | Win Rate vs NormDial |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 3 of 3 |
| 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

- [Win Rate vs NormDial](<https://www.sota2.com/research/sota/dialogue-generation-on-norm-grounded-dialogue-chinese-test.md?metric=Win%20Rate%20vs%20NormDial>): Higher is better; unit: PERCENT.

- [Win Rate vs SODA](<https://www.sota2.com/research/sota/dialogue-generation-on-norm-grounded-dialogue-chinese-test.md?metric=Win%20Rate%20vs%20SODA>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Win Rate vs NormDial | Result date | Report paper |
| --- | --- | --- | --- |
| Qwen-2.5-32B | 79 | Not recorded | 68fa7b7f1e209131cea88b10 |
| GPT-4o-mini | 75 | Not recorded | 68fa7b7f1e209131cea88b10 |
| Qwen-2.5-14B | 71 | Not recorded | 68fa7b7f1e209131cea88b10 |

### Qwen-2.5-32B

Win Rate vs NormDial: 79. Result ID: 69b3b5f847f34ce9249d2d53.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7b7f1e209131cea88b10.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7b7f1e209131cea88b10.md>)

Recorded evaluation configuration: {"Model":"Qwen-2.5-32B"}. This record does not assert independent reproduction.

### GPT-4o-mini

Win Rate vs NormDial: 75. Result ID: 69b3b5f847f34ce9249d2d51.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7b7f1e209131cea88b10.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7b7f1e209131cea88b10.md>)

Recorded evaluation configuration: {"Model":"GPT-4o-mini"}. This record does not assert independent reproduction.

### Qwen-2.5-14B

Win Rate vs NormDial: 71. Result ID: 69b3b5f847f34ce9249d2d52.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7b7f1e209131cea88b10.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7b7f1e209131cea88b10.md>)

Recorded evaluation configuration: {"Model":"Qwen-2.5-14B"}. This record does not assert independent reproduction.

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