# Dialogue on Dialogue

Source: [Dialogue on Dialogue](<https://www.sota2.com/research/sota/dialogue-on-dialogue>)

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

Data updated: 2026-02-26T05:57:15.819Z

Evaluation of dialogue systems on the Dialogue dataset, measured using PandaLM, Reward, and Win metrics.

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 |
| Dataset (including any stated split) | Dialogue |
| Metric | PandaLM |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 18 of 18 |
| 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

- [PandaLM](<https://www.sota2.com/research/sota/dialogue-on-dialogue.md?metric=PandaLM>): Higher is better; unit: SCALAR.

- [Reward](<https://www.sota2.com/research/sota/dialogue-on-dialogue.md?metric=Reward>): Higher is better; unit: SCALAR.

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

## Recorded results

| Method | PandaLM | Result date | Report paper |
| --- | --- | --- | --- |
| HBAT | 77.79 | Not recorded | 68fa73b51e209131cea5e0b2 |
| HBAT-Freeze | 76.87 | Not recorded | 68fa73b51e209131cea5e0b2 |
| DPO | 75 | Not recorded | 68fa73b51e209131cea5e0b2 |
| HBAT | 74.78 | Not recorded | 68fa73b51e209131cea5e0b2 |
| HBAT | 72.83 | Not recorded | 68fa73b51e209131cea5e0b2 |
| RLHF | 72.21 | Not recorded | 68fa73b51e209131cea5e0b2 |
| HBAT-Freeze | 71.31 | Not recorded | 68fa73b51e209131cea5e0b2 |
| HBAT | 70.88 | Not recorded | 68fa73b51e209131cea5e0b2 |
| DPO | 70.74 | Not recorded | 68fa73b51e209131cea5e0b2 |
| HBAT-Freeze | 69.91 | Not recorded | 68fa73b51e209131cea5e0b2 |
| RLHF | 69.79 | Not recorded | 68fa73b51e209131cea5e0b2 |
| HBAT-Freeze | 68.78 | Not recorded | 68fa73b51e209131cea5e0b2 |
| RLHF+pt | 63.67 | Not recorded | 68fa73b51e209131cea5e0b2 |
| SFT | 62.16 | Not recorded | 68fa73b51e209131cea5e0b2 |
| SFT | 54.76 | Not recorded | 68fa73b51e209131cea5e0b2 |
| SFT+ppo | 54 | Not recorded | 68fa73b51e209131cea5e0b2 |
| RLHF+pt | 53.52 | Not recorded | 68fa73b51e209131cea5e0b2 |
| SFT+ppo | 45.32 | Not recorded | 68fa73b51e209131cea5e0b2 |

### HBAT

PandaLM: 77.79. Result ID: 69931a113a6c31484913b8d5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Implementation](<https://github.com/wangclnlp/DeepSpeed-Chat-Extension/tree/main/examples/hybrid_alignment_training>)

Recorded evaluation configuration: {"#Param":"13B","PPO":false,"DPO":true}. This record does not assert independent reproduction.

### HBAT-Freeze

PandaLM: 76.87. Result ID: 69931a113a6c31484913b8d4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Implementation](<https://github.com/wangclnlp/DeepSpeed-Chat-Extension/tree/main/examples/hybrid_alignment_training>)

Recorded evaluation configuration: {"#Param":"13B","PPO":false,"DPO":true}. This record does not assert independent reproduction.

### DPO

PandaLM: 75. Result ID: 69931a113a6c31484913b8d3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74511e209131cea6105d.md>)

Recorded evaluation configuration: {"#Param":"13B","PPO":false,"DPO":true}. This record does not assert independent reproduction.

### HBAT

PandaLM: 74.78. Result ID: 69931a113a6c31484913b8cc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Implementation](<https://github.com/wangclnlp/DeepSpeed-Chat-Extension/tree/main/examples/hybrid_alignment_training>)

Recorded evaluation configuration: {"#Param":"7B","PPO":false,"DPO":true}. This record does not assert independent reproduction.

### HBAT

PandaLM: 72.83. Result ID: 69931a113a6c31484913b8d2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Implementation](<https://github.com/wangclnlp/DeepSpeed-Chat-Extension/tree/main/examples/hybrid_alignment_training>)

Recorded evaluation configuration: {"#Param":"13B","PPO":true,"DPO":false}. This record does not assert independent reproduction.

### RLHF

PandaLM: 72.21. Result ID: 69931a113a6c31484913b8ce.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

Recorded evaluation configuration: {"#Param":"13B","PPO":true,"DPO":false}. This record does not assert independent reproduction.

### HBAT-Freeze

PandaLM: 71.31. Result ID: 69931a113a6c31484913b8d1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Implementation](<https://github.com/wangclnlp/DeepSpeed-Chat-Extension/tree/main/examples/hybrid_alignment_training>)

Recorded evaluation configuration: {"#Param":"13B","PPO":true,"DPO":false}. This record does not assert independent reproduction.

### HBAT

PandaLM: 70.88. Result ID: 69931a113a6c31484913b8c9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Implementation](<https://github.com/wangclnlp/DeepSpeed-Chat-Extension/tree/main/examples/hybrid_alignment_training>)

Recorded evaluation configuration: {"#Param":"7B","PPO":true,"DPO":false}. This record does not assert independent reproduction.

### DPO

PandaLM: 70.74. Result ID: 69931a113a6c31484913b8ca.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74511e209131cea6105d.md>)

Recorded evaluation configuration: {"#Param":"7B","PPO":false,"DPO":true}. This record does not assert independent reproduction.

### HBAT-Freeze

PandaLM: 69.91. Result ID: 69931a113a6c31484913b8c8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Implementation](<https://github.com/wangclnlp/DeepSpeed-Chat-Extension/tree/main/examples/hybrid_alignment_training>)

Recorded evaluation configuration: {"#Param":"7B","PPO":true,"DPO":false}. This record does not assert independent reproduction.

### RLHF

PandaLM: 69.79. Result ID: 69931a113a6c31484913b8c5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

Recorded evaluation configuration: {"#Param":"7B","PPO":true,"DPO":false}. This record does not assert independent reproduction.

### HBAT-Freeze

PandaLM: 68.78. Result ID: 69931a113a6c31484913b8cb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Implementation](<https://github.com/wangclnlp/DeepSpeed-Chat-Extension/tree/main/examples/hybrid_alignment_training>)

Recorded evaluation configuration: {"#Param":"7B","PPO":false,"DPO":true}. This record does not assert independent reproduction.

### RLHF+pt

PandaLM: 63.67. Result ID: 69931a113a6c31484913b8cf.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bf71e209131cea314c0.md>)

[Implementation](<https://github.com/openai/following-instructions-human-feedback>)

Recorded evaluation configuration: {"#Param":"13B","PPO":true,"DPO":false}. This record does not assert independent reproduction.

### SFT

PandaLM: 62.16. Result ID: 69931a113a6c31484913b8cd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

Recorded evaluation configuration: {"#Param":"13B","PPO":false,"DPO":false}. This record does not assert independent reproduction.

### SFT

PandaLM: 54.76. Result ID: 69931a113a6c31484913b8c4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

Recorded evaluation configuration: {"#Param":"7B","PPO":false,"DPO":false}. This record does not assert independent reproduction.

### SFT+ppo

PandaLM: 54. Result ID: 69931a113a6c31484913b8d0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79a11e209131cea7e87b.md>)

Recorded evaluation configuration: {"#Param":"13B","PPO":true,"DPO":false}. This record does not assert independent reproduction.

### RLHF+pt

PandaLM: 53.52. Result ID: 69931a113a6c31484913b8c6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bf71e209131cea314c0.md>)

[Implementation](<https://github.com/openai/following-instructions-human-feedback>)

Recorded evaluation configuration: {"#Param":"7B","PPO":true,"DPO":false}. This record does not assert independent reproduction.

### SFT+ppo

PandaLM: 45.32. Result ID: 69931a113a6c31484913b8c7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73b51e209131cea5e0b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79a11e209131cea7e87b.md>)

Recorded evaluation configuration: {"#Param":"7B","PPO":true,"DPO":false}. This record does not assert independent reproduction.

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