# Dialogue Comprehension on DREAM

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

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

Data updated: 2026-02-26T05:28:41.622Z

Evaluation of model performance on dialogue comprehension tasks using the DREAM dataset, measured by Accuracy.

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 Comprehension |
| Dataset (including any stated split) | DREAM |
| Metric | Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 15 of 15 |
| 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

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

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| RoBERTa^ASA | 69.2 | Not recorded | 68fa6e7b1e209131cea40278 |
| RoBERTa | 67 | Not recorded | 68fa6e7b1e209131cea40278 |
| BERT^ASA | 64.3 | Not recorded | 68fa6e7b1e209131cea40278 |
| BERT†^ASA | 63.3 | Not recorded | 68fa6e7b1e209131cea40278 |
| BERT | 62.9 | Not recorded | 68fa6e7b1e209131cea40278 |
| BART-FT | 62.56 | Not recorded | 68fa6c141e209131cea31ff0 |
| BERT† | 61.2 | Not recorded | 68fa6e7b1e209131cea40278 |
| Narrator | 53.41 | Not recorded | 68fa6c141e209131cea31ff0 |
| BART-CRD3-GEN | 52.96 | Not recorded | 68fa6c141e209131cea31ff0 |
| BART-CNN-GEN | 52.72 | Not recorded | 68fa6c141e209131cea31ff0 |
| BART-CNN-CLS | 50.46 | Not recorded | 68fa6c141e209131cea31ff0 |
| BART-DIAL-DE | 46.69 | Not recorded | 68fa6c141e209131cea31ff0 |
| ROBERTa | 45.22 | Not recorded | 68fa6c141e209131cea31ff0 |
| BART | 45.07 | Not recorded | 68fa6c141e209131cea31ff0 |
| GPT-2 | 41.99 | Not recorded | 68fa6c141e209131cea31ff0 |

### RoBERTa^ASA

Accuracy: 69.2. Result ID: 6990248d3a6c314849fe6979.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6e7b1e209131cea40278.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e7b1e209131cea40278.md>)

[Implementation](<https://github.com/gingasan/adversarialSA>)

Recorded evaluation configuration: {"Backbone architecture":"RoBERTa","ASA":"true"}. This record does not assert independent reproduction.

### RoBERTa

Accuracy: 67. Result ID: 6990248d3a6c314849fe6978.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6e7b1e209131cea40278.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66131e209131cea0e18b.md>)

[Implementation](<https://github.com/pytorch/fairseq>)

Recorded evaluation configuration: {"Backbone architecture":"RoBERTa"}. This record does not assert independent reproduction.

### BERT^ASA

Accuracy: 64.3. Result ID: 6990248d3a6c314849fe6975.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6e7b1e209131cea40278.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e7b1e209131cea40278.md>)

[Implementation](<https://github.com/gingasan/adversarialSA>)

Recorded evaluation configuration: {"Backbone architecture":"BERT","ASA":"true"}. This record does not assert independent reproduction.

### BERT†^ASA

Accuracy: 63.3. Result ID: 6990248d3a6c314849fe6977.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6e7b1e209131cea40278.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e7b1e209131cea40278.md>)

[Implementation](<https://github.com/gingasan/adversarialSA>)

Recorded evaluation configuration: {"Backbone architecture":"BERT","Pre-training strategy":"longer-trained with MLM","ASA":"true"}. This record does not assert independent reproduction.

### BERT

Accuracy: 62.9. Result ID: 6990248d3a6c314849fe6974.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6e7b1e209131cea40278.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65c31e209131cea0c4ed.md>)

[Implementation](<https://github.com/google-research/bert>)

Recorded evaluation configuration: {"Backbone architecture":"BERT"}. This record does not assert independent reproduction.

### BART-FT

Accuracy: 62.56. Result ID: 698eca06ffc3901de0a13242.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6c141e209131cea31ff0.md>)

Recorded evaluation configuration: {"training\_mode":"supervised finetuning"}. This record does not assert independent reproduction.

### BERT†

Accuracy: 61.2. Result ID: 6990248d3a6c314849fe6976.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6e7b1e209131cea40278.md>)

Recorded evaluation configuration: {"Backbone architecture":"BERT","Pre-training strategy":"longer-trained with MLM"}. This record does not assert independent reproduction.

### Narrator

Accuracy: 53.41. Result ID: 698eca06ffc3901de0a1324a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6c141e209131cea31ff0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c141e209131cea31ff0.md>)

[Implementation](<https://github.com/zhaochaocs/Diana>)

Recorded evaluation configuration: {"pre-training\_data":"DIANA","pre-training\_task":"GEN","training\_mode":"zero-shot"}. This record does not assert independent reproduction.

### BART-CRD3-GEN

Accuracy: 52.96. Result ID: 698eca06ffc3901de0a13249.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6c141e209131cea31ff0.md>)

Recorded evaluation configuration: {"pre-training\_data":"CRD3","pre-training\_task":"GEN","training\_mode":"zero-shot"}. This record does not assert independent reproduction.

### BART-CNN-GEN

Accuracy: 52.72. Result ID: 698eca06ffc3901de0a13248.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6c141e209131cea31ff0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa634a1e209131ce9fdbfb.md>)

[Implementation](<https://www.sota2.com/www.github.com/abisee/pointer-generator>)

Recorded evaluation configuration: {"pre-training\_data":"CNN","pre-training\_task":"GEN","training\_mode":"zero-shot"}. This record does not assert independent reproduction.

### BART-CNN-CLS

Accuracy: 50.46. Result ID: 698eca06ffc3901de0a13247.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6c141e209131cea31ff0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa634a1e209131ce9fdbfb.md>)

[Implementation](<https://www.sota2.com/www.github.com/abisee/pointer-generator>)

Recorded evaluation configuration: {"pre-training\_data":"CNN","pre-training\_task":"CLS","training\_mode":"zero-shot"}. This record does not assert independent reproduction.

### BART-DIAL-DE

Accuracy: 46.69. Result ID: 698eca06ffc3901de0a13246.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6c141e209131cea31ff0.md>)

Recorded evaluation configuration: {"pre-training\_data":"DIAL","pre-training\_task":"DE","training\_mode":"zero-shot"}. This record does not assert independent reproduction.

### ROBERTa

Accuracy: 45.22. Result ID: 698eca06ffc3901de0a13244.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6c141e209131cea31ff0.md>)

Recorded evaluation configuration: {"training\_mode":"zero-shot"}. This record does not assert independent reproduction.

### BART

Accuracy: 45.07. Result ID: 698eca06ffc3901de0a13245.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6c141e209131cea31ff0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66921e209131cea10fff.md>)

Recorded evaluation configuration: {"training\_mode":"zero-shot"}. This record does not assert independent reproduction.

### GPT-2

Accuracy: 41.99. Result ID: 698eca06ffc3901de0a13243.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6c141e209131cea31ff0.md>)

Recorded evaluation configuration: {"training\_mode":"zero-shot"}. This record does not assert independent reproduction.

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