# Conversational Question Answering on CHATRAG BENCH

Source: [Conversational Question Answering on CHATRAG BENCH](<https://www.sota2.com/research/sota/conversational-question-answering-on-chatrag-bench>)

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

Data updated: 2026-02-26T05:35:49.196Z

Evaluation of model performance on the CHATRAG BENCH dataset for Conversational Question Answering, aggregating results across constituent evaluations like Doc2Dial, QuAC, CoQA, and others.

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 | Conversational Question Answering |
| Dataset (including any stated split) | CHATRAG BENCH |
| Metric | Average Score |
| 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

- [Average Score](<https://www.sota2.com/research/sota/conversational-question-answering-on-chatrag-bench.md?metric=Average%20Score>): Higher is better; unit: PERCENT.

- [Doc2Dial Score](<https://www.sota2.com/research/sota/conversational-question-answering-on-chatrag-bench.md?metric=Doc2Dial%20Score>): Higher is better; unit: PERCENT.

- [QuAC Score](<https://www.sota2.com/research/sota/conversational-question-answering-on-chatrag-bench.md?metric=QuAC%20Score>): Higher is better; unit: PERCENT.

- [QReCC Score](<https://www.sota2.com/research/sota/conversational-question-answering-on-chatrag-bench.md?metric=QReCC%20Score>): Higher is better; unit: PERCENT.

- [CoQA Score](<https://www.sota2.com/research/sota/conversational-question-answering-on-chatrag-bench.md?metric=CoQA%20Score>): Higher is better; unit: PERCENT.

- [DoQA Score](<https://www.sota2.com/research/sota/conversational-question-answering-on-chatrag-bench.md?metric=DoQA%20Score>): Higher is better; unit: PERCENT.

- [ConvFinQA Score](<https://www.sota2.com/research/sota/conversational-question-answering-on-chatrag-bench.md?metric=ConvFinQA%20Score>): Higher is better; unit: PERCENT.

- [SQA Score](<https://www.sota2.com/research/sota/conversational-question-answering-on-chatrag-bench.md?metric=SQA%20Score>): Higher is better; unit: PERCENT.

- [TopiOCQA Score](<https://www.sota2.com/research/sota/conversational-question-answering-on-chatrag-bench.md?metric=TopiOCQA%20Score>): Higher is better; unit: PERCENT.

- [HybriDial Score](<https://www.sota2.com/research/sota/conversational-question-answering-on-chatrag-bench.md?metric=HybriDial%20Score>): Higher is better; unit: PERCENT.

- [INSCIT Score](<https://www.sota2.com/research/sota/conversational-question-answering-on-chatrag-bench.md?metric=INSCIT%20Score>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Average Score | Result date | Report paper |
| --- | --- | --- | --- |
| ChatQA-1.0-70B vs GPT-4-0613 | 69.09 | Not recorded | 68fa757c1e209131cea6847a |
| GPT-4-0613 | 17.1 | Not recorded | 68fa757c1e209131cea6847a |
| ChatQA-1.0-70B | 13.81 | Not recorded | 68fa757c1e209131cea6847a |

### ChatQA-1.0-70B vs GPT-4-0613

Average Score: 69.09. Result ID: 6993cbc63a6c314849189b13.

[Reporting paper](<https://www.sota2.com/research/paper/68fa757c1e209131cea6847a.md>)

Recorded evaluation configuration: {"Evaluation Protocol":"Human A/B Testing","Outcome":"Tie Rate"}. This record does not assert independent reproduction.

### GPT-4-0613

Average Score: 17.1. Result ID: 6993cbc63a6c314849189b12.

[Reporting paper](<https://www.sota2.com/research/paper/68fa757c1e209131cea6847a.md>)

Recorded evaluation configuration: {"Evaluation Protocol":"Human A/B Testing","Outcome":"Win Rate vs ChatQA-1.0-70B"}. This record does not assert independent reproduction.

### ChatQA-1.0-70B

Average Score: 13.81. Result ID: 6993cbc63a6c314849189b11.

[Reporting paper](<https://www.sota2.com/research/paper/68fa757c1e209131cea6847a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa757c1e209131cea6847a.md>)

[Implementation](<https://chatqa-project.github.io/>)

Recorded evaluation configuration: {"Evaluation Protocol":"Human A/B Testing","Outcome":"Win Rate vs GPT-4-0613"}. This record does not assert independent reproduction.

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