# Multi-turn conversation on MT-Bench-101

Source: [Multi-turn conversation on MT-Bench-101](<https://www.sota2.com/research/sota/multi-turn-conversation-on-mt-bench-101>)

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

Data updated: 2026-02-26T05:46:04.398Z

Evaluation of multi-turn conversational ability on the MT-Bench-101 dataset, reporting detailed component scores (e.g., Grounding, Instruction Following, Helpfulness) and an overall average.

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 |
| Dataset (including any stated split) | MT-Bench-101 |
| Metric | Grounding Score |
| Unit | SCALAR |
| 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

- [Grounding Score](<https://www.sota2.com/research/sota/multi-turn-conversation-on-mt-bench-101.md?metric=Grounding%20Score>): Higher is better; unit: SCALAR.

- [Instruction Following Score](<https://www.sota2.com/research/sota/multi-turn-conversation-on-mt-bench-101.md?metric=Instruction%20Following%20Score>): Higher is better; unit: SCALAR.

- [Answer Relevance Score](<https://www.sota2.com/research/sota/multi-turn-conversation-on-mt-bench-101.md?metric=Answer%20Relevance%20Score>): Higher is better; unit: SCALAR.

- [Factuality Score](<https://www.sota2.com/research/sota/multi-turn-conversation-on-mt-bench-101.md?metric=Factuality%20Score>): Higher is better; unit: SCALAR.

- [Meaningfulness Score](<https://www.sota2.com/research/sota/multi-turn-conversation-on-mt-bench-101.md?metric=Meaningfulness%20Score>): Higher is better; unit: SCALAR.

- [Coherence Score](<https://www.sota2.com/research/sota/multi-turn-conversation-on-mt-bench-101.md?metric=Coherence%20Score>): Higher is better; unit: SCALAR.

- [Tone Score](<https://www.sota2.com/research/sota/multi-turn-conversation-on-mt-bench-101.md?metric=Tone%20Score>): Higher is better; unit: SCALAR.

- [Completeness Score](<https://www.sota2.com/research/sota/multi-turn-conversation-on-mt-bench-101.md?metric=Completeness%20Score>): Higher is better; unit: SCALAR.

- [Safety Score](<https://www.sota2.com/research/sota/multi-turn-conversation-on-mt-bench-101.md?metric=Safety%20Score>): Higher is better; unit: SCALAR.

- [Specificity Score](<https://www.sota2.com/research/sota/multi-turn-conversation-on-mt-bench-101.md?metric=Specificity%20Score>): Higher is better; unit: SCALAR.

- [Cohesiveness Score](<https://www.sota2.com/research/sota/multi-turn-conversation-on-mt-bench-101.md?metric=Cohesiveness%20Score>): Higher is better; unit: SCALAR.

- [Prompt Interpretation Score](<https://www.sota2.com/research/sota/multi-turn-conversation-on-mt-bench-101.md?metric=Prompt%20Interpretation%20Score>): Higher is better; unit: SCALAR.

- [Style Consistency Score](<https://www.sota2.com/research/sota/multi-turn-conversation-on-mt-bench-101.md?metric=Style%20Consistency%20Score>): Higher is better; unit: SCALAR.

- [Average Score](<https://www.sota2.com/research/sota/multi-turn-conversation-on-mt-bench-101.md?metric=Average%20Score>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Grounding Score | Result date | Report paper |
| --- | --- | --- | --- |
| TCA-Attention | 8.5 | Not recorded | 693a50246cfd488761a54ba7 |
| Qwen2.5-7B-Instruct | 8.2 | Not recorded | 693a50246cfd488761a54ba7 |
| FlexPrefill | 7.81 | Not recorded | 693a50246cfd488761a54ba7 |

### TCA-Attention

Grounding Score: 8.5. Result ID: 6996c5ae5698665ec4112a1a.

[Reporting paper](<https://www.sota2.com/research/paper/693a50246cfd488761a54ba7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/693a50246cfd488761a54ba7.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2.5-7B-Instruct"}. This record does not assert independent reproduction.

### Qwen2.5-7B-Instruct

Grounding Score: 8.2. Result ID: 6996c5ae5698665ec4112a18.

[Reporting paper](<https://www.sota2.com/research/paper/693a50246cfd488761a54ba7.md>)

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

Recorded evaluation configuration: {"Backbone":"Qwen2.5-7B-Instruct","Attention":"Full-Attention"}. This record does not assert independent reproduction.

### FlexPrefill

Grounding Score: 7.81. Result ID: 6996c5ae5698665ec4112a19.

[Reporting paper](<https://www.sota2.com/research/paper/693a50246cfd488761a54ba7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77281e209131cea7173b.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2.5-7B-Instruct"}. This record does not assert independent reproduction.

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