# Open-ended Question Answering on CrossAlpaca-Eval zh 2.0

Source: [Open-ended Question Answering on CrossAlpaca-Eval zh 2.0](<https://www.sota2.com/research/sota/open-ended-question-answering-on-crossalpaca-eval-zh-2-0>)

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

Data updated: 2026-04-15T07:12:08.677Z

Evaluation of model performance on the Chinese version 2.0 of the CrossAlpaca evaluation set for open-ended question answering, typically measured using LLM-as-a-judge scores like GPT-4o.

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 | Open-ended Question Answering |
| Dataset (including any stated split) | CrossAlpaca-Eval zh 2.0 |
| Metric | GPT-4o Score |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 4 of 4 |
| 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

- [GPT-4o Score](<https://www.sota2.com/research/sota/open-ended-question-answering-on-crossalpaca-eval-zh-2-0.md?metric=GPT-4o%20Score>): Higher is better; unit: SCALAR.

## Recorded results

| Method | GPT-4o Score | Result date | Report paper |
| --- | --- | --- | --- |
| Qwen2.5-7B-AdaMCOT | 8.53 | Not recorded | 68fa7a601e209131cea82b5f |
| Qwen2.5-7B-Instruction | 8.47 | Not recorded | 68fa7a601e209131cea82b5f |
| LLaMA3.1-8B-AdaMCOT | 8.13 | Not recorded | 68fa7a601e209131cea82b5f |
| LLaMA3.1-8B-Instruction | 7.53 | Not recorded | 68fa7a601e209131cea82b5f |

### Qwen2.5-7B-AdaMCOT

GPT-4o Score: 8.53. Result ID: 69df390f47f34ce924c78602.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a601e209131cea82b5f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a601e209131cea82b5f.md>)

Recorded evaluation configuration: {"Backbone":"Qwen 2.5-7B","Reasoning Method":"AdaMCOT"}. This record does not assert independent reproduction.

### Qwen2.5-7B-Instruction

GPT-4o Score: 8.47. Result ID: 69df390f47f34ce924c78601.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a601e209131cea82b5f.md>)

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

Recorded evaluation configuration: {"Backbone":"Qwen 2.5-7B","Reasoning Method":"Instruction-tuned"}. This record does not assert independent reproduction.

### LLaMA3.1-8B-AdaMCOT

GPT-4o Score: 8.13. Result ID: 69df390f47f34ce924c78600.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a601e209131cea82b5f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a601e209131cea82b5f.md>)

Recorded evaluation configuration: {"Backbone":"LLaMA 3.1-8B","Reasoning Method":"AdaMCOT"}. This record does not assert independent reproduction.

### LLaMA3.1-8B-Instruction

GPT-4o Score: 7.53. Result ID: 69df390f47f34ce924c785ff.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a601e209131cea82b5f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75d11e209131cea6a028.md>)

[Implementation](<https://github.com/meta-llama/llama-agentic-system>)

Recorded evaluation configuration: {"Backbone":"LLaMA 3.1-8B","Reasoning Method":"Instruction-tuned"}. This record does not assert independent reproduction.

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