# Compound LLM Collaboration System on Human-preferred responses dataset

Source: [Compound LLM Collaboration System on Human-preferred responses dataset](<https://www.sota2.com/research/sota/compound-llm-collaboration-system-on-human-preferred-responses-dataset>)

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

Data updated: 2026-03-09T06:39:14.760Z

Evaluation of model performance on the Compound LLM Collaboration System task using the Human-preferred responses dataset, measured by Win Rate metrics (WR-Chosen, WR-Prompted).

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 | Compound LLM Collaboration System |
| Dataset (including any stated split) | Human-preferred responses dataset |
| Metric | Win Rate (Chosen) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 5 of 5 |
| 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

- [Win Rate (Chosen)](<https://www.sota2.com/research/sota/compound-llm-collaboration-system-on-human-preferred-responses-dataset.md?metric=Win%20Rate%20(Chosen)>): Higher is better; unit: PERCENT.

- [Win Rate (Prompted)](<https://www.sota2.com/research/sota/compound-llm-collaboration-system-on-human-preferred-responses-dataset.md?metric=Win%20Rate%20(Prompted)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Win Rate (Chosen) | Result date | Report paper |
| --- | --- | --- | --- |
| SysDPO-Sampling | 19.8 | Not recorded | 68fa790f1e209131cea7be6f |
| SysDPO-ψ2 | 18.1 | Not recorded | 68fa790f1e209131cea7be6f |
| Separate-DPO | 16.6 | Not recorded | 68fa790f1e209131cea7be6f |
| SysDPO-ψ1 | 16 | Not recorded | 68fa790f1e209131cea7be6f |
| Prompted System | 12.8 | Not recorded | 68fa790f1e209131cea7be6f |

### SysDPO-Sampling

Win Rate (Chosen): 19.8. Result ID: 69ae6ae66dbb7075b955da83.

[Reporting paper](<https://www.sota2.com/research/paper/68fa790f1e209131cea7be6f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa790f1e209131cea7be6f.md>)

Recorded evaluation configuration: {"Backbone":"Qwen1.5-1.8B-Chat"}. This record does not assert independent reproduction.

### SysDPO-ψ2

Win Rate (Chosen): 18.1. Result ID: 69ae6ae66dbb7075b955da87.

[Reporting paper](<https://www.sota2.com/research/paper/68fa790f1e209131cea7be6f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa790f1e209131cea7be6f.md>)

Recorded evaluation configuration: {"Stage":"ψ2 training only"}. This record does not assert independent reproduction.

### Separate-DPO

Win Rate (Chosen): 16.6. Result ID: 69ae6ae66dbb7075b955da85.

[Reporting paper](<https://www.sota2.com/research/paper/68fa790f1e209131cea7be6f.md>)

Recorded evaluation configuration: {"Backbone":"Qwen1.5-1.8B-Chat"}. This record does not assert independent reproduction.

### SysDPO-ψ1

Win Rate (Chosen): 16. Result ID: 69ae6ae66dbb7075b955da86.

[Reporting paper](<https://www.sota2.com/research/paper/68fa790f1e209131cea7be6f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa790f1e209131cea7be6f.md>)

Recorded evaluation configuration: {"Stage":"ψ1 training only"}. This record does not assert independent reproduction.

### Prompted System

Win Rate (Chosen): 12.8. Result ID: 69ae6ae66dbb7075b955da84.

[Reporting paper](<https://www.sota2.com/research/paper/68fa790f1e209131cea7be6f.md>)

Recorded evaluation configuration: {"Backbone":"Qwen1.5-1.8B-Chat"}. This record does not assert independent reproduction.

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