List your company

Compound LLM Collaboration System on Human-preferred responses dataset

19.8Win Rate (Chosen)SysDPO-Sampling
12.5214.4116.318.19Feb 24, 2025

Evaluation results

MethodLinks
SysDPO-SamplingBackbone=Qwen1.5-1.8B-Chat2025.0219.866.4
SysDPO-ψ2Stage=ψ2 training only2025.0218.163.9
Separate-DPOBackbone=Qwen1.5-1.8B-Chat2025.0216.657.3
SysDPO-ψ1Stage=ψ1 training only2025.021660.4
Prompted SystemBackbone=Qwen1.5-1.8B-Chat2025.0212.8—