List your company

Multi-turn Dialogue on MT-Bench (GPT-4 Score)

8.9GPT-4 ScoreUAPO
7.1327.5918.058.509Sep 3, 2025

Evaluation results

MethodLinks
UAPOBackbone=Gemma-2-9B-Instruct2025.098.9
SimUAPOBackbone=Gemma-2-9B-Instruct2025.098.9
SimPOBackbone=Gemma-2-9B-Instruct2025.098.8
IPOBackbone=Gemma-2-9B-Instruct2025.098.7
CPOBackbone=Gemma-2-9B-Instruct2025.098.7
DPOBackbone=Gemma-2-9B-Instruct2025.098.7
SFTBackbone=Gemma-2-9B-Instruct2025.098.6
KTOBackbone=Gemma-2-9B-Instruct2025.098.5
R-DPOBackbone=Gemma-2-9B-Instruct2025.098.5
ORPOBackbone=Llama-3-8B-Instruct2025.098.3
SimUAPOBackbone=Llama-3-8B-Instruct2025.098.3
ORPOBackbone=Gemma-2-9B-Instruct2025.098.3
IPOBackbone=Llama-3-8B-Instruct2025.098.2
CPOBackbone=Llama-3-8B-Instruct2025.098.2
KTOBackbone=Llama-3-8B-Instruct2025.098.2
R-DPOBackbone=Llama-3-8B-Instruct2025.098.2
DPOBackbone=Llama-3-8B-Instruct, Source=Meng et al. (2024)2025.098.2
DPOBackbone=Llama-3-8B-Instruct2025.098.2
UAPOBackbone=Llama-3-8B-Instruct2025.098.2
SFTBackbone=Llama-3-8B-Instruct2025.098.1
SimPOBackbone=Llama-3-8B-Instruct, Source=Meng et al. (2024)2025.098
IPOBackbone=Mistral-7B-Instruct2025.097.8
KTOBackbone=Mistral-7B-Instruct2025.097.7
ORPOBackbone=Mistral-7B-Instruct2025.097.7
UAPOBackbone=Mistral-7B-Instruct2025.097.7
DPOBackbone=Mistral-7B-Instruct, Source=Meng et al. (2024)2025.097.6
DPOBackbone=Mistral-7B-Instruct2025.097.6
SimPOBackbone=Mistral-7B-Instruct, Source=Meng et al. (2024)2025.097.6
SimUAPOBackbone=Mistral-7B-Instruct2025.097.6
SFTBackbone=Mistral-7B-Instruct2025.097.5
CPOBackbone=Mistral-7B-Instruct2025.097.5
R-DPOBackbone=Mistral-7B-Instruct2025.097.5
SimPOBackbone=Llama-3-8B-Instruct2025.097.4
SimPOBackbone=Mistral-7B-Instruct2025.097.2