List your company
ResearchFormerly WizWand

Social Dialogue on SOTOPIA Interaction with GPT-4o

8.21Goal ScoreLLaMA-8B+BC+ALPO
7.14927.42467.77.9754Jan 3, 2025Mar 9, 2025May 14, 2025Jul 19, 2025Sep 22, 2025Nov 27, 2025Feb 1, 2026
Updated 8mo ago

Evaluation results

MethodLinks
LLaMA-8B+BC+ALPOBackbone=LLaMA-8B, Alignment=ALPO, Interaction Partner=GPT-4o2026.028.213.19
SDPOBase Model=Mistral-v0.32025.018.143.06
Llama-8B+BC+SDPOAlignment=BC + Segment-Level Direct Preference Optimization2025.018.133.16
LLaMA-8B+BC+SDPOBackbone=LLaMA-8B, Alignment=SDPO, Interaction Partner=GPT-4o2026.028.133.16
Llama-8B+BC+ETOAlignment=BC + ETO2025.018.023.03
LLaMA-8B+BC+ETOBackbone=LLaMA-8B, Alignment=ETO, Interaction Partner=GPT-4o2026.028.023.03
Llama-8B+BC+DMPOAlignment=BC + DMPO2025.0182.98
LLaMA-8B+BC+DMPOBackbone=LLaMA-8B, Alignment=DMPO, Interaction Partner=GPT-4o2026.0282.98
DMPOBase Model=Mistral-v0.32025.017.972.94
ETOBase Model=Mistral-v0.32025.017.942.94
GPT-4-turbo2025.017.922.79
GPT-4-turboInteraction Partner=GPT-4o2026.027.922.79
GPT-4o2025.017.92.67
GPT-4oInteraction Partner=GPT-4o2026.027.92.67
DPOBase Model=Mistral-v0.32025.017.832.86
Llama-8B+BC+DPOAlignment=BC + DPO2025.017.82.97
LLaMA-8B+BC+DPOBackbone=LLaMA-8B, Alignment=DPO, Interaction Partner=GPT-4o2026.027.82.97
Llama-8B2025.017.72.49
LLaMA-8BBackbone=LLaMA-8B, Interaction Partner=GPT-4o2026.027.72.49
Llama-8B+BC+Preferred-SFTAlignment=BC + Preferred-SFT2025.017.652.88
LLaMA-8B+BC+Preferred-SFTBackbone=LLaMA-8B, Training=BC + Preferred-SFT, Interaction Partner=GPT-4o2026.027.652.88
BCBase Model=Mistral-v0.32025.017.62.81
Llama-8B+BCAlignment=Behavioral Cloning2025.017.532.78
LLaMA-8B+BCBackbone=LLaMA-8B, Training=Behavioral Cloning, Interaction Partner=GPT-4o2026.027.532.78
GPT-4o-mini2025.017.442.36
GPT-4o-miniInteraction Partner=GPT-4o2026.027.442.36
GPT-3.5-turbo2025.017.192.05
GPT-3.5-turboInteraction Partner=GPT-4o2026.027.192.05