List your company
ResearchFormerly WizWand

Multimodal Conversation on PCogAlignBench LS1

0.903LLM Judge ScoreDAPO (Latent Action)
0.6690.729750.79050.85125Jan 12, 2026
Updated 8mo ago

Evaluation results

MethodLinks
DAPO (Latent Action)Backbone=Qwen2.5-VL-7B-Instruct, Training Method=DAPO, Action Space=Latent Action2026.010.903
BNPO (Latent Action)Backbone=Qwen2.5-VL-7B-Instruct, Training Method=BNPO, Action Space=Latent Action2026.010.901
GRPO (Latent Action)Backbone=Qwen2.5-VL-7B-Instruct, Training Method=GRPO, Action Space=Latent Action2026.010.898
Dr.GRPO (Latent Action)Backbone=Qwen2.5-VL-7B-Instruct, Training Method=Dr.GRPO, Action Space=Latent Action2026.010.897
DAPO (Latent Action)Backbone=Qwen2.5-VL-3B-Instruct, Training Method=DAPO, Action Space=Latent Action2026.010.879
Dr.GRPO (Latent Action)Backbone=Qwen2.5-VL-3B-Instruct, Training Method=Dr.GRPO, Action Space=Latent Action2026.010.874
BNPO (Latent Action)Backbone=Qwen2.5-VL-3B-Instruct, Training Method=BNPO, Action Space=Latent Action2026.010.872
GRPO (Latent Action)Backbone=Qwen2.5-VL-3B-Instruct, Training Method=GRPO, Action Space=Latent Action2026.010.871
GRPO (Token)Backbone=Qwen2.5-VL-7B-Instruct, Training Method=GRPO, Action Space=Token2026.010.87
Dr.GRPO (Token)Backbone=Qwen2.5-VL-7B-Instruct, Training Method=Dr.GRPO, Action Space=Token2026.010.854
BNPO (Token)Backbone=Qwen2.5-VL-7B-Instruct, Training Method=BNPO, Action Space=Token2026.010.85
BNPO (Token)Backbone=Qwen2.5-VL-3B-Instruct, Training Method=BNPO, Action Space=Token2026.010.849
GRPO (Token)Backbone=Qwen2.5-VL-3B-Instruct, Training Method=GRPO, Action Space=Token2026.010.845
DAPO (Token)Backbone=Qwen2.5-VL-7B-Instruct, Training Method=DAPO, Action Space=Token2026.010.844
Dr.GRPO (Token)Backbone=Qwen2.5-VL-3B-Instruct, Training Method=Dr.GRPO, Action Space=Token2026.010.835
DAPO (Token)Backbone=Qwen2.5-VL-3B-Instruct, Training Method=DAPO, Action Space=Token2026.010.835
SFTBackbone=Qwen2.5-VL-3B-Instruct, Training Method=SFT, Action Space=Token2026.010.808
SFTBackbone=Qwen2.5-VL-7B-Instruct, Training Method=SFT, Action Space=Token2026.010.808
PromptBackbone=Qwen2.5-VL-7B-Instruct, Training Method=Prompt, Action Space=Token2026.010.721
PromptBackbone=Qwen2.5-VL-3B-Instruct, Training Method=Prompt, Action Space=Token2026.010.678