List your company
ResearchFormerly WizWand

Multimodal Conversation on MMRole OOD

91.6LLM-as-a-Judge ScoreDr.GRPO (Latent Action)
67.78473.96780.1586.333Jan 12, 2026
Updated 8mo ago

Evaluation results

MethodLinks
Dr.GRPO (Latent Action)Backbone=Qwen2.5-VL-3B-Instruct, Training Method=Dr.GRPO, Action Space=Latent Action2026.0191.6
GRPO (Latent Action)Backbone=Qwen2.5-VL-3B-Instruct, Training Method=GRPO, Action Space=Latent Action2026.0191.5
BNPO (Latent Action)Backbone=Qwen2.5-VL-3B-Instruct, Training Method=BNPO, Action Space=Latent Action2026.0190.1
DAPO (Latent Action)Backbone=Qwen2.5-VL-3B-Instruct, Training Method=DAPO, Action Space=Latent Action2026.0188.9
GRPO (Latent Action)Backbone=Qwen2.5-VL-7B-Instruct, Training Method=GRPO, Action Space=Latent Action2026.0187.2
Dr.GRPO (Latent Action)Backbone=Qwen2.5-VL-7B-Instruct, Training Method=Dr.GRPO, Action Space=Latent Action2026.0186.4
DAPO (Latent Action)Backbone=Qwen2.5-VL-7B-Instruct, Training Method=DAPO, Action Space=Latent Action2026.0186.3
BNPO (Token)Backbone=Qwen2.5-VL-7B-Instruct, Training Method=BNPO, Action Space=Token2026.0185.9
SFTBackbone=Qwen2.5-VL-7B-Instruct, Training Method=SFT, Action Space=Token2026.0185.6
Dr.GRPO (Token)Backbone=Qwen2.5-VL-7B-Instruct, Training Method=Dr.GRPO, Action Space=Token2026.0185.4
DAPO (Token)Backbone=Qwen2.5-VL-7B-Instruct, Training Method=DAPO, Action Space=Token2026.0184.2
BNPO (Latent Action)Backbone=Qwen2.5-VL-7B-Instruct, Training Method=BNPO, Action Space=Latent Action2026.0184.2
GRPO (Token)Backbone=Qwen2.5-VL-7B-Instruct, Training Method=GRPO, Action Space=Token2026.0184
Dr.GRPO (Token)Backbone=Qwen2.5-VL-3B-Instruct, Training Method=Dr.GRPO, Action Space=Token2026.0182.3
PromptBackbone=Qwen2.5-VL-7B-Instruct, Training Method=Prompt, Action Space=Token2026.0182.1
SFTBackbone=Qwen2.5-VL-3B-Instruct, Training Method=SFT, Action Space=Token2026.0180.9
DAPO (Token)Backbone=Qwen2.5-VL-3B-Instruct, Training Method=DAPO, Action Space=Token2026.0180.5
BNPO (Token)Backbone=Qwen2.5-VL-3B-Instruct, Training Method=BNPO, Action Space=Token2026.0180.1
GRPO (Token)Backbone=Qwen2.5-VL-3B-Instruct, Training Method=GRPO, Action Space=Token2026.0179.6
PromptBackbone=Qwen2.5-VL-3B-Instruct, Training Method=Prompt, Action Space=Token2026.0168.7