List your company
ResearchFormerly WizWand

Visual Grounding on Out-of-Distribution OOD Avg (test)

78.4Accuracy (OOD Avg)VRRL (Ours)
1.54421.49741.4561.403Jul 2, 2026
Updated 4mo ago

Evaluation results

MethodLinks
VRRL (Ours)Evaluation Setting=RL, Model Backbone=Qwen2.5-VL-7B-Instruct2026.0778.4
Multi-SFT -> GRPOEvaluation Setting=RL, Model Backbone=Qwen2.5-VL-7B-Instruct2026.0773.2
Single-SFT -> GRPOEvaluation Setting=RL, Model Backbone=Qwen2.5-VL-7B-Instruct2026.0770.2
Reflection TuningEvaluation Setting=SFT, Model Backbone=Qwen2.5-VL-7B-Instruct2026.0755.6
Multi-SFTEvaluation Setting=SFT, Model Backbone=Qwen2.5-VL-7B-Instruct2026.0749.9
Single-SFTEvaluation Setting=SFT, Model Backbone=Qwen2.5-VL-7B-Instruct2026.0746.5
VRRL (Ours)Evaluation Setting=RL, Model Backbone=Qwen2.5-VL-3B-Instruct2026.0745.7
Multi-SFT -> GRPOEvaluation Setting=RL, Model Backbone=Qwen2.5-VL-3B-Instruct2026.0740
Single-SFT -> GRPOEvaluation Setting=RL, Model Backbone=Qwen2.5-VL-3B-Instruct2026.0730.1
Reflection TuningEvaluation Setting=SFT, Model Backbone=Qwen2.5-VL-3B-Instruct2026.0728
Single-SFTEvaluation Setting=SFT, Model Backbone=Qwen2.5-VL-3B-Instruct2026.0724.5
Multi-SFTEvaluation Setting=SFT, Model Backbone=Qwen2.5-VL-3B-Instruct2026.0722.5
VL-Rethinker-32BEvaluation Setting=Zero-shot, Backbone=32B2026.0722.1
Qwen2.5-VL-7BSingleEvaluation Setting=Zero-shot, Backbone=7B, Strategy=Single-turn2026.078.8
Qwen2.5-VL-7B*MultiEvaluation Setting=Zero-shot, Backbone=7B, Strategy=Multi-turn reflection2026.078.8
Qwen2.5-VL-3BSingleEvaluation Setting=Zero-shot, Backbone=3B, Strategy=Single-turn2026.076
VL-Rethinker-7BEvaluation Setting=Zero-shot, Backbone=7B2026.074.6
Qwen2.5-VL-3B*MultiEvaluation Setting=Zero-shot, Backbone=3B, Strategy=Multi-turn reflection2026.074.5