List your company
ResearchFormerly WizWand

Visual Grounding on Cell Query Out-of-Distribution (test)

77.3AccuracyVRRL (Ours)
-1.7418.7839.359.82Jul 2, 2026
Updated 4mo ago

Evaluation results

MethodLinks
VRRL (Ours)Evaluation Setting=RL, Model Backbone=Qwen2.5-VL-7B-Instruct2026.0777.3—
Multi-SFT -> GRPOEvaluation Setting=RL, Model Backbone=Qwen2.5-VL-7B-Instruct2026.0768.4—
Single-SFT -> GRPOEvaluation Setting=RL, Model Backbone=Qwen2.5-VL-7B-Instruct2026.0768.3—
Reflection TuningEvaluation Setting=SFT, Model Backbone=Qwen2.5-VL-7B-Instruct2026.0751.8—
Multi-SFTEvaluation Setting=SFT, Model Backbone=Qwen2.5-VL-7B-Instruct2026.0739.1—
Single-SFTEvaluation Setting=SFT, Model Backbone=Qwen2.5-VL-7B-Instruct2026.0734—
VL-Rethinker-32BEvaluation Setting=Zero-shot, Backbone=32B2026.0727.6—
VRRL (Ours)Evaluation Setting=RL, Model Backbone=Qwen2.5-VL-3B-Instruct2026.0720.3—
Multi-SFT -> GRPOEvaluation Setting=RL, Model Backbone=Qwen2.5-VL-3B-Instruct2026.0713.5—
Qwen2.5-VL-7B*MultiEvaluation Setting=Zero-shot, Backbone=7B, Strategy=Multi-turn reflection2026.077.276.8
Reflection TuningEvaluation Setting=SFT, Model Backbone=Qwen2.5-VL-3B-Instruct2026.077—
Qwen2.5-VL-7BSingleEvaluation Setting=Zero-shot, Backbone=7B, Strategy=Single-turn2026.076.1—
VL-Rethinker-7BEvaluation Setting=Zero-shot, Backbone=7B2026.075.7—
Single-SFT -> GRPOEvaluation Setting=RL, Model Backbone=Qwen2.5-VL-3B-Instruct2026.075.3—
Qwen2.5-VL-3BSingleEvaluation Setting=Zero-shot, Backbone=3B, Strategy=Single-turn2026.073.4—
Single-SFTEvaluation Setting=SFT, Model Backbone=Qwen2.5-VL-3B-Instruct2026.072.4—
Multi-SFTEvaluation Setting=SFT, Model Backbone=Qwen2.5-VL-3B-Instruct2026.071.6—
Qwen2.5-VL-3B*MultiEvaluation Setting=Zero-shot, Backbone=3B, Strategy=Multi-turn reflection2026.071.386.1