List your company
ResearchFormerly WizWand

Visual Grounding on In-Distribution (test)

99.7AccuracyVRRL (Ours)
1.52427.01252.577.988Jul 2, 2026
Updated 4mo ago

Evaluation results

MethodLinks
VRRL (Ours)Evaluation Setting=RL, Model Backbone=Qwen2.5-VL-7B-Instruct2026.0799.7—
Multi-SFT -> GRPOEvaluation Setting=RL, Model Backbone=Qwen2.5-VL-3B-Instruct2026.0799.6—
VRRL (Ours)Evaluation Setting=RL, Model Backbone=Qwen2.5-VL-3B-Instruct2026.0799.6—
Single-SFT -> GRPOEvaluation Setting=RL, Model Backbone=Qwen2.5-VL-7B-Instruct2026.0799.6—
Multi-SFT -> GRPOEvaluation Setting=RL, Model Backbone=Qwen2.5-VL-7B-Instruct2026.0799.6—
Single-SFT -> GRPOEvaluation Setting=RL, Model Backbone=Qwen2.5-VL-3B-Instruct2026.0796.2—
Reflection TuningEvaluation Setting=SFT, Model Backbone=Qwen2.5-VL-7B-Instruct2026.0795.3—
Reflection TuningEvaluation Setting=SFT, Model Backbone=Qwen2.5-VL-3B-Instruct2026.0792.7—
Multi-SFTEvaluation Setting=SFT, Model Backbone=Qwen2.5-VL-7B-Instruct2026.0784.8—
Multi-SFTEvaluation Setting=SFT, Model Backbone=Qwen2.5-VL-3B-Instruct2026.0784.7—
Single-SFTEvaluation Setting=SFT, Model Backbone=Qwen2.5-VL-7B-Instruct2026.0783.6—
Single-SFTEvaluation Setting=SFT, Model Backbone=Qwen2.5-VL-3B-Instruct2026.0780.4—
VL-Rethinker-32BEvaluation Setting=Zero-shot, Backbone=32B2026.0742.4—
Qwen2.5-VL-7B*MultiEvaluation Setting=Zero-shot, Backbone=7B, Strategy=Multi-turn reflection2026.0719.858
Qwen2.5-VL-7BSingleEvaluation Setting=Zero-shot, Backbone=7B, Strategy=Single-turn2026.0717.9—
VL-Rethinker-7BEvaluation Setting=Zero-shot, Backbone=7B2026.0715.1—
Qwen2.5-VL-3B*MultiEvaluation Setting=Zero-shot, Backbone=3B, Strategy=Multi-turn reflection2026.075.677.3
Qwen2.5-VL-3BSingleEvaluation Setting=Zero-shot, Backbone=3B, Strategy=Single-turn2026.075.3—