Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

VLABench

Benchmarks

Task NameDataset NameSOTA ResultTrend
Language-conditioned Manipulation PlanningVLABench
Success Rate64.83
35
High-level planningVLABench vlm_evaluation 1.0 (test)
AUC85.8
25
Success Rate EvaluationVLABench
Average Success Rate46.3
19
Failure DetectionVLABench (Unseen Tasks)
bACC71.3
12
Failure DetectionVLABench (Seen Tasks)
Balanced Accuracy (bACC)85.6
12
Embodied Task ExecutionVLABench Avg.
Progress Score (PS)65.9
10
Embodied Task ExecutionVLABench Instruction
Progress Score (PS)70.2
10
Embodied Task ExecutionVLABench Commonsense
Progress Score (PS)57.3
10
Embodied Task ExecutionVLABench Cross Category
Primitive Success (PS)61
10
Embodied Task ExecutionVLABench In-dist.
Progress Score (PS)81.1
10
Embodied Task ExecutionVLABench Texture
Progress Score (PS)62.3
9
Robotic Manipulation ReasoningVLABench
In Distribution Accuracy78
9
Robotic ManipulationVLABench seen object configurations π0.5
Success Rate38.8
8
Robot Action PredictionVLABench T2
Success Rate39.9
8
Robot Action PredictionVLABench T1
Success Rate (SR)64.8
8
Robotic ManipulationVLABench GR00T-N1.6 (seen object configurations)
Success Rate35.7
5
Robotic ManipulationVLABench Average (test)
IS71.2
5
Robotic ManipulationVLABench Unseen Texture (test)
IS76
5
Robotic ManipulationVLABench Semantic Instruction (test)
IS Score82
5
Robotic ManipulationVLABench Common Sense (test)
IS74.2
5
Robotic ManipulationVLABench Cross-Category (test)
IS54.1
5
Robotic ManipulationVLABench Real-world (Average)
Success Rate (SR)55
5
Robotic ManipulationVLABench Real-world (Long Horizon Tier)
Success Rate (SR)38
5
Robotic ManipulationVLABench Real-world (Semantic Tier)
Success Rate (SR)42
5
Robotic ManipulationVLABench Real-world (Basic Tier)
Success Rate (SR)97
5
Showing 25 of 42 rows