User Study
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
User Study n=90 trials per comparison
3.3Loss Rate
3
User Study 30 participants (test)
89.9Prompt Alignment
3
User Study 60–120° angular range
31.9Fraction of accepted trades decreasing user benefit
3
User Study (120–180° alignment range)
608.64Computer Benefit Value
3
User Study 60–120° alignment range
650.93Computer Benefit Value
3
User Study 0–60° alignment range
614.86Computer Benefit Value
3
User Study 120-180° angular range
1,025Human Benefit Value
3
User Study 60-120° angular range
64.28Human Benefit Value
3
User Study angular range (0-60°)
-774.44Human Benefit Value
3
User Study 30 participants 1.0 (test)
2.9Confidence in Understanding
3
User Study
51.15Controllability
3
User study 12 participants 30 rigid objects
28.9Raw NASA TLX
3
User Study Layout Generation
52.1Physical Plausibility
3
User Study 5-point Likert scale
3.66Subject Consistency
3
User Study Graph-conditional 100 samples (test)
51Winning Rate
3
User Study IGG (test)
46.9Preference (%)
3
User Study PST (test)
41.4Preference (%)
3
User Study Vlog
54Visual Quality
3
User Study (test)
2.72GQ
3
User Study
4.06Concept Consistency
3
User Study Prolific (results)
70.4Lip Sync Score
3
User Study
90.4Semantic Consistency
3
User Study
68.9Audio Quality Score
3
User Study SD 3
54.17Preference Rate
3
User Study SD 2.1
76.67Preference Rate
3