Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
High-Resolution Visual Reasoning on HR-Bench 4K
Loading...
83.4
Accuracy
GazeVLM (Ours)
58.024
64.612
71.2
77.788
Jul 8, 2025
Aug 27, 2025
Oct 17, 2025
Dec 7, 2025
Jan 26, 2026
Mar 18, 2026
May 8, 2026
Accuracy
Updated 23d ago
Evaluation Results
Method
Method
Links
Accuracy
GazeVLM (Ours)
Gaze Bias=Enabled, Bas...
2026.05
83.4
Qwen3-VL-4B
Training Paradigm=Vani...
2026.05
79.5
GazeVLM (w/o gaze bias)
Gaze Bias=None, Base M...
2026.05
79
Qwen3.5-4B-Inst
Training Paradigm=Vani...
2026.05
78.6
DeepEyes
Training Paradigm=RL-t...
2026.05
75.1
Ground-R1-7B
Training Paradigm=RL-t...
2026.05
75
MGPO
Base Model=Qwen2.5-VL-7B
2025.07
74.2
Pixel-Reasoner-7B
Training Paradigm=RL-t...
2026.05
74
MGPO
Base Model=Qwen2.5-VL-3B
2025.07
70.9
SPARC-4B
Training Paradigm=SFT-...
2026.05
70.5
GRPO
Base Model=Qwen2.5-VL-7B
2025.07
69.8
Qwen2.5-VL-7B
Training Paradigm=Vani...
2026.05
68.5
GRPO
Base Model=Qwen2.5-VL-3B
2025.07
67.9
Gemini2.5-Flash-Lite
Training Paradigm=Clos...
2026.05
67.2
GPT-4o
Training Paradigm=Clos...
2026.05
59
Feedback
Search any
task
Search any
task