Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Multimodal Long-Document Question Answering on DocBench
Loading...
82.1
LasJ Score
MARDoc
62.34
67.47
72.6
77.73
Jun 4, 2026
LasJ Score
Updated 1mo ago
Evaluation Results
Method
Method
Links
LasJ Score
MARDoc
Model=Qwen3-VL-30B-A3B...
2026.06
82.1
DocDancer
Model=Qwen3-VL-30B-A3B...
2026.06
81.2
Human Baseline
2026.06
81.2
DocAgent
Model=GPT-4o
2026.06
79.9
DocDancer
Model=Qwen3-VL-30B-A3B...
2026.06
74.1
DocDancer
Model=GPT-4o
2026.06
73.5
DocAgent
Model=Qwen3-VL-30B-A3B...
2026.06
73.2
MARDoc
Model=Qwen3-VL-8B-Inst...
2026.06
72.3
DocAgent
Model=Qwen3-VL-8B-Inst...
2026.06
67.1
Vanilla VLM
Model=Qwen3-VL-8B-Inst...
2026.06
64.6
RAGAnything
Model=GPT-4o-mini
2026.06
63.4
Vanilla VLM
Model=Qwen3-VL-30B-A3B...
2026.06
63.2
Vanilla VLM
Model=GPT-4o
2026.06
63.1
Feedback
Search any
task
Search any
task