List your company

Multimodal Hallucination Assessment on HallusionBench

70AccuracyQwen3.5-27B
46.360852.497958.63564.7721Apr 9, 2026Apr 14, 2026Apr 19, 2026Apr 24, 2026Apr 29, 2026May 4, 2026May 9, 2026

Evaluation results

MethodLinks
Qwen3.5-27BMode=REASONING, Architecture=Dense, # Total Params=27B, # Activated Params=27B2026.0470
Qwen3-VL-32BMode=Thinking, Architecture=Dense, # Total Params=33B, # Activated Params=33B2026.0467.4
Qwen3-VL-235B-A22BMode=Thinking, Architecture=MoE, # Total Params=236B, # Activated Params=23B2026.0466.7
EXAONE 4.5 33BMode=REASONING, Architecture=Dense, # Total Params=33B, # Activated Params=33B2026.0463.7
GPT-5 miniMode=REASONING: HIGH, Architecture=-, # Total Params=-, # Activated Params=-2026.0463.2
F3ARatio=60%, Backbone=InternVL3.5-8B2026.0554.88
InternVL3.5-8BRatio=100%, Backbone=InternVL3.5-8B2026.0554.77
VisionZipRatio=60%, Backbone=InternVL3.5-8B2026.0554.22
FastVRatio=60%, Backbone=InternVL3.5-8B2026.0553.35
DivPruneRatio=60%, Backbone=InternVL3.5-8B2026.0553.35
DivPruneRatio=40%, Backbone=InternVL3.5-8B2026.0552.42
CDPrunerRatio=60%, Backbone=InternVL3.5-8B2026.0552.35
VisionZipRatio=40%, Backbone=InternVL3.5-8B2026.0552.25
F3ARatio=40%, Backbone=InternVL3.5-8B2026.0552.25
CDPrunerRatio=40%, Backbone=InternVL3.5-8B2026.0551.83
FastVRatio=40%, Backbone=InternVL3.5-8B2026.0551.76
F3ARatio=20%, Backbone=InternVL3.5-8B2026.0550.55
CDPrunerRatio=20%, Backbone=InternVL3.5-8B2026.0548.46
DivPruneRatio=20%, Backbone=InternVL3.5-8B2026.0548.42
VisionZipRatio=20%, Backbone=InternVL3.5-8B2026.0548.31
FastVRatio=20%, Backbone=InternVL3.5-8B2026.0547.27