Compositional Hallucination Evaluation on OmniVCHall Average benchmark leaderboard · SOTA2 Research

Compositional Hallucination Evaluation on OmniVCHall Average

95AccuracyHuman

Evaluation Results

Method	Links
Human2026.01		95
Gemini-3-pro2026.01		77
Qwen3-VL-Thinking2026.01		76
GPT-5.22026.01		76
Doubao-Seed-1.82026.01		75
Qwen3-VL-Thinking2026.01		74
Qwen3-VL-Instruct2026.01		74
GPT-52026.01		73
Gemini-2.5-flash2026.01		73
Gemini-2.5-pro2026.01		73
Doubao-Seed-1.62026.01		73
Qwen3-VL-Instruct2026.01		72
Qwen3-VL-Thinking2026.01		71
Kimi-VL-Thinking2026.01		71
GPT-4o2026.01		71
Qwen3-VL-Thinking2026.01		70
Qwen3-VL-Instruct2026.01		70
Qwen3-VL-Instruct2026.01		68
Kimi-VL-Instruct2026.01		68
Molmo22026.01		67
Qwen2.5-VL-Instruct2026.01		67
Molmo22026.01		66
MiniCPM-V-4.52026.01		66
Qwen3-VL-Thinking2026.01		65
InternVL3.52026.01		65
InternVL3.52026.01		64
Qwen2.5-VL-Instruct2026.01		64
InternVL3.52026.01		64
Qwen3-VL-Instruct2026.01		63
VideoLLaMA32026.01		62
MiniCPM-V-42026.01		60
VideoChat-Flash2026.01		59
Qwen2.5-VL-Instruct2026.01		57
VideoChat-Flash2026.01		56
VideoLLaMA32026.01		54
GLM-4.6v2026.01		51
LLaVA-NeXT-Video2026.01		50
GLM-4.5v2026.01		49
LLaVA-NeXT-Video2026.01		47
GLM-4.6v-flash2026.01		46