List your company

Visual Hallucination Evaluation on AMBER Discriminative (Acc, Prec, Rec, F1)

89AccuracyHeRA
83.38484.84286.387.758Jun 22, 2026

Evaluation results

MethodLinks
HeRABackbone=Qwen3-8B2026.068992.490.991.6
HeRABackbone=Qwen2.5-14B2026.068990.493.491.9
HeRABackbone=Qwen3-14B2026.0688.990.792.791.7
Qwen3-14BBackbone=Qwen3-14B2026.0688.891.691.591.5
HeRABackbone=Qwen2.5-7B2026.0687.789.192.890.9
Qwen2.5-7BBackbone=Qwen2.5-7B2026.0687.689.691.990.7
Qwen3-8BBackbone=Qwen3-8B2026.0687.390.390.690.4
Qwen3-4BBackbone=Qwen3-4B2026.0686.589.989.689.7
HeRABackbone=Qwen3-4B2026.0686.590.189.489.7
Qwen2.5-14BBackbone=Qwen2.5-14B2026.0686.287.991.889.8
HeRABackbone=Vicuna-7B2026.0685.990.887.789.2
LLama3-8BBackbone=LLama3-8B2026.0685.989.289.689.4
HeRABackbone=LLama3-8B2026.0685.988.390.889.5
HeRABackbone=Vicuna-13B2026.0685.791.786.288.9
HeRABackbone=Qwen2.5-3B2026.0685.686.592.889.5
Vicuna-13BBackbone=Vicuna-13B2026.0684.893.982.487.8
Vicuna-7BBackbone=Vicuna-7B2026.0684.29085.887.8
Qwen2.5-3BBackbone=Qwen2.5-3B2026.0683.683.993.288.3