List your company

Generative Hallucination Evaluation on AMBER-g

2.2CHAIR ScoreUE-DPO
1.93.9255.957.975Apr 1, 2026Apr 6, 2026Apr 12, 2026Apr 18, 2026Apr 24, 2026Apr 30, 2026May 6, 2026

Evaluation results

MethodLinks
UE-DPOBackbone=LLaVA-v1.5-13B, Datasize=5.7k2026.052.2—0.914.4
UE-DPOBackbone=LLaVA-v1.5-7B, Datasize=16k, Training Data Source=RLAIF-V Data2026.052.5—0.912.5
UE-DPOBackbone=LLaVA-v1.5-7B, Datasize=5.7k, Training Data Source=RLHF-V Data2026.052.9—117.4
RLAIF-VBackbone=LLaVA-v1.5-7B, Datasize=16k2026.053—116.2
UE-DPOBackbone=Qwen2.5-VL-3B, Datasize=5.7k2026.053—0.614.7
mDPOBackbone=LLaVA-v1.5-7B, Datasize=10k2026.054.4—2.424.5
Qwen2.5VL-7B + ACTArchitecture=Qwen2.5VL-7B, ACT=Enabled2026.044.521.81.3—
LLaVA-1.5-13B + ACTArchitecture=LLaVA-1.5-13B, ACT=Enabled2026.044.625.32.3—
Qwen2VL-7B + ACTArchitecture=Qwen2VL-7B, ACT=Enabled2026.044.920.41.2—
LLaVA-1.5-7B + ACTArchitecture=LLaVA-1.5-7B, ACT=Enabled2026.045.127.42.7—
Qwen2.5VL-7BArchitecture=Qwen2.5VL-7B, ACT=Disabled2026.045.422.61—
V-DPOBackbone=LLaVA-v1.5-7B, Datasize=5.7k2026.055.6—2.727.3
DPOBackbone=Qwen2.5-VL-3B, Datasize=5.7k2026.055.7—1.422.6
RLHF-V(HD)Backbone=LLaVA-v1.5-13B, Datasize=1.4k2026.056.3—2.125.1
HALVABackbone=LLaVA-v1.5-13B, Datasize=22.5k2026.056.4—3.230.4
Qwen2VL-7BArchitecture=Qwen2VL-7B, ACT=Disabled2026.046.542.84—
HALVABackbone=LLaVA-v1.5-7B, Datasize=21.5k2026.056.6—3.432.2
LLaVA-v1.5-13BBackbone=LLaVA-v1.5-13B2026.056.8—3.331.8
InternVL2.5-8B + ACTArchitecture=InternVL2.5-8B, ACT=Enabled2026.047.163.87.6—
LLaVA-1.5-13BArchitecture=LLaVA-1.5-13B, ACT=Disabled2026.047.234.23.8—
POVIDBackbone=LLaVA-v1.5-7B, Datasize=17k2026.057.4—3.934.3
LLaVA-1.5-7BArchitecture=LLaVA-1.5-7B, ACT=Disabled2026.047.535.94.3—
InternVL2-8B + ACTArchitecture=InternVL2-8B, ACT=Enabled2026.047.656.26.6—
InternVL2.5-8BArchitecture=InternVL2.5-8B, ACT=Disabled2026.047.663.57.3—
LLaVA-v1.5-7BBackbone=LLaVA-v1.5-7B2026.057.7—4.234.7
LLaVA-RLHFBackbone=LLaVA-v1.5-13B, Datasize=122k2026.057.7—438.6
Qwen2.5-VL-3BBackbone=Qwen2.5-VL-3B2026.058—4.344.4
InternVL2-8BArchitecture=InternVL2-8B, ACT=Disabled2026.048.368.28.1—
LLaVA-RLHFBackbone=LLaVA-v1.5-7B, Datasize=122k2026.059.7—5.346.6