List your company
ResearchFormerly WizWand

VQA Hallucination on MMHal

3.87ScoreGPT-4o
1.85242.37622.93.4238Jul 29, 2025
Updated 7mo ago

Evaluation results

MethodLinks
GPT-4o2025.073.8724
Intern-VL2.5-7BModel Scale=7B2025.073.5426
Qwen-VL2.5-8BModel Scale=8B2025.073.2927
OPA-DPOBase Model=LLaVA-v1.5-13B, RL Technique=OPA-DPO2025.073.0240
RLAIFBase Model=LLaVA-v1.5-7B, RL Technique=RLAIF2025.072.8942
TARS (Mask)Base Model=LLaVA-v1.5-13B, Perturbation Strategy=Mask2025.072.8945
DeepSeek-VL2-27BModel Scale=27B2025.072.8427
OPA-DPOBase Model=LLaVA-v1.5-7B, RL Technique=OPA-DPO2025.072.7846
CHiP-DPOBase Model=LLaVA-v1.5-13B, RL Technique=CHiP-DPO2025.072.746
TARS (Replace)Base Model=LLaVA-v1.5-13B, Perturbation Strategy=Replace2025.072.6347
TARS (Replace)Base Model=LLaVA-v1.5-7B, Perturbation Strategy=Replace2025.072.5446
TARS (Mask)Base Model=LLaVA-v1.5-7B, Perturbation Strategy=Mask2025.072.4845
DPOBase Model=LLaVA-v1.5-13B, RL Technique=DPO2025.072.4850
LLaVA-v1.5-13BModel Scale=13B2025.072.3953
CHiP-DPOBase Model=LLaVA-v1.5-7B, RL Technique=CHiP-DPO2025.072.3257
HALVABase Model=LLaVA-v1.5-13B, RL Technique=HALVA2025.072.2856
DPOBase Model=LLaVA-v1.5-7B, RL Technique=DPO2025.072.1961
HALVABase Model=LLaVA-v1.5-7B, RL Technique=HALVA2025.072.1259
RLHFBase Model=LLaVA-v1.5-13B, RL Technique=RLHF2025.072.167
LLaVA-v1.5-7BModel Scale=7B2025.072.0261
RLHFBase Model=LLaVA-v1.5-7B, RL Technique=RLHF2025.071.9367