Hallucination Evaluation on POPE v1.0 (test)
87POPE ScoreMoE-LLaVA-2B
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MoE-LLaVA-2BLLM=Qwen-1.5-1.8B, #Samples=2.2M2026.03 | 87 | — | — | |
| MobileVLM V2 7BLLM=Vicuna-7B, #Samples=3.6M2026.03 | 86.6 | — | — | |
| Cosine KD w/ Beta-KD (Instance)LLM=MobileLLaMA 1.4B, #Samples=3.6M2026.03 | 86.6 | — | — | |
| LLaVA-1.5LLM=Vicuna-7B, #Samples=1.2M2026.03 | 85.9 | — | — | |
| ShareGPT4VLLM=Vicuna-7B, #Samples=1.9M2026.03 | 85.7 | — | — | |
| Align-KDLLM=MobileLLaMA 1.4B, #Samples=3.6M2026.03 | 85.4 | — | — | |
| Align-KD w/ Beta-KD (Task)LLM=MobileLLaMA 1.4B, #Samples=3.6M2026.03 | 85.4 | — | — | |
| Align-KD w/ Beta-KD (Instance)LLM=MobileLLaMA 1.4B, #Samples=3.6M2026.03 | 85.4 | — | — | |
| Cosine KD w/ Beta-KD (Task)LLM=MobileLLaMA 1.4B, #Samples=3.6M2026.03 | 85.4 | — | — | |
| MobileVLM -V2 1.7BLLM=MobileLLaMA 1.4B, #Samples=3.6M2026.03 | 85.3 | — | — | |
| MobileVLM 3BLLM=MobileLLaMA 2.7B, #Samples=3.6M2026.03 | 84.9 | — | — | |
| LLAVADILLM=MobileLLaMA 1.4B, #Samples=3.6M2026.03 | 84.7 | — | — | |
| Cosine KDLLM=MobileLLaMA 1.4B, #Samples=3.6M2026.03 | 84.6 | — | — | |
| MobileVLM 1.7BLLM=MobileLLaMA 1.4B, #Samples=3.6M2026.03 | 84.5 | — | — | |
| MoE-LLaVA-1.6BLLM=StableLM-1.6B, #Samples=2.2M2026.03 | 84.3 | — | — | |
| LLaVA-1.5 13BVIG training=false2026.02 | — | 85.72 | 87.05 | |
| LLaVA-1.5 13B + VIG trainingVIG training=true2026.02 | — | 86.95 | 87.53 | |
| LLaVA-1.5 7BVIG training=false2026.02 | — | 85.9 | 87.08 | |
| LLaVA-1.5 7B + VIG trainingVIG training=true2026.02 | — | 85.93 | 87.47 | |
| ShareGPT4V 7BVIG training=false2026.02 | — | 85.69 | 86.98 | |
| ShareGPT4V 7B + VIG trainingVIG training=true2026.02 | — | 87.15 | 87.24 |