unable to reproduce the results of Table 7

DAMO-NLP-SG / VCD

[CVPR 2024 Highlight] Mitigating Object Hallucinations in Large Vision-Language Models through Visual Contrastive Decoding

Apache License 2.0

219 stars 11 forks source link

Closed XueJiang16 closed 5 months ago

XueJiang16 commented 5 months ago

May you provide the codes how to combine VCD with other decoding strategies (like greedy)?

LengSicong commented 5 months ago

Hi, you can refer to the discussions in the following issue. Thanks. https://github.com/DAMO-NLP-SG/VCD/issues/8