Multi-modal Preference Alignment Remedies Degradation of Visual Instruction Tuning on Language Models.
Shengzhi LiRongyu LinShichao PeiPublished in: ACL (1) (2024)
Keyphrases
- multi modal
- language model
- cross modal
- language modeling
- n gram
- video search
- information retrieval
- probabilistic model
- speech recognition
- document retrieval
- language modelling
- single modality
- smoothing methods
- retrieval model
- test collection
- query expansion
- statistical language models
- high dimensional
- audio visual
- image annotation
- multi modality
- visual information
- visual features
- relevance model
- language models for information retrieval
- document ranking
- multiple modalities
- translation model
- co occurrence
- low level
- multimedia