LongProLIP: A Probabilistic Vision-Language Model with Long Context Text
This work addresses a specific bottleneck in vision-language models for researchers and practitioners, but it is incremental as it fine-tunes an existing method rather than introducing a new paradigm.
The paper tackles the limitation of Probabilistic Language-Image Pre-Training (ProLIP) models in handling long context texts beyond 64 tokens by proposing a fine-tuning strategy to extend this to 256 tokens, showing improved long-context understanding on Urban-1k and DataComp benchmarks while noting a trade-off with general zero-shot capability.
Recently, Probabilistic Language-Image Pre-Training (ProLIP) has been proposed to tackle the multiplicity issue of vision-language (VL) tasks. Despite their success in probabilistic representation learning at a scale, the ProLIP models cannot handle long context texts longer than 64 context length, which limits their ability to capture rich contextual information from longer text sequences. To address this issue, this paper proposes a fine-tuning strategy for ProLIP to accept longer texts, e.g., 256 text tokens. Experimental results on Urban-1k and the DataComp evaluation suite show that the proposed LongProLIP recipe can improve understanding of long contexts while minimizing the negative effect of fine-tuning.We also observe a trade-off between the long context understanding (measured by Urban-1k) and general zero-shot capability (measured by evaluation datasets by DataComp). Code is available at https://github.com/naver-ai/prolip