From Pixels and Words to Waves: A Unified Framework for Spectral Dictionary vLLMs
Fuente:
arXiv
Salvato in:
| Autori principali: | , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866912445972348928 |
|---|---|
| author | Kiruluta, Andrew Burity, Priscilla |
| author_facet | Kiruluta, Andrew Burity, Priscilla |
| contents | Vision-language models (VLMs) unify computer vision and natural language processing in a single architecture capable of interpreting and describing images. Most state-of-the-art systems rely on two computationally intensive components: convolutions in the vision encoder and quadratic self-attention for multimodal fusion. This work removes both by introducing a spectral dictionary token mixer, which represents each image patch or wordpiece as a sparse combination of learnable frequency atoms. Our 1.1B-parameter prototype, SDict-VLM, achieves BLEU-4 of 39.2, CIDEr of 127.5, and SPICE of 27.0 on MS-COCO captioning, along with 50.3 percent accuracy on VQAv2. These results close approximately 85 percent of the performance gap to BLIP-2 while using 60 percent fewer parameters, 2.3 times less peak GPU memory, and 2.2 times faster inference than PaLI-3. To our knowledge, this is the first VLM to eliminate both convolutions and self-attention while matching mid-scale transformer baselines. In addition to its O(L log L) complexity, the shared frequency dictionary enables transparent cross-modal alignment and offers a tunable trade-off between accuracy and compute, paving the way for efficient and interpretable VLMs. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_18943 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | From Pixels and Words to Waves: A Unified Framework for Spectral Dictionary vLLMs Kiruluta, Andrew Burity, Priscilla Computer Vision and Pattern Recognition Vision-language models (VLMs) unify computer vision and natural language processing in a single architecture capable of interpreting and describing images. Most state-of-the-art systems rely on two computationally intensive components: convolutions in the vision encoder and quadratic self-attention for multimodal fusion. This work removes both by introducing a spectral dictionary token mixer, which represents each image patch or wordpiece as a sparse combination of learnable frequency atoms. Our 1.1B-parameter prototype, SDict-VLM, achieves BLEU-4 of 39.2, CIDEr of 127.5, and SPICE of 27.0 on MS-COCO captioning, along with 50.3 percent accuracy on VQAv2. These results close approximately 85 percent of the performance gap to BLIP-2 while using 60 percent fewer parameters, 2.3 times less peak GPU memory, and 2.2 times faster inference than PaLI-3. To our knowledge, this is the first VLM to eliminate both convolutions and self-attention while matching mid-scale transformer baselines. In addition to its O(L log L) complexity, the shared frequency dictionary enables transparent cross-modal alignment and offers a tunable trade-off between accuracy and compute, paving the way for efficient and interpretable VLMs. |
| title | From Pixels and Words to Waves: A Unified Framework for Spectral Dictionary vLLMs |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2506.18943 |