Papers › Enabling Multimodal Generation on CLIP via Vision-Language Knowledge Distillation

Enabling Multimodal Generation on CLIP via Vision-Language Knowledge Distillation

16 Nov 2021ACL ARR November 2021 11archive 2025-07-28

Anonymous

The recent large-scale vision-language pre-training (VLP) of dual-stream architectures (e.g., CLIP) with a tremendous amount of image-text pair data, has shown its superiority on various multimodal alignment tasks. Despite its success, the resulting models are not capable of generative multimodal tasks due to the weak text encoder. To tackle this problem, we propose to augment the dual-stream VLP model with a textual pre-trained language model (PLM) via vision-language knowledge distillation (VLKD), enabling the capability for multimodal generation. VLKD is pretty data- and computation-efficient compared to the pre-training from scratch. Experimental results show that the resulting model has strong zero-shot performance on multimodal generation tasks, such as open-ended visual question answering and image captioning. For example, it achieves 39.7% zero-shot accuracy on the VQA 2.0 dataset, surpassing the previous state-of-the-art zero-shot model with 14x fewer parameters. Furthermore, the original text processing ability of the PLM is maintained after VLKD, which makes our model versatile for both multimodal and unimodal tasks.

PaperPDF

Code

No code repository is listed for this paper in the archive or in Syntology's graph.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Image CaptioningKnowledge DistillationLanguage ModelingLanguage ModellingQuestion AnsweringVisual Question AnsweringVisual Question Answering (VQA)multimodal generation

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Image Captioning COCO Captions VLKD (ViT-B/16) BLEU-4 16.7 #34 of 41 Archive leaderboard report
Image Captioning COCO Captions VLKD (ViT-B/16) CIDER 58.3 #34 of 41 Archive leaderboard report
Image Captioning COCO Captions VLKD (ViT-B/16) METEOR 19.7 #34 of 41 Archive leaderboard report
Image Captioning COCO Captions VLKD (ViT-B/16) SPICE 13.4 #34 of 41 Archive leaderboard report
Visual Question Answering (VQA) OK-VQA VLKD(ViT-B/16) Accuracy 10.5 #36 of 37 Archive leaderboard report
Visual Question Answering (VQA) VQA v2 test-dev VLKD Accuracy 44.5 #56 of 56 Archive leaderboard report
Visual Question Answering (VQA) VQA v2 val VLKD(ViT-B/16) Accuracy 38.6 #10 of 11 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Knowledge Distillation

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections