Multimodal Neurons in Pretrained Text-Only Transformers

Sarah Schwettmann, Neil Chowdhury, Samuel Klein, David Bau, Antonio Torralba; Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops, 2023, pp. 2862-2867


Language models demonstrate remarkable capacity to generalize representations learned in one modality to downstream tasks in other modalities. Can we trace this ability to individual neurons? We study the case where a frozen text transformer is augmented with vision using a self-supervised visual encoder and a single linear projection learned on an image-to-text task. Outputs of the projection layer are not immediately decodable into language describing image content; instead, we find that translation between modalities occurs deeper within the transformer. We introduce a procedure for identifying "multimodal neurons" that convert visual representations into corresponding text, and decoding the concepts they inject into the model's residual stream. In a series of experiments, we show that multimodal neurons operate on specific visual concepts across inputs, and have a systematic causal effect on image captioning.

Related Material

[pdf] [supp]
@InProceedings{Schwettmann_2023_ICCV, author = {Schwettmann, Sarah and Chowdhury, Neil and Klein, Samuel and Bau, David and Torralba, Antonio}, title = {Multimodal Neurons in Pretrained Text-Only Transformers}, booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops}, month = {October}, year = {2023}, pages = {2862-2867} }