Tsimpoukelli et al. show a way to adapt few-shot learning from language models to the multimodal setting, enabling strong few-shot/zero-shot performance on tasks like visual question answering. They introduce a method called Frozen, where images are encoded in the word embedding space of a…

SIGIR, the premier Information Retrieval conference was held last week. Ivan Vendrov and Chris Painter have an interesting substack post where they summarize papers at SIGIR that are related to recommender systems alignment and ethics.

Github Copilot is astonishingly good in many instances, and hilariously bad at several others. Open AI, the creators of Codex, the language model that powers Github Copilot, have released an evaluation framework for evaluating models on their ability to synthesize code from docstrings…