2022–11–21 | Your Daily AI Research tl;dr 🧠

Galactica and controversies with large language models, Token Turing Machines and improving robustness with Vision Transformers.

Welcome to your official daily AI research tl;dr (often with code and news) for AI professionals where I share the most exciting papers I find daily, along with a one-liner summary to help you quickly determine if the article (and code) is worth investigating.

1️⃣ Optimizing Relevance Maps of Vision Transformers Improves Robustness

Monitoring the model’s relevancy signal and manipulating it such that the model is focused on the foreground object rather than its background- which hurts the model’s ability to generalize.

Link to the paper: https://arxiv.org/abs/2206.01161

2️⃣ Token Turing Machines

“We propose Token Turing Machines (TTM), a sequential, autoregressive Transformer model with memory for real-world sequential visual understanding.”

Link to the paper: https://arxiv.org/abs/2211.09119

Join the discussion with fellow AI enthusiasts

🆕 Galactica and controversies with…