Sign in Subscribe

Plain English Papers

Showing the result of adding explicit registers

Researchers discover explicit registers eliminate vision transformer attention spikes

When visualizing the inner workings of vision transformers (ViTs), researchers noticed weird spikes of attention on random background patches. Here's how they fixed them.

Examples

Marrying Pixel and Latent Diffusion Models for Efficient and High-Quality Text-to-Video Generation

A new paper proposes Show-1, a hybrid model that combines pixel and latent diffusion for efficient high-quality text-to-video generation.

VideoDirectorGPT diagram showing how the tool works

UNC Researchers Present VideoDirectorGPT: Using AI to Generate Multi-Scene Videos from Text

The key innovation proposed is decomposing multi-scene video generation into two steps: a director step and a "film" step.

Kohlberg model.

Microsoft Researchers Propose AI Morality Test for LLMs

The authors of a new paper combined human psychology and AI research to create a "defining issues test" for LLMs.

DeepMind finds a way to study large model instabilities without a ton of GPUs

How to use small models to study large ones (saving compute in the process).

CodePlan flow diagram.

Microsoft Researchers Announce CodePlan: Automating Complex Software Engineering Tasks with AI

Automating complex software engineering tasks with AI: A deep dive into CodePlan.

Example generations from GTP4Tools

Meet GPT4Tools: teaching existing LLMs how to use tools for visual tasks

GPT4Tools: using ChatGPT as a teacher to show other LLMs how to use other tools for visual tasks.