Sign in Subscribe

Plain English Papers

Example from the paper

Meet ALMA: A New Training Method That Boosts Translation Performance for Large Language Models

Researchers from Johns Hopkins and Microsoft propose a new 2-stage fine-tuning method that unlocks stronger translation abilities in smaller models with just 7-13 billion parameters.

Example performance data from the paper

LongLoRA: A New, More Efficient Way to Fine-Tune LLMs

Meet LongLoRA, an efficient fine-tuning approach that extends the context sizes of pre-trained large language models (LLMs), with limited computation cost.

Example DreamLLM generation

By Teaching AI to Make Pictures and Write, Scientists Improve Its Grasp of Vision and Language

Generating images enhances this AI's ability to comprehend text, researchers find.

Example Neural Motion Texture generation

Bringing Still Pictures to Life with Neural Motion Textures

Google researchers published a new technique to animate still photos by predicting how each pixel should move over time

New Research Says LLMs Are Surprisingly Good at Compressing... Images and Audio?

New Research Says LLMs Are Surprisingly Good at Compressing... Images and Audio?

New research shows LLMs like GPT-3 have excellent compression capabilities, able to shrink down text, images, audio and more.

Example LLM improvement

Contrastive Decoding: A Promising New Technique for Boosting Reasoning in Large Language Models

How using an "expert" and an "amateur" LLM can improve performance with the contrastive decoding method.

Paper screenshot

A New Multilingual Dataset for Training AI Models in 167 Languages

How can the CulturaX dataset help people who don't speak English access the benefits of AI?