Enabling Language Models to Implicitly Learn Self-Improvement
Rather than manually distilling criteria into prompts, implicit information in preference data can be leveraged.
Latest
LLMs can be extended to infinite sequence lengths without fine-tuning
LLMs trained with a finite attention window can be extended to infinite sequence lengths without any fine-tuning.
Tool-Integrated Reasoning: A New Approach for Math-Savvy LLMs
TORA combines both rationale-based and program-based reasoning to deliver results to math problems that were previously too difficult for LLMs to solve
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.
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.
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.
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.