AI研究論文デイリー

OpenELM: An Efficient Language Model Family with Open-source Training and Inference Framework

Multi-Head Mixture-of-Experts

FlashSpeech: Efficient Zero-Shot Speech Synthesis

Pegasus-v1 Technical Report

SnapKV: LLM Knows What You are Looking for Before Generation

Align Your Steps: Optimizing Sampling Schedules in Diffusion Models

Transformers Can Represent n-gram Language Models