A comprehensive index of artificial intelligence and machine-learning research with AI-generated summaries, citation metrics, and direct links to papers and code.
Runze Liu, Junqi Gao, Jian Zhao, et al.
This paper investigates compute-optimal test-time scaling for LLMs, showing that smaller models with optimal TTS can outperform much larger ones.
Unknown
SoViT introduces compute-optimal shape scaling for vision transformers, achieving performance of models twice its size with equivalent compute.
Unknown
AgentTTS introduces an LLM agent that integrates prior insights to achieve compute-optimal scaling for test-time allocation in complex tasks.