INSANE NEW TTS: Recurrent Depth (New Transformer)

Overview of TEST-TIME-Compute Scaling (TTS) from a simple search algo with Process reward Models (PRM) and Policy Models (PM) to the latest AI research on TTS via recurrent depth blocks that build a new transformer architecture for deep reasoning (math, science). Latent Space. Latent reasoning.

All rights w/ authors:
Scaling up Test-Time Compute with Latent Reasoning:
A Recurrent Depth Approach
Jonas Geiping, Sean McLeish, Neel Jain, John Kirchenbauer, Siddharth Singh, Brian R. Bartoldson,
Bhavya Kailkhura, Abhinav Bhatele, Tom Goldstein
from Max-Planck Institute for Intelligent Systems,
University of Maryland, College Park,
Lawrence Livermore National Laboratory @LivermoreLab

Can 1B LLM Surpass 405B LLM? Rethinking Compute-Optimal Test-Time Scaling
Runze Liu, Junqi Gao, Jian Zhao, Kaiyan Zhang, Xiu Li, Biqing Qi, Wanli Ouyang and Bowen Zhou
from Shanghai AI Laboratory,
Tsinghua University,
Harbin Institute of Technology,
BUPT

Scaling LLM Test-Time Compute Optimally can
be More Effective than Scaling Model Parameters
Charlie Snell, Jaehoon Lee, Kelvin Xu, and Aviral Kumar
from UC Berkeley and Google DeepMind

@UCBerkeley @Google_DeepMind

#airesearch
#testtime
#inference
#newtech