3 papers
cs.LG2025
A Unified Noise-Curvature View of Loss of Trainability
Gunbir Singh Baveja, Alex Lewandowski, Mark Schmidt
Loss of trainability refers to a phenomenon in continual learning where parameter updates no longer make progress on the optimization objective, so accuracy stalls or degrades as t…
cs.AI2025
ReMA: Learning to Meta-think for LLMs with Multi-Agent Reinforcement Learning
Ziyu Wan, Yunxiang Li, Xiaoyu Wen +8
Recent research on Reasoning of Large Language Models (LLMs) has sought to further enhance their performance by integrating meta-thinking -- enabling models to monitor, evaluate, a…
cs.LG2024
BlockLLM: Memory-Efficient Adaptation of LLMs by Selecting and Optimizing the Right Coordinate Blocks
Amrutha Varshini Ramesh, Vignesh Ganapathiraman, Issam H. Laradji +1
Training large language models (LLMs) for pretraining or adapting to new tasks and domains has become increasingly critical as their applications expand. However, as the model and…