English
Related papers

Related papers: Mpemba Effect in Large-Language Model Training Dyn…

200 papers

High-quality data is scarce in large language model (LLM) training, yet how to schedule its use jointly with training dynamics lacks theoretical guidance. We extend functional scaling laws by incorporating a data-quality dimension, and…

Machine Learning · Computer Science 2026-05-26 Zhitao Zhu , Xili Wang , Shizhe Wu , Jiawei Fu , Xiaoqing Liu

Parameter-efficient methods like LoRA have revolutionised large language model (LLM) fine-tuning. ReLoRA extends this idea to pretraining by repeatedly merging and reinitialising low-rank adapters, increasing cumulative rank while keeping…

Computation and Language · Computer Science 2025-10-03 Yuval Weiss , David Demitri Africa , Paula Buttery , Richard Diehl Martinez

Training Transformers on algorithmic tasks frequently demonstrates an intriguing abrupt learning phenomenon: an extended performance plateau followed by a sudden, sharp improvement. This work investigates the underlying mechanisms for such…

Machine Learning · Computer Science 2025-10-24 Pulkit Gopalani , Wei Hu

In classical thermodynamics, the Mpemba effect refers to the counterintuitive observation that hot water can freeze faster than cold water, manifesting as an anomalous crossing of dynamical trajectories. While analogues of this phenomenon…

Quantum Physics · Physics 2026-05-25 Zhilong Liu , Zehua Tian , Jieci Wang

Large Language Model (LLM) pretraining, finetuning, and evaluation rely on input-space reconstruction and generative capabilities. Yet, it has been observed in vision that embedding-space training objectives, e.g., with Joint Embedding…

Computation and Language · Computer Science 2025-10-08 Hai Huang , Yann LeCun , Randall Balestriero

Learning dynamics, which describes how the learning of specific training examples influences the model's predictions on other examples, gives us a powerful tool for understanding the behavior of deep learning systems. We study the learning…

Machine Learning · Computer Science 2025-07-01 Yi Ren , Danica J. Sutherland

Traditional Large Language Model (LLM) pretraining relies on autoregressive language modeling with randomly sampled data from web-scale datasets. Inspired by human learning techniques like spaced repetition, we hypothesize that random…

Computation and Language · Computer Science 2025-01-30 Neha Prakriya , Jui-Nan Yen , Cho-Jui Hsieh , Jason Cong

As both model and dataset sizes continue to scale rapidly, conventional pretraining strategies with fixed compute budgets-such as cosine learning rate schedules-are increasingly inadequate for large-scale training. Recent alternatives,…

Machine Learning · Computer Science 2025-11-04 Minhak Song , Beomhan Baek , Kwangjun Ahn , Chulhee Yun

The highly complicated nature of far from equilibrium systems can lead to a complete breakdown of the physical intuition developed in equilibrium. A famous example of this is the Mpemba effect, which states that non-equilibrium states may…

Statistical Mechanics · Physics 2024-07-03 Colin Rylands , Katja Klobas , Filiberto Ares , Pasquale Calabrese , Sara Murciano , Bruno Bertini

Large language models (LLMs) bear promise as a fast and accurate material modeling paradigm for evaluation, analysis, and design. Their vast number of trainable parameters necessitates a wealth of data to achieve accuracy and mitigate…

Machine Learning · Computer Science 2024-07-04 Ning Liu , Siavash Jafarzadeh , Brian Y. Lattimer , Shuna Ni , Jim Lua , Yue Yu

Training Large Language Models (LLMs) from scratch requires immense computational resources, making it prohibitively expensive. Model scaling-up offers a promising solution by leveraging the parameters of smaller models to create larger…

Machine Learning · Computer Science 2025-02-20 Yifei Yang , Zouying Cao , Xinbei Ma , Yao Yao , Libo Qin , Zhi Chen , Hai Zhao

Unlike most of the research on the Mpemba effect which has focused on verifying the observation that warm water freezes faster than cold water, our work quantitatively investigates the rates at which hot and cold water cool and the point at…

Popular Physics · Physics 2011-01-17 Andrew Wang , Monica Chen , Yanni Vourgourakis , Antonio Nassar

The scaling trend in Large Language Models (LLMs) has prioritized increasing the maximum context window to facilitate complex, long-form reasoning and document analysis. However, managing this expanded context introduces severe…

Computation and Language · Computer Science 2026-01-21 Ahilan Ayyachamy Nadar Ponnusamy , Karthic Chandran , M Maruf Hossain

The Mpemba Effect can be studied with Markovian dynamics in a non-equilibrium thermodynamics framework. The Markovian Mpemba Effect can be observed in a variety of systems including the Ising model. We demonstrate that the Markovian Mpemba…

Disordered Systems and Neural Networks · Physics 2024-03-26 Felipe Amorim , Joey Wisely , Nathan Buckley , Christiana DiNardo , Daniel Sadasivan

Multimodal Large Language Models (MLLMs) hold great promise for advanced reasoning at the intersection of text and images, yet they have not fully realized this potential. MLLMs typically integrate an LLM, a vision encoder, and a connector…

Machine Learning · Computer Science 2025-11-07 Nikita Rajaneesh , Thomas Zollo , Richard Zemel

Learning rate warmup is a popular and practical technique in training large-scale deep neural networks. Despite the huge success in practice, the theoretical advantages of this strategy of gradually increasing the learning rate at the…

Machine Learning · Computer Science 2025-09-10 Yuxing Liu , Yuze Ge , Rui Pan , An Kang , Tong Zhang

The rapid advancement of Large Language Models (LLMs) has improved text understanding and generation but poses challenges in computational resources. This study proposes a curriculum learning-inspired, data-centric training strategy that…

Computation and Language · Computer Science 2024-05-14 Jisu Kim , Juhwan Lee

The quicker freezing of hotter water, than a colder sample, when quenched to a common lower temperature, is referred to as the Mpemba effect (ME). While this counter-intuitive fact remains a surprize since long, efforts have begun to…

Statistical Mechanics · Physics 2023-09-08 Sohini Chatterjee , Soumik Ghosh , Nalina Vadakkayil , Tanay Paul , Sanat K. Singha , Subir K. Das

Learning Rate Warmup is a popular heuristic for training neural networks, especially at larger batch sizes, despite limited understanding of its benefits. Warmup decreases the update size $\Delta \mathbf{w}_t = \eta_t \mathbf{u}_t$ early in…

Machine Learning · Computer Science 2024-11-01 Atli Kosson , Bettina Messmer , Martin Jaggi

Why do neural networks trained with large learning rates for a longer time often lead to better generalization? In this paper, we delve into this question by examining the relation between training and testing loss in neural networks.…

Machine Learning · Computer Science 2024-01-23 Yinuo Ren , Chao Ma , Lexing Ying