arXiv subjects

Deyi Liu

Publications and source records attributed to Deyi Liu.

12 recordsLinked to original sources

arXiv2026

LLMs as Noisy Channels: A Shannon Perspective on Model Capacity and Scaling Laws

Existing scaling laws for Large Language Models (LLMs), predominantly monotonic power laws, fail to explain emerging non-monotonic phenomena such as catastrophic overtraining and quantization-induced degradation, where performance deteriorates despite increased compute. We propose the Shannon Scaling Law, a unified theoretical framework that models LLM training as information transmission over a noisy channel, grounded in the Shannon-Hartley theorem. By mapping model parameters to channel bandwidth and training tokens to signal power, our formulation explicitly captures the interaction between learning signal and intrinsic noise. This perspective reveals a fundamental Shannon capacity for LLMs: scaling model size or data without preserving a sufficient signal-to-noise ratio (SNR) inevitably amplifies noise, inducing a transition from monotonic improvement to U-shaped performance degradation. We validate our theory through experiments on Pythia and OLMo2 under perturbations, including Gaussian noise, quantization and supervised fine-tuning on math, QA and code tasks. The Shannon Scaling Law consistently outperforms classical scaling laws and recent perturbation-aware laws, achieving strong $R^2$ scores and accurately capturing loss basins missed by prior approaches. It also extrapolates: fitted on $\leq$6.9B Pythia models with $\leq$180B tokens, it predicts the unseen 12B model up to 307B tokens at pooled $R^2{=}0.847$, while monotonic baselines collapse.

cs.LG↗

arXiv2026

SPARKLING: Balancing Signal Preservation and Symmetry Breaking for Width-Progressive Learning

Progressive Learning (PL) reduces pre-training computational overhead by gradually increasing model scale. While prior work has extensively explored depth expansion, width expansion remains significantly understudied, with the few existing methods limited to the early stages of training. However, expanding width during the mid-stage is essential for maximizing computational savings, yet it remains a formidable challenge due to severe training instabilities. Empirically, we show that naive initialization at this stage disrupts activation statistics, triggering loss spikes, while copy-based initialization introduces gradient symmetry that hinders feature diversity. To address these issues, we propose SPARKLING (balancing {S}ignal {P}reservation {A}nd symmet{R}y brea{K}ing for width-progressive {L}earn{ING}), a novel framework for mid-stage width expansion. Our method achieves signal preservation via RMS-scale consistency, stabilizing activation statistics during expansion. Symmetry breaking is ensured through asymmetric optimizer state reset and asymmetric learning rate re-warmup. Extensive experiments on dense and Mixture-of-Experts (MoE) models demonstrate that, across multiple width axes and optimizer families, SPARKLING consistently outperforms training from scratch and reduces training cost by up to 35% under $2\times$ width expansion.

cs.LG↗

arXiv2025

GatePro: Parameter-Free Expert Selection Optimization for Mixture-of-Experts Models

Modern large language models leverage Mixture-of-Experts (MoE) architectures for efficient scaling, but face a critical challenge: functionally similar experts are often selected simultaneously, creating redundant computation and limiting effective model capacity. Existing auxiliary balance loss methods improve token distribution but fail to address the underlying expert diversity problem. We introduce GatePro, a novel parameter-free method that directly promotes expert selection diversity. GatePro identifies the most similar expert pairs and introduces localized competition mechanisms, preventing redundant expert co-activation while maintaining natural expert specialization. Our comprehensive evaluation demonstrates GatePro's effectiveness across model scales and benchmarks. Analysis demonstrates GatePro's ability to achieve enhanced expert diversity, where experts develop more distinct and complementary capabilities, avoiding functional redundancy. This approach can be deployed hot-swappable during any training phase without additional learnable parameters, offering a practical solution for improving MoE effectiveness.

cs.CL↗

arXiv2025

Balanced Actor Initialization: Stable RLHF Training of Distillation-Based Reasoning Models

The development of alignment and reasoning capabilities in large language models has seen remarkable progress through two paradigms: instruction tuning and reinforcement learning from human feedback (RLHF) alignment paradigm, and distillation-based reasoning fine-tuning paradigm. While both approaches prove effective independently, the third paradigm of applying RLHF to distillation-trained models presents significant challenges. Our investigation reveals two critical phenomena that emerge in this paradigm: Sequence Length Collapse, where language generation dramatically reduces during early RLHF training, and the Reward Hockey Stick Curve, featuring severe reward score drops followed by gradual recovery. These instabilities fundamentally compromise the model's alignment and reasoning capabilities. To address these challenges, we propose Balanced Actor Initialization (BAI), a two-stage weighted model merging approach. BAI first merges instruction-following and distillation-based reasoning fine-tuned models, then further combines this intermediate model with the pretrained model to preserve foundational knowledge. Through comprehensive experiments across diverse benchmarks and detailed analysis of training experiments, we demonstrate that BAI resolves Sequence Length Collapse, mitigates the Reward Hockey Stick Curve, and enables continuous sequence length improvement during training. Additionally, our analysis reveals that balanced merging ratios achieve optimal trade-offs between training stability and reasoning capability preservation. Our work provides the effective solution for stable training in this third paradigm, enabling more capable reasoning models that combine distillation efficiency with RLHF alignment.

cs.CL↗

arXiv2025

Model Merging in Pre-training of Large Language Models

Model merging has emerged as a promising technique for enhancing large language models, though its application in large-scale pre-training remains relatively unexplored. In this paper, we present a comprehensive investigation of model merging techniques during the pre-training process. Through extensive experiments with both dense and Mixture-of-Experts (MoE) architectures ranging from millions to over 100 billion parameters, we demonstrate that merging checkpoints trained with constant learning rates not only achieves significant performance improvements but also enables accurate prediction of annealing behavior. These improvements lead to both more efficient model development and significantly lower training costs. Our detailed ablation studies on merging strategies and hyperparameters provide new insights into the underlying mechanisms while uncovering novel applications. Through comprehensive experimental analysis, we offer the open-source community practical pre-training guidelines for effective model merging.

cs.CL↗

arXiv2021

A New Randomized Primal-Dual Algorithm for Convex Optimization with Optimal Last Iterate Rates

We develop a novel unified randomized block-coordinate primal-dual algorithm to solve a class of nonsmooth constrained convex optimization problems, which covers different existing variants and model settings from the literature. We prove that our algorithm achieves optimal $\mathcal{O}(n/k)$ and $\mathcal{O}(n^2/k^2)$ convergence rates (up to a constant factor) in two cases: general convexity and strong convexity, respectively, where $k$ is the iteration counter and n is the number of block-coordinates. Our convergence rates are obtained through three criteria: primal objective residual and primal feasibility violation, dual objective residual, and primal-dual expected gap. Moreover, our rates for the primal problem are on the last iterate sequence. Our dual convergence guarantee requires additionally a Lipschitz continuity assumption. We specify our algorithm to handle two important special cases, where our rates are still applied. Finally, we verify our algorithm on two well-studied numerical examples and compare it with two existing methods. Our results show that the proposed method has encouraging performance on different experiments.

math.OC↗

arXiv2021

Robust and Generalizable Visual Representation Learning via Random Convolutions

While successful for various computer vision tasks, deep neural networks have shown to be vulnerable to texture style shifts and small perturbations to which humans are robust. In this work, we show that the robustness of neural networks can be greatly improved through the use of random convolutions as data augmentation. Random convolutions are approximately shape-preserving and may distort local textures. Intuitively, randomized convolutions create an infinite number of new domains with similar global shapes but random local textures. Therefore, we explore using outputs of multi-scale random convolutions as new images or mixing them with the original images during training. When applying a network trained with our approach to unseen domains, our method consistently improves the performance on domain generalization benchmarks and is scalable to ImageNet. In particular, in the challenging scenario of generalizing to the sketch domain in PACS and to ImageNet-Sketch, our method outperforms state-of-art methods by a large margin. More interestingly, our method can benefit downstream tasks by providing a more robust pretrained visual representation.

cs.CV↗

arXiv2021

A New Primal-Dual Algorithm for a Class of Nonlinear Compositional Convex Optimization Problems

We develop a novel primal-dual algorithm to solve a class of nonsmooth and nonlinear compositional convex minimization problems, which covers many existing and brand-new models as special cases. Our approach relies on a combination of a new nonconvex potential function, Nesterov's accelerated scheme, and an adaptive parameter updating strategy. Our algorithm is single-loop and has low per-iteration complexity. Under only general convexity and mild assumptions, our algorithm achieves $\mathcal{O}(1/k)$ convergence rates through three different criteria: primal objective residual, dual objective residual, and primal-dual gap, where $k$ is the iteration counter. Our rates are both ergodic (i.e., on an averaging sequence) and non-ergodic (i.e., on the last-iterate sequence). These convergence rates can be accelerated up to $\mathcal{O}(1/k^2)$ if only one objective term is strongly convex (or equivalently, its conjugate is $L$-smooth). To the best of our knowledge, this is the first algorithm achieving optimal rates on the primal last-iterate sequence for nonlinear compositional convex minimization. As a by-product, we specify our algorithm to solve a general convex cone constrained program with both ergodic and non-ergodic rate guarantees. We test our algorithms and compare them with two recent methods on a binary classification and a convex-concave game model.

math.OC↗

arXiv2020

Hybrid Variance-Reduced SGD Algorithms For Nonconvex-Concave Minimax Problems

We develop a novel and single-loop variance-reduced algorithm to solve a class of stochastic nonconvex-convex minimax problems involving a nonconvex-linear objective function, which has various applications in different fields such as machine learning and robust optimization. This problem class has several computational challenges due to its nonsmoothness, nonconvexity, nonlinearity, and non-separability of the objective functions. Our approach relies on a new combination of recent ideas, including smoothing and hybrid biased variance-reduced techniques. Our algorithm and its variants can achieve $\mathcal{O}(T^{-2/3})$-convergence rate and the best known oracle complexity under standard assumptions, where $T$ is the iteration counter. They have several computational advantages compared to existing methods such as simple to implement and less parameter tuning requirements. They can also work with both single sample or mini-batch on derivative estimators, and with constant or diminishing step-sizes. We demonstrate the benefits of our algorithms over existing methods through two numerical examples, including a nonsmooth and nonconvex-non-strongly concave minimax model.

math.OC↗

arXiv2020

An Optimal Hybrid Variance-Reduced Algorithm for Stochastic Composite Nonconvex Optimization

In this note we propose a new variant of the hybrid variance-reduced proximal gradient method in [7] to solve a common stochastic composite nonconvex optimization problem under standard assumptions. We simply replace the independent unbiased estimator in our hybrid- SARAH estimator introduced in [7] by the stochastic gradient evaluated at the same sample, leading to the identical momentum-SARAH estimator introduced in [2]. This allows us to save one stochastic gradient per iteration compared to [7], and only requires two samples per iteration. Our algorithm is very simple and achieves optimal stochastic oracle complexity bound in terms of stochastic gradient evaluations (up to a constant factor). Our analysis is essentially inspired by [7], but we do not use two different step-sizes.

math.OC↗

arXiv2020

A Newton Frank-Wolfe Method for Constrained Self-Concordant Minimization

We demonstrate how to scalably solve a class of constrained self-concordant minimization problems using linear minimization oracles (LMO) over the constraint set. We prove that the number of LMO calls of our method is nearly the same as that of the Frank-Wolfe method in the L-smooth case. Specifically, our Newton Frank-Wolfe method uses $\mathcal{O}(ε^{-ν})$ LMO's, where $ε$ is the desired accuracy and $ν:= 1 + o(1)$. In addition, we demonstrate how our algorithm can exploit the improved variants of the LMO-based schemes, including away-steps, to attain linear convergence rates. We also provide numerical evidence with portfolio design with the competitive ratio, D-optimal experimental design, and logistic regression with the elastic net where Newton Frank-Wolfe outperforms the state-of-the-art.

math.OC↗

arXiv2019

An Inexact Interior-Point Lagrangian Decomposition Algorithm with Inexact Oracles

We develop a new inexact interior-point Lagrangian decomposition method to solve a wide range class of constrained composite convex optimization problems. Our method relies on four techniques: Lagrangian dual decomposition, self-concordant barrier smoothing, path-following, and proximal-Newton technique. It also allows one to approximately compute the solution of the primal subproblems (called the slave problems), which leads to inexact oracles (i.e., inexact gradients and Hessians) of the smoothed dual problem (called the master problem). The smoothed dual problem is nonsmooth, we propose to use an inexact proximal-Newton method to solve it. By appropriately controlling the inexact computation at both levels: the slave and master problems, we still estimate a polynomial-time iteration-complexity of our algorithm as in standard short-step interior-point methods. We also provide a strategy to recover primal solutions and establish complexity to achieve an approximate primal solution. We illustrate our method through two numerical examples on well-known models with both synthetic and real data and compare it with some existing state-of-the-art methods.

math.OC↗