arXiv subjects

Tong Yang

Publications and source records attributed to Tong Yang.

At least 127 records · Page 7Linked to original sources

arXiv2025

Faster WIND: Accelerating Iterative Best-of-$N$ Distillation for LLM Alignment

Recent advances in aligning large language models with human preferences have corroborated the growing importance of best-of-N distillation (BOND). However, the iterative BOND algorithm is prohibitively expensive in practice due to the sample and computation inefficiency. This paper addresses the problem by revealing a unified game-theoretic connection between iterative BOND and self-play alignment, which unifies seemingly disparate algorithmic paradigms. Based on the connection, we establish a novel framework, WIN rate Dominance (WIND), with a series of efficient algorithms for regularized win rate dominance optimization that approximates iterative BOND in the parameter space. We provides provable sample efficiency guarantee for one of the WIND variant with the square loss objective. The experimental results confirm that our algorithm not only accelerates the computation, but also achieves superior sample efficiency compared to existing methods.

cs.LG↗

arXiv2025

Incentivize without Bonus: Provably Efficient Model-based Online Multi-agent RL for Markov Games

Multi-agent reinforcement learning (MARL) lies at the heart of a plethora of applications involving the interaction of a group of agents in a shared unknown environment. A prominent framework for studying MARL is Markov games, with the goal of finding various notions of equilibria in a sample-efficient manner, such as the Nash equilibrium (NE) and the coarse correlated equilibrium (CCE). However, existing sample-efficient approaches either require tailored uncertainty estimation under function approximation, or careful coordination of the players. In this paper, we propose a novel model-based algorithm, called VMG, that incentivizes exploration via biasing the empirical estimate of the model parameters towards those with a higher collective best-response values of all the players when fixing the other players' policies, thus encouraging the policy to deviate from its current equilibrium for more exploration. VMG is oblivious to different forms of function approximation, and permits simultaneous and uncoupled policy updates of all players. Theoretically, we also establish that VMG achieves a near-optimal regret for finding both the NEs of two-player zero-sum Markov games and CCEs of multi-player general-sum Markov games under linear function approximation in an online environment, which nearly match their counterparts with sophisticated uncertainty quantification.

cs.LG↗

arXiv2025

LLM-Sketch: Enhancing Network Sketches with LLM

Network stream mining is fundamental to many network operations. Sketches, as compact data structures that offer low memory overhead with bounded accuracy, have emerged as a promising solution for network stream mining. Recent studies attempt to optimize sketches using machine learning; however, these approaches face the challenges of lacking adaptivity to dynamic networks and incurring high training costs. In this paper, we propose LLM-Sketch, based on the insight that fields beyond the flow IDs in packet headers can also help infer flow sizes. By using a two-tier data structure and separately recording large and small flows, LLM-Sketch improves accuracy while minimizing memory usage. Furthermore, it leverages fine-tuned large language models (LLMs) to reliably estimate flow sizes. We evaluate LLM-Sketch on three representative tasks, and the results demonstrate that LLM-Sketch outperforms state-of-the-art methods by achieving a $7.5\times$ accuracy improvement.

cs.NI↗

arXiv2025

Structural stability of boundary layers in the entire subsonic regime

Despite the physical importance, there are limited mathematical theories for the compressible Navier-Stokes equations with strong boundary layers. This is mainly due to the absence of a stream function structure, unlike the extensively studied incompressible fluid dynamics in two dimensions. This paper aims to establish the structural stability of boundary layer profiles in the form of shear flow for the two-dimensional steady compressible Navier-Stokes equations. Our estimates are uniform across the entire subsonic regime, where the Mach number $m\in (0,1)$. As a byproduct, we provide the first result concerning the low Mach number limit in the presence of Prandtl boundary layers. The proof relies on the quasi-compressible-Stokes iteration introduced in [38], along with a subtle analysis of the interplay between density and velocity variables in different frequency regimes, and the identification of cancellations in higher-order estimates.

math.AP↗

arXiv2025

Rank Also Matters: Hierarchical Configuration for Mixture of Adapter Experts in LLM Fine-Tuning

Large language models (LLMs) have demonstrated remarkable success across various tasks, accompanied by a continuous increase in their parameter size. Parameter-efficient fine-tuning (PEFT) methods, such as Low-Rank Adaptation (LoRA), address the challenges of fine-tuning LLMs by significantly reducing the number of trainable parameters. Recent studies have integrated LoRA with Mixture of Experts (MoE) architectures, leveraging multiple adapter experts and gating mechanisms to further improve fine-tuning performance. However, existing approaches primarily focus on adjusting the allocations of adapter experts per layer to optimize the introduced trainable parameter size, while neglecting a critical factor of adapters' rank. To this end, we propose a hierarchical scheme for expert allocation and rank configuration, HILO, which dynamically adjusts the number and rank of adapter experts across layers, matching the varying representational complexity of model layers in adapter-granularity. Extensive experiments on multiple benchmark tasks demonstrate that HILO outperforms existing methods in accuracy while introducing fewer trainable parameters, providing an efficient and practical solution for fine-tuning LLMs.

cs.LG↗

arXiv2025

Predicting 3D representations for Dynamic Scenes

We present a novel framework for dynamic radiance field prediction given monocular video streams. Unlike previous methods that primarily focus on predicting future frames, our method goes a step further by generating explicit 3D representations of the dynamic scene. The framework builds on two core designs. First, we adopt an ego-centric unbounded triplane to explicitly represent the dynamic physical world. Second, we develop a 4D-aware transformer to aggregate features from monocular videos to update the triplane. Coupling these two designs enables us to train the proposed model with large-scale monocular videos in a self-supervised manner. Our model achieves top results in dynamic radiance field prediction on NVIDIA dynamic scenes, demonstrating its strong performance on 4D physical world modeling. Besides, our model shows a superior generalizability to unseen scenarios. Notably, we find that our approach emerges capabilities for geometry and semantic learning.

cs.CV↗

arXiv2025

Small polarons mediated near-room-temperature metal-insulator transition in vanadium dioxide and their hopping dynamics

Researchers pursuing advanced photoelectric devices have discovered near room-temperature metal-insulator transitions (MIT) in non-volatile VO2. Despite theoretical investigations suggesting that polaron dynamics mediate the MIT, direct experimental evidence remains scarce. In this study, we present direct evidence of the polaron state in insulating VO2 through high-resolution spectroscopic ellipsometry measurements and first-principles calculations. We illustrate the complementary role of polaron dynamics in facilitating Peierls and Mott transitions, thereby contributing to the MIT processes. Furthermore, our observations and characterizations of conventional metallic and correlated plasmons in the respective phases of the VO2 film offer valuable insights into their electron structures. This investigation enhances comprehension of the MIT mechanism in correlated systems and underscores the roles of polarons, lattice distortions, and electron correlations in facilitating phase transition processes in strongly-correlated systems. Additionally, the detailed detection of small polarons and plasmons serves as inspiration for the development of new device functionalities.

physics.app-ph↗

arXiv2025

Knudsen boundary layer equations with incoming boundary condition: full range of cutoff collision kernels and Mach numbers of the far field

This paper establishes tahe existence and uniqueness of the nonlinear Knudsen layer equation with incoming boundary conditions. It is well-known that the solvability conditions of the problem vary with the Mach number of the far Maxwellian $\mathcal{M}^\infty$. We consider full ranges of cutoff collision kernels (i.e., $- 3 < γ\leq 1$) and all the Mach numbers of the far field in the $L^\infty_{x,v}$ framework. Additionally, the solution exhibits exponential decay $\exp \{- c x^\frac{2}{3 - γ} - c |v|^2 \}$ for some $c > 0$. To address the general angular cutoff collision kernel, we introduce a $(x,v)$-mixed weight $σ$. The proof is essentially bsed on adding an artificial damping term.

math.AP↗

arXiv2024

Carbonyl4: A Sketch for Set-Increment Mixed Updates

In the realm of data stream processing, the advent of SET-INCREMENT Mixed (SIM) data streams necessitates algorithms that efficiently handle both SET and INCREMENT operations. We present Carbonyl4, an innovative algorithm designed specifically for SIM data streams, ensuring accuracy, unbiasedness, and adaptability. Carbonyl4 introduces two pioneering techniques: the Balance Bucket for refined variance optimization, and the Cascading Overflow for maintaining precision amidst overflow scenarios. Our experiments across four diverse datasets establish Carbonyl4's supremacy over existing algorithms, particularly in terms of accuracy for item-level information retrieval and adaptability to fluctuating memory requirements. The versatility of Carbonyl4 is further demonstrated through its dynamic memory shrinking capability, achieved via a re-sampling and a heuristic approach. The source codes of Carbonyl4 are available at GitHub.

cs.DS↗

arXiv2024

1D flat bands in phosphorene nanoribbons with pentagonal nature

Materials with flat bands can serve as a promising platform to investigate strongly interacting phenomena. However, experimental realization of ideal flat bands is mostly limited to artificial lattices or moiré systems. Here we report a general way to construct one-dimensional (1D) flat bands in phosphorene nanoribbons (PNRs) with pentagonal nature: penta-hexa-PNRs and penta-dodeca-PNRs, wherein the corresponding 1D flat bands are directly verified by using angle-resolved photoemission spectroscopy. We confirm that the observed 1D flat bands originate from the electronic 1D zigzag and Lieb lattices, respectively, as revealed by the combination of bond-resolved scanning tunneling microscopy, scanning tunneling spectroscopy, tight-binding models, and first-principles calculations. Our study demonstrates a general way to construct 1D flat bands in 1D solid materials system, which provides a robust platform to explore strongly interacting phases of matter.

cond-mat.mtrl-sci↗

arXiv2024

CuckooGraph: A Scalable and Space-Time Efficient Data Structure for Large-Scale Dynamic Graphs

Graphs play an increasingly important role in various big data applications. However, existing graph data structures cannot simultaneously address the performance bottlenecks caused by the dynamic updates, large scale, and high query complexity of current graphs. This paper proposes a novel data structure for large-scale dynamic graphs called CuckooGraph. It does not require any prior knowledge of the upcoming graphs, and can adaptively resize to the most memory-efficient form while requiring few memory accesses for very fast graph data processing. The key techniques of CuckooGraph include TRANSFORMATION and DENYLIST. TRANSFORMATION fully utilizes the limited memory by designing related data structures that allow flexible space transformations to smoothly expand/tighten the required space depending on the number of incoming items. DENYLIST efficiently handles item insertion failures and further improves processing speed. Our experimental results show that compared with the most competitive solution Spruce, CuckooGraph achieves about $33\times$ higher insertion throughput while requiring only about $68\%$ of the memory space.

cs.DB↗

arXiv2024

Inference-to-complete: A High-performance and Programmable Data-plane Co-processor for Neural-network-driven Traffic Analysis

Neural-networks-driven intelligent data-plane (NN-driven IDP) is becoming an emerging topic for excellent accuracy and high performance. Meanwhile we argue that NN-driven IDP should satisfy three design goals: the flexibility to support various NNs models, the low-latency-high-throughput inference performance, and the data-plane-unawareness harming no performance and functionality. Unfortunately, existing work either over-modify NNs for IDP, or insert inline pipelined accelerators into the data-plane, failing to meet the flexibility and unawareness goals. In this paper, we propose Kaleidoscope, a flexible and high-performance co-processor located at the bypass of the data-plane. To address the challenge of meeting three design goals, three key techniques are presented. The programmable run-to-completion accelerators are developed for flexible inference. To further improve performance, we design a scalable inference engine which completes low-latency and low-cost inference for the mouse flows, and perform complex NNs with high-accuracy for the elephant flows. Finally, raw-bytes-based NNs are introduced, which help to achieve unawareness. We prototype Kaleidoscope on both FPGA and ASIC library. In evaluation on six NNs models, Kaleidoscope reaches 256-352 ns inference latency and 100 Gbps throughput with negligible influence on the data-plane. The on-board tested NNs perform state-of-the-art accuracy among other NN-driven IDP, exhibiting the the significant impact of flexibility on enhancing traffic analysis accuracy.

cs.NI↗

arXiv2024

BATON: Enhancing Batch-wise Inference Efficiency for Large Language Models via Dynamic Re-batching

The advanced capabilities of Large Language Models (LLMs) have inspired the development of various interactive web services or applications, such as ChatGPT, which offer query inference services for users. Unlike traditional DNN model, the inference of LLM entails different iterations of forward computation for different queries, which result in efficiency challenges for existing run-to-completion batch-wise inference. Hence, some methods refine batch-wise inference to iteration-level by duplicating all nonlinear layers of LLM. However, this approach not only increases resource usage but also introduces idle computations to the batch due to the prefilling of newly added queries. Therefore, we propose BATON, an efficient batch-wise LLM inference scheme by dynamically adjusting processing batch, which can achieve near-zero idle computations without incurring additional resource consumption. To do so, BATON 1) shapes the vectors involved in the inference of the newly inserted query and processing batch to align dimensions and generates a new attention mask based on vector shaping to ensure inference correctness, which enables query inserting without consuming additional resource; 2) embeds prefilled Keys and Values of the new query into the KV_Cache of the processing batch by leveraging the prefilling and decoding separation mechanism, eliminating idle computations to the batch introduced by the prefilling process of the new query. Experimental results show that compared to the state-of-the-art solution Orca, BATON improves query processing by up to 1.75 times.

cs.LG↗

arXiv2024

INT-FlashAttention: Enabling Flash Attention for INT8 Quantization

As the foundation of large language models (LLMs), self-attention module faces the challenge of quadratic time and memory complexity with respect to sequence length. FlashAttention accelerates attention computation and reduces its memory usage by leveraging the GPU memory hierarchy. A promising research direction is to integrate FlashAttention with quantization methods. This paper introduces INT-FlashAttention, the first INT8 quantization architecture compatible with the forward workflow of FlashAttention, which significantly improves the inference speed of FlashAttention on Ampere GPUs. We implement our INT-FlashAttention prototype with fully INT8 activations and general matrix-multiplication (GEMM) kernels, making it the first attention operator with fully INT8 input. As a general token-level post-training quantization framework, INT-FlashAttention is also compatible with other data formats like INT4, etc. Experimental results show INT-FlashAttention achieves 72% faster inference speed and 82% smaller quantization error compared to standard FlashAttention with FP16 and FP8 data format.

cs.LG↗

arXiv2024

In-Context Learning with Representations: Contextual Generalization of Trained Transformers

In-context learning (ICL) refers to a remarkable capability of pretrained large language models, which can learn a new task given a few examples during inference. However, theoretical understanding of ICL is largely under-explored, particularly whether transformers can be trained to generalize to unseen examples in a prompt, which will require the model to acquire contextual knowledge of the prompt for generalization. This paper investigates the training dynamics of transformers by gradient descent through the lens of non-linear regression tasks. The contextual generalization here can be attained via learning the template function for each task in-context, where all template functions lie in a linear space with $m$ basis functions. We analyze the training dynamics of one-layer multi-head transformers to in-contextly predict unlabeled inputs given partially labeled prompts, where the labels contain Gaussian noise and the number of examples in each prompt are not sufficient to determine the template. Under mild assumptions, we show that the training loss for a one-layer multi-head transformer converges linearly to a global minimum. Moreover, the transformer effectively learns to perform ridge regression over the basis functions. To our knowledge, this study is the first provable demonstration that transformers can learn contextual (i.e., template) information to generalize to both unseen examples and tasks when prompts contain only a small number of query-answer pairs.

cs.LG↗

arXiv2024

Art and Science of Quantizing Large-Scale Models: A Comprehensive Overview

This paper provides a comprehensive overview of the principles, challenges, and methodologies associated with quantizing large-scale neural network models. As neural networks have evolved towards larger and more complex architectures to address increasingly sophisticated tasks, the computational and energy costs have escalated significantly. We explore the necessity and impact of model size growth, highlighting the performance benefits as well as the computational challenges and environmental considerations. The core focus is on model quantization as a fundamental approach to mitigate these challenges by reducing model size and improving efficiency without substantially compromising accuracy. We delve into various quantization techniques, including both post-training quantization (PTQ) and quantization-aware training (QAT), and analyze several state-of-the-art algorithms such as LLM-QAT, PEQA(L4Q), ZeroQuant, SmoothQuant, and others. Through comparative analysis, we examine how these methods address issues like outliers, importance weighting, and activation quantization, ultimately contributing to more sustainable and accessible deployment of large-scale models.

cs.LG↗

arXiv2024

The spatially inhomogeneous Vlasov-Nordström-Fokker-Planck system in the intrinsic weak diffusion regime

The spatially homogeneous Vlasov-Nordström-Fokker-Planck system is known to exhibit nontrivial large time behavior, naturally leading to weak diffusion of the Fokker-Planck operator. This weak diffusion, combined with the singularity of relativistic velocity, present a significant challenge in analysis for the spatially inhomogeneous counterpart. In this paper, we demonstrate that the Cauchy problem for the spatially inhomogeneous Vlasov-Nordström-Fokker-Planck system, without friction, maintains dynamically stable relative to the corresponding spatially homogeneous system. Our results are twofold: (1) we establish the existence of a unique global classical solution and characterize the asymptotic behavior of the spatially inhomogeneous system using a refined weighted energy method; (2) we directly verify the dynamic stability of the spatially inhomogeneous system in the framework of self-similar solutions.

math.AP↗

arXiv2024

QET: Enhancing Quantized LLM Parameters and KV cache Compression through Element Substitution and Residual Clustering

The matrix quantization entails representing matrix elements in a more space-efficient form to reduce storage usage, with dequantization restoring the original matrix for use. We formulate the Quantization Error Minimization (QEM) problem as minimizing the distance between a matrix before and after quantization, under the condition that the quantized matrix occupies the same memory space. Matrix quantization is crucial in various applications, including Large Language Models (LLMs) weight quantization, vector databases, KV cache quantization, graph compression, and image compression. Recent advancements in LLMs, such as GPT-4 and BERT, have highlighted the importance of matrix compression due to the large size of parameters and KV cache, which are stored as matrices. We propose Quantum Entanglement Trees (QET) to address the QEM problem by leveraging the local orderliness of matrix elements, involving iterative element swapping to form a locally ordered matrix. This matrix is then grouped and quantized by columns. To enhance QET, we introduce two optimizations: further quantizing residuals to reduce MSE, and using masking and batch processing to accelerate the algorithm. Experimental results demonstrate that QET can effectively reduce MSE to 5.05%, 13.33%, and 11.89% of the current best method on the LLM dataset, K cache, and V cache, respectively. Our contributions include the abstraction of the QEM problem, the design of the QET algorithm, and the proposal of two optimizations to improve accuracy and speed.

cs.LG↗