arXiv subjects

Zhen Jia

Publications and source records attributed to Zhen Jia.

At least 19 recordsLinked to original sources

arXiv2025

AccelOpt: A Self-Improving LLM Agentic System for AI Accelerator Kernel Optimization

We present AccelOpt, a self-improving large language model (LLM) agentic system that autonomously optimizes kernels for emerging AI acclerators, eliminating the need for expert-provided hardware-specific optimization knowledge. AccelOpt explores the kernel optimization space through iterative generation, informed by an optimization memory that curates experiences and insights from previously encountered slow-fast kernel pairs. We build NKIBench, a new benchmark suite of AWS Trainium accelerator kernels with varying complexity extracted from real-world LLM workloads to evaluate the effectiveness of AccelOpt. Our evaluation confirms that AccelOpt's capability improves over time, boosting the average percentage of peak throughput from $49\%$ to $61\%$ on Trainium 1 and from $45\%$ to $59\%$ on Trainium 2 for NKIBench kernels. Moreover, AccelOpt is highly cost-effective: using open-source models, it matches the kernel improvements of Claude Sonnet 4 while being $26\times$ cheaper. The code is open-sourced at https://github.com/zhang677/AccelOpt.

cs.LG↗

arXiv2025

SMoRFFI: A Large-Scale Same-Model 2.4 GHz Wi-Fi Dataset and Reproducible Framework for RF Fingerprinting

Radio frequency (RF) fingerprinting exploits hardware imperfections for device identification, but distinguishing between same-model devices remains challenging due to their minimal hardware variations. Existing datasets for RF fingerprinting are constrained by small device scales and heterogeneous models, which hinder robust training and fair evaluation of machine learning methods. To address this gap, we introduce a large-scale dataset of same-model devices along with an open-source experimental framework. The dataset is built using 123 same-model commercial IEEE 802.11g devices, which contain 35.42 million raw I/Q samples from the preambles and corresponding 1.85 million RF features. The accompanying framework further provides a fully reproducible pipeline from data collection to performance evaluation. Within this framework, a Random Forest-based algorithm is implemented as a baseline to achieve 89.06% identification accuracy on this dataset.

cs.NI↗

arXiv2025

Traceable Drug Recommendation over Medical Knowledge Graphs

Drug recommendation (DR) systems aim to support healthcare professionals in selecting appropriate medications based on patients' medical conditions. State-of-the-art approaches utilize deep learning techniques for improving DR, but fall short in providing any insights on the derivation process of recommendations -- a critical limitation in such high-stake applications. We propose TraceDR, a novel DR system operating over a medical knowledge graph (MKG), which ensures access to large-scale and high-quality information. TraceDR simultaneously predicts drug recommendations and related evidence within a multi-task learning framework, enabling traceability of medication recommendations. For covering a more diverse set of diseases and drugs than existing works, we devise a framework for automatically constructing patient health records and release DrugRec, a new large-scale testbed for DR.

cs.IR↗

arXiv2025

Not-a-Bandit: Provably No-Regret Drafter Selection in Speculative Decoding for LLMs

Speculative decoding is widely used in accelerating large language model (LLM) inference. In this work, we focus on the online draft model selection problem in speculative decoding. We design an algorithm that provably competes with the best draft model in hindsight for each query in terms of either the token acceptance probability or expected acceptance length. In particular, we show that we can accurately evaluate all draft models, instead of only the chosen model without incurring additional queries to the target model, which allows us to improve exponentially over the existing bandit-based approach as the number of draft models increases. Our approach is generically applicable with any speculative decoding methods (single draft, multi-drafts and draft-trees). Moreover, we design system-efficient versions of online learners and demonstrate that the overhead in computation and latency can be substantially reduced. We conduct extensive experiments on open-source LLMs and diverse datasets, demonstrating that our methods substantially outperform the state-of-the-art EAGLE3 and the BanditSpec baseline in a variety of domains where specialized domain-expert drafters are available, especially when long reasoning chains are required.

cs.LG↗

arXiv2025

TritonRL: Training LLMs to Think and Code Triton Without Cheating

The rapid evolution of Large Language Models (LLMs) has driven a growing demand for automated, high-performance system kernels to accelerate machine learning workloads. We introduce TritonRL, a domain-specialized 8B-scale LLM for Triton programming, trained via a novel reinforcement learning (RL) framework. While Triton synthesis faces unique challenges, including data scarcity and a high susceptibility to reward hacking, our approach enables robust kernel generation through two primary innovations. First, we implement a multi-layered verification system that provides high-fidelity reward signals, ensuring that generated kernels are both syntactically and functionally valid. Second, we propose Hierarchical Reward Decomposition (HRD), which decouples reinforcement for high-level reasoning and low-level implementation to resolve the credit assignment problem in long-sequence generation. Comprehensive evaluations on KernelBench demonstrate that TritonRL achieves state-of-the-art correctness and runtime speedup, outperforming concurrent Triton-specific models and matching the performance of frontier models with over 100B parameters. Our results highlight the effectiveness of hardware-aware RL paradigms in specialized domain adaptation.

cs.SE↗

arXiv2025

DCP: Addressing Input Dynamism In Long-Context Training via Dynamic Context Parallelism

Context parallelism has emerged as a key technique to support long-context training, a growing trend in generative AI for modern large models. However, existing context parallel methods rely on static parallelization configurations that overlook the dynamic nature of training data, specifically, the variability in sequence lengths and token relationships (i.e., attention patterns) across samples. As a result, these methods often suffer from unnecessary communication overhead and imbalanced computation. In this paper, we present DCP, a dynamic context parallel training framework that introduces fine-grained blockwise partitioning of both data and computation. By enabling flexible mapping of data and computation blocks to devices, DCP can adapt to varying sequence characteristics, effectively reducing communication and improving memory and computation balance. Micro-benchmarks demonstrate that DCP accelerates attention by 1.19x~2.45x under causal masks and 2.15x~3.77x under sparse attention patterns. Additionally, we observe up to 0.94x~1.16x end-to-end training speed-up for causal masks, and 1.00x~1.46x for sparse masks.

cs.DC↗

arXiv2025

PLoRA: Efficient Concurrent LoRA Training for Large Language Models

Low-Rank Adaptation (LoRA) has gained popularity as a fine-tuning approach for Large Language Models (LLMs) due to its low resource requirements and good performance. While numerous studies have investigated ways to improve LoRA serving efficiency by serving multiple LoRAs concurrently, existing methods assume that a wide range of LoRA adapters are available for serving. In our work, we conduct extensive empirical studies to show that current LoRA training paradigms do not efficiently utilize hardware resources and incur high overhead to obtain a performant LoRA adapter. Leveraging these insights, we propose PLoRA, which automatically orchestrates concurrent LoRA fine-tuning jobs under given hardware and model constraints and develops performant kernels to improve training efficiency. Across a range of LLMs and LoRA configurations, PLoRA improves training throughput by up to 12.8x and reduces the overall fine-tuning makespan by up to 7.52x compared to existing approaches.

cs.LG↗

arXiv2025

ProxSparse: Regularized Learning of Semi-Structured Sparsity Masks for Pretrained LLMs

Large Language Models (LLMs) have demonstrated exceptional performance in natural language processing tasks, yet their massive size makes serving them inefficient and costly. Semi-structured pruning has emerged as an effective method for model acceleration, but existing approaches are suboptimal because they focus on local, layer-wise optimizations using heuristic rules, failing to leverage global feedback. We present ProxSparse, a learning-based framework for mask selection enabled by regularized optimization. ProxSparse transforms the rigid, non-differentiable mask selection process into a smoother optimization procedure, allowing gradual mask exploration with flexibility. ProxSparse does not involve additional weight updates once the mask is determined. Our extensive evaluations on 7 widely used models show that ProxSparse consistently outperforms previously proposed semi-structured mask selection methods with significant improvement, demonstrating the effectiveness of our learned approach towards semi-structured pruning.

cs.LG↗

arXiv2025

TTrace: Lightweight Error Checking and Diagnosis for Distributed Training

Distributed training is essential for scaling the training of large neural network models, such as large language models (LLMs), across thousands of GPUs. However, the complexity of distributed training programs makes them particularly prone to silent bugs, which do not produce explicit error signals but lead to incorrect training outcomes. Effectively detecting and localizing such silent bugs in distributed training is challenging. Common debugging practices based on monitoring training loss or gradient norm curves are indirect, inefficient, and provide no way to localize bugs. To address those challenges, we design and implement TTrace, the first systematic differential testing system for detecting and localizing silent bugs in distributed training. TTrace aligns intermediate tensors from distributed training with those from a trusted reference implementation. To properly compare the floating-point values in the corresponding tensors, we propose a novel mathematical analysis that provides a guideline for setting tolerances, enabling TTrace to distinguish bug-induced errors from numerical errors. Experimental results demonstrate that TTrace effectively detects 11 existing bugs and 3 new bugs in the widely used Megatron-LM framework, while requiring fewer than 10 lines of code changes. TTrace is effective in various training recipes, including low-precision recipes involving BF16 and FP8. Notably, a popular open-source training framework has already adopted the method proposed by TTrace in its development workflow.

cs.DC↗

arXiv2025

Marconi: Prefix Caching for the Era of Hybrid LLMs

Hybrid models that combine the language modeling capabilities of Attention layers with the efficiency of Recurrent layers (e.g., State Space Models) have gained traction in practically supporting long contexts in Large Language Model serving. Yet, the unique properties of these models complicate the usage of complementary efficiency optimizations such as prefix caching that skip redundant computations across requests. Most notably, their use of in-place state updates for recurrent layers precludes rolling back cache entries for partial sequence overlaps, and instead mandates only exact-match cache hits; the effect is a deluge of (large) cache entries per sequence, most of which yield minimal reuse opportunities. We present Marconi, the first system that supports efficient prefix caching with Hybrid LLMs. Key to Marconi are its novel admission and eviction policies that more judiciously assess potential cache entries based not only on recency, but also on (1) forecasts of their reuse likelihood across a taxonomy of different hit scenarios, and (2) the compute savings that hits deliver relative to memory footprints. Across diverse workloads and Hybrid models, Marconi achieves up to 34.4$\times$ higher token hit rates (71.1% or 617 ms lower TTFT) compared to state-of-the-art prefix caching systems.

cs.DC↗

arXiv2025

Extrapolated Urban View Synthesis Benchmark

Photorealistic simulators are essential for the training and evaluation of vision-centric autonomous vehicles (AVs). At their core is Novel View Synthesis (NVS), a crucial capability that generates diverse unseen viewpoints to accommodate the broad and continuous pose distribution of AVs. Recent advances in radiance fields, such as 3D Gaussian Splatting, achieve photorealistic rendering at real-time speeds and have been widely used in modeling large-scale driving scenes. However, their performance is commonly evaluated using an interpolated setup with highly correlated training and test views. In contrast, extrapolation, where test views largely deviate from training views, remains underexplored, limiting progress in generalizable simulation technology. To address this gap, we leverage publicly available AV datasets with multiple traversals, multiple vehicles, and multiple cameras to build the first Extrapolated Urban View Synthesis (EUVS) benchmark. Meanwhile, we conduct both quantitative and qualitative evaluations of state-of-the-art NVS methods across different evaluation settings. Our results show that current NVS methods are prone to overfitting to training views. Besides, incorporating diffusion priors and improving geometry cannot fundamentally improve NVS under large view changes, highlighting the need for more robust approaches and large-scale training. We will release the data to help advance self-driving and urban robotics simulation technology.

cs.CV↗

arXiv2024

Non-uniform Cross-intersecting Families

Let $m\geq 2$, $n$ be positive integers, and $R_i=\{k_{i,1} >k_{i,2} >\cdots> k_{i,t_i}\}$ be subsets of $[n]$ for $i=1,2,\ldots,m$. The families $\mathcal{F}_1\subseteq \binom{[n]}{R_1},\mathcal{F}_2\subseteq \binom{[n]}{R_2},\ldots,\mathcal{F}_m\subseteq \binom{[n]}{R_m}$ are said to be non-empty cross-intersecting if for each $i\in [m]$, $\mathcal{F}_i\neq\emptyset$ and for any $A\in \mathcal{F}_i,B\in\mathcal{F}_j$, $1\leq i<j\leq m$, $|A\bigcap B|\geq1$. In this paper, we determine the maximum value of $\sum_{j=1}^{m}|\mathcal{F}_j|$ for non-empty cross-intersecting family $\mathcal{F}_1, \mathcal{F}_2,\ldots,\mathcal{F}_m$ when $n\geq k_1+k_2$, where $k_1$ (respectively, $k_2$) is the largest (respectively, second largest) value in $\{k_{1,1},k_{2,1},\ldots,k_{m,1}\}$. This result is a generalization of the results by Shi, Frankl and Qian \cite{shi2022non} on non-empty cross-intersecting families. Moreover, the extremal families are completely characterized.

math.CO↗

arXiv2024

Human Image Generation: A Comprehensive Survey

Image and video synthesis has become a blooming topic in computer vision and machine learning communities along with the developments of deep generative models, due to its great academic and application value. Many researchers have been devoted to synthesizing high-fidelity human images as one of the most commonly seen object categories in daily lives, where a large number of studies are performed based on various models, task settings and applications. Thus, it is necessary to give a comprehensive overview on these variant methods on human image generation. In this paper, we divide human image generation techniques into three paradigms, i.e., data-driven methods, knowledge-guided methods and hybrid methods. For each paradigm, the most representative models and the corresponding variants are presented, where the advantages and characteristics of different methods are summarized in terms of model architectures. Besides, the main public human image datasets and evaluation metrics in the literature are summarized. Furthermore, due to the wide application potentials, the typical downstream usages of synthesized human images are covered. Finally, the challenges and potential opportunities of human image generation are discussed to shed light on future research.

cs.CV↗

arXiv2024

Uncertainty-aware self-training with expectation maximization basis transformation

Self-training is a powerful approach to deep learning. The key process is to find a pseudo-label for modeling. However, previous self-training algorithms suffer from the over-confidence issue brought by the hard labels, even some confidence-related regularizers cannot comprehensively catch the uncertainty. Therefore, we propose a new self-training framework to combine uncertainty information of both model and dataset. Specifically, we propose to use Expectation-Maximization (EM) to smooth the labels and comprehensively estimate the uncertainty information. We further design a basis extraction network to estimate the initial basis from the dataset. The obtained basis with uncertainty can be filtered based on uncertainty information. It can then be transformed into the real hard label to iteratively update the model and basis in the retraining process. Experiments on image classification and semantic segmentation show the advantages of our methods among confidence-aware self-training algorithms with 1-3 percentage improvement on different datasets.

cs.CV↗

arXiv2024

Potential Energy based Mixture Model for Noisy Label Learning

Training deep neural networks (DNNs) from noisy labels is an important and challenging task. However, most existing approaches focus on the corrupted labels and ignore the importance of inherent data structure. To bridge the gap between noisy labels and data, inspired by the concept of potential energy in physics, we propose a novel Potential Energy based Mixture Model (PEMM) for noise-labels learning. We innovate a distance-based classifier with the potential energy regularization on its class centers. Embedding our proposed classifier with existing deep learning backbones, we can have robust networks with better feature representations. They can preserve intrinsic structures from the data, resulting in a superior noisy tolerance. We conducted extensive experiments to analyze the efficiency of our proposed model on several real-world datasets. Quantitative results show that it can achieve state-of-the-art performance.

cs.LG↗

arXiv2024

DiffusionPipe: Training Large Diffusion Models with Efficient Pipelines

Diffusion models have emerged as dominant performers for image generation. To support training large diffusion models, this paper studies pipeline parallel training of diffusion models and proposes DiffusionPipe, a synchronous pipeline training system that advocates innovative pipeline bubble filling technique, catering to structural characteristics of diffusion models. State-of-the-art diffusion models typically include trainable (the backbone) and non-trainable (e.g., frozen input encoders) parts. We first unify optimal stage partitioning and pipeline scheduling of single and multiple backbones in representative diffusion models with a dynamic programming approach. We then propose to fill the computation of non-trainable model parts into idle periods of the pipeline training of the backbones by an efficient greedy algorithm, thus achieving high training throughput. Extensive experiments show that DiffusionPipe can achieve up to 1.41x speedup over pipeline parallel methods and 1.28x speedup over data parallel training on popular diffusion models.

cs.DC↗

arXiv2024

Lancet: Accelerating Mixture-of-Experts Training via Whole Graph Computation-Communication Overlapping

The Mixture-of-Expert (MoE) technique plays a crucial role in expanding the size of DNN model parameters. However, it faces the challenge of extended all-to-all communication latency during the training process. Existing methods attempt to mitigate this issue by overlapping all-to-all with expert computation. Yet, these methods frequently fall short of achieving sufficient overlap, consequently restricting the potential for performance enhancements. In our study, we extend the scope of this challenge by considering overlap at the broader training graph level. During the forward pass, we enable non-MoE computations to overlap with all-to-all through careful partitioning and pipelining. In the backward pass, we achieve overlap with all-to-all by scheduling gradient weight computations. We implement these techniques in Lancet, a system using compiler-based optimization to automatically enhance MoE model training. Our extensive evaluation reveals that Lancet significantly reduces the time devoted to non-overlapping communication, by as much as 77%. Moreover, it achieves a notable end-to-end speedup of up to 1.3 times when compared to the state-of-the-art solutions.

cs.DC↗

arXiv2024

Faithful Temporal Question Answering over Heterogeneous Sources

Temporal question answering (QA) involves time constraints, with phrases such as "... in 2019" or "... before COVID". In the former, time is an explicit condition, in the latter it is implicit. State-of-the-art methods have limitations along three dimensions. First, with neural inference, time constraints are merely soft-matched, giving room to invalid or inexplicable answers. Second, questions with implicit time are poorly supported. Third, answers come from a single source: either a knowledge base (KB) or a text corpus. We propose a temporal QA system that addresses these shortcomings. First, it enforces temporal constraints for faithful answering with tangible evidence. Second, it properly handles implicit questions. Third, it operates over heterogeneous sources, covering KB, text and web tables in a unified manner. The method has three stages: (i) understanding the question and its temporal conditions, (ii) retrieving evidence from all sources, and (iii) faithfully answering the question. As implicit questions are sparse in prior benchmarks, we introduce a principled method for generating diverse questions. Experiments show superior performance over a suite of baselines.

cs.IR↗