Author name cluster

Yanjun Wu

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

8 papers

1 author row

AAAI Conference 2026 Conference Paper

QiMeng-Kernel: Macro-Thinking Micro-Coding Paradigm for LLM-Based High-Performance GPU Kernel Generation

Xinguo Zhu
Shaohui Peng
Jiaming Guo
Yunji Chen
Qi Guo
Yuanbo Wen
Hang Qin
Ruizhi Chen

Developing high-performance GPU kernels is critical for AI and scientific computing, but remains challenging due to its reliance on expert crafting and poor portability. While large language models (LLMs) offer promise for automation, both general-purpose and finetuned LLMs suffer from two fundamental and conflicting limitations: correctness and efficiency. The key reason is that existing LLM-based approaches directly generate the entire optimized low-level programs, requiring exploration of an extremely vast space encompassing both optimization policies and implementation codes. To address the challenge of exploring an intractable space, we propose Macro Thinking Micro Coding (MTMC), a hierarchical framework inspired by the staged optimization strategy of human experts. It decouples optimization strategy from implementation details, ensuring efficiency through high-level strategy and correctness through low-level implementation. Specifically, Macro Thinking employs reinforcement learning to guide lightweight LLMs in efficiently exploring and learning semantic optimization strategies that maximize hardware utilization. Micro Coding leverages general-purpose LLMs to incrementally implement the stepwise optimization proposals from Macro Thinking, avoiding full-kernel generation errors. Together, they effectively navigate the vast optimization space and intricate implementation details, enabling LLMs for high-performance GPU kernel generation. Comprehensive results on widely adopted benchmarks demonstrate the superior performance of MTMC on GPU kernel generation in both accuracy and running time. On KernelBench, MTMC achieves near 100% and 70% accuracy at Levels 1-2 and 3, over 50% than SOTA general-purpose and domain-finetuned LLMs, with up to 7.3× speedup over LLMs, and 2.2× over expert-optimized PyTorch Eager kernels. On the more challenging TritonBench, MTMC attains up to 59.64% accuracy and 34× speedup. All models and datasets will be made publicly available.

PDF Details DOI

NeurIPS Conference 2025 Conference Paper

Compiler-R1: Towards Agentic Compiler Auto-tuning with Reinforcement Learning

Haolin Pan
Hongyu Lin
Haoran Luo
Yang Liu
Kaichun Yao
Libo Zhang
Mingjie Xing
Yanjun Wu

Compiler auto-tuning optimizes pass sequences to improve performance metrics such as Intermediate Representation (IR) instruction count. Although recent advances leveraging Large Language Models (LLMs) have shown promise in automating compiler tuning, two significant challenges still remain: the absence of high-quality reasoning datasets for agents training, and limited effective interactions with the compilation environment. In this work, we introduce Compiler-R1, the first reinforcement learning (RL)-driven framework specifically augmenting LLM capabilities for compiler auto-tuning. Compiler-R1 features a curated, high-quality reasoning dataset and a novel two-stage end-to-end RL training pipeline, enabling efficient environment exploration and learning through an outcome-based reward. Extensive experiments across seven datasets demonstrate Compiler-R1 achieving an average 8. 46\% IR instruction count reduction compared to opt -Oz, showcasing the strong potential of RL-trained LLMs for compiler optimization. Our code and datasets are publicly available at https: //github. com/Panhaolin2001/Compiler-R1.

PDF Details

NeurIPS Conference 2025 Conference Paper

EasySpec: Layer-Parallel Speculative Decoding for Efficient Multi-GPU Utilization

Yize Wu
Ke Gao
Ling Li
Yanjun Wu

Speculative decoding is an effective and lossless method for Large Language Model (LLM) inference acceleration. It employs a smaller model to generate a draft token sequence, which is then verified by the original base model. In multi-GPU systems, inference latency can be further reduced through tensor parallelism (TP), while the optimal TP size of the draft model is typically smaller than that of the base model, leading to GPU idling during the drafting stage. We observe that such inefficiency stems from the sequential execution of layers, which is seemingly natural but actually unnecessary. Therefore, we propose EasySpec, a layer-parallel speculation strategy that optimizes the efficiency of multi-GPU utilization. EasySpec breaks the inter-layer data dependencies in the draft model, enabling multiple layers to run simultaneously across multiple devices as ``fuzzy'' speculation. After each drafting-and-verification iteration, the draft model’s key-value cache is calibrated in a single forward pass, preventing long-term fuzzy-error accumulation at minimal additional latency. EasySpec is a training-free and plug-in method. We evaluated EasySpec on several mainstream open-source LLMs, using smaller versions of models from the same series as drafters. The results demonstrate that EasySpec can achieve a peak speedup of 4. 17x compared to vanilla decoding, while preserving the original distributions of the base LLMs. Specifically, the drafting stage can be accelerated by up to 1. 62x with a maximum speculation accuracy drop of only 7\%. The code is available at https: //github. com/Yize-Wu/EasySpec.

PDF Details

AAAI Conference 2025 Conference Paper

QiMeng-GEMM: Automatically Generating High-Performance Matrix Multiplication Code by Exploiting Large Language Models

Qirui Zhou
Yuanbo Wen
Ruizhi Chen
Ke Gao
Weiqiang Xiong
Ling Li
Qi Guo
Yanjun Wu

As a crucial operator in numerous scientific and engineering computing applications, the automatic optimization of General Matrix Multiplication (GEMM) with full utilization of ever-evolving hardware architectures (e.g. GPUs and RISC-V) is of paramount importance. While Large Language Models (LLMs) can generate functionally correct code for simple tasks, they have yet to produce high-performance code. The key challenge resides in deeply understanding diverse hardware architectures and crafting prompts that effectively unleash the potential of LLMs to generate high-performance code. In this paper, we propose a novel prompt mechanism called QiMeng-GEMM which enables LLMs to comprehend the architectural characteristics of different hardware platforms and automatically search for the optimization combinations for GEMM. The key of QiMeng-GEMM is a set of informative, adaptive, and iterative meta-prompts. Based on this, a searching strategy for optimal combinations of meta-prompts is used to iteratively generate high-performance code. Extensive experiments conducted on 4 leading LLMs, various paradigmatic hardware platforms, and representative matrix dimensions unequivocally demonstrate QiMeng-GEMM’s superior performance in auto-generating optimized GEMM code. Compared to vanilla prompts, our method achieves a performance enhancement of up to 113×. Even when compared to human experts, our method can reach 115% of cuBLAS on NVIDIA GPUs and 211% of OpenBLAS on RISC-V CPUs. Notably, while human experts often take months to optimize GEMM, our approach reduces the development cost by over 240×.

PDF Details DOI

IJCAI Conference 2025 Conference Paper

QiMeng-TensorOp: One-Line Prompt is Enough for High-Performance Tensor Operator Generation with Hardware Primitives

Xuzhi Zhang
Shaohui Peng
Qirui Zhou
Yuanbo Wen
Qi Guo
Ruizhi Chen
Xinguo Zhu
Weiqiang Xiong

Computation-intensive tensor operators constitute over 90% of the computations in Large Language Models (LLMs) and Deep Neural Networks. Automatically and efficiently generating high-performance tensor operators with hardware primitives is crucial for diverse and ever-evolving hardware architectures like RISC-V, ARM, and GPUs, as manually optimized implementation takes at least months and lacks portability. LLMs excel at generating high-level language codes, but they struggle to fully comprehend hardware characteristics and produce high-performance tensor operators. We introduce a tensor-operator auto-generation framework with a one-line user prompt (QiMeng-TensorOp), which enables LLMs to automatically exploit hardware characteristics to generate tensor operators with hardware primitives, and tune parameters for optimal performance across diverse hardware. Experimental results on various hardware platforms, SOTA LLMs, and typical tensor operators demonstrate that QiMeng-TensorOp effectively unleashes the computing capability of various hardware platforms, and automatically generates tensor operators of superior performance. Compared with vanilla LLMs, QiMeng-TensorOp achieves up to 1291× performance improvement. Even compared with human experts, QiMeng-TensorOp could reach 251% of OpenBLAS on RISC-V CPUs, and 124% of cuBLAS on NVIDIA GPUs. Additionally, QiMeng-TensorOp also significantly reduces development costs by 200× compared with human experts.

PDF Details DOI

AAAI Conference 2019 Conference Paper

Scale Invariant Fully Convolutional Network: Detecting Hands Efficiently

Dan Liu
Dawei Du
Libo Zhang
Tiejian Luo
Yanjun Wu
Feiyue Huang
Siwei Lyu

Existing hand detection methods usually follow the pipeline of multiple stages with high computation cost, i. e. , feature extraction, region proposal, bounding box regression, and additional layers for rotated region detection. In this paper, we propose a new Scale Invariant Fully Convolutional Network (SIFCN) trained in an end-to-end fashion to detect hands efficiently. Specifically, we merge the feature maps from high to low layers in an iterative way, which handles different scales of hands better with less time overhead comparing to concatenating them simply. Moreover, we develop the Complementary Weighted Fusion (CWF) block to make full use of the distinctive features among multiple layers to achieve scale invariance. To deal with rotated hand detection, we present the rotation map to get rid of complex rotation and derotation layers. Besides, we design the multi-scale loss scheme to accelerate the training process significantly by adding supervision to the intermediate layers of the network. Compared with the state-of-the-art methods, our algorithm shows comparable accuracy and runs a 4. 23 times faster speed on the VIVA dataset and achieves better average precision on Oxford hand detection dataset at a speed of 62. 5 fps.

PDF Details

IJCAI Conference 2019 Conference Paper

VulSniper: Focus Your Attention to Shoot Fine-Grained Vulnerabilities

Xu Duan
Jingzheng Wu
Shouling Ji
Zhiqing Rui
Tianyue Luo
Mutian Yang
Yanjun Wu

With the explosive development of information technology, vulnerabilities have become one of the major threats to computer security. Most vulnerabilities with similar patterns can be detected effectively by static analysis methods. However, some vulnerable and non-vulnerable code is hardly distinguishable, resulting in low detection accuracy. In this paper, we define the accurate identification of vulnerabilities in similar code as a fine-grained vulnerability detection problem. We propose VulSniper which is designed to detect fine-grained vulnerabilities more effectively. In VulSniper, attention mechanism is used to capture the critical features of the vulnerabilities. Especially, we use bottom-up and top-down structures to learn the attention weights of different areas of the program. Moreover, in order to fully extract the semantic features of the program, we generate the code property graph, design a 144-dimensional vector to describe the relation between the nodes, and finally encode the program as a feature tensor. VulSniper achieves F1-scores of 80. 6% and 73. 3% on the two benchmark datasets, the SARD Buffer Error dataset and the SARD Resource Management Error dataset respectively, which are significantly higher than those of the state-of-the-art methods.

PDF Details

IJCAI Conference 2018 Conference Paper

Teaching Machines to Ask Questions

Kaichun Yao
Libo Zhang
Tiejian Luo
Lili Tao
Yanjun Wu

We propose a novel neural network model that aims to generate diverse and human-like natural language questions. Our model not only directly captures the variability in possible questions by using a latent variable, but also generates certain types of questions by introducing an additional observed variable. We deploy our model in the generative adversarial network (GAN) framework and modify the discriminator which not only allows evaluating the question authenticity, but predicts the question type. Our model is trained and evaluated on a question-answering dataset SQuAD, and the experimental results shown the proposed model is able to generate diverse and readable questions with the specific attribute.

PDF Details