arXiv subjects

Zili Shao

Publications and source records attributed to Zili Shao.

8 recordsLinked to original sources

arXiv2026

Kernel Foundry: A Diagnosis-driven Evolutionary Kernel Optimizer with Multi-Experts

Generating high-performance GPU kernels remains challenging due to the need for both correctness and hardware-aware optimization. While large language models (LLMs) show promise in code generation, they often fail to produce kernels that are both correct and efficient. We propose Kernel Foundry, a diagnosis-driven evolutionary framework for automatic GPU kernel optimization. Our method combines expert-guided, retrieval-augmented initialization with a multi-island evolutionary search, where candidate kernels are iteratively refined using structured diagnostic feedback. A centralized experience library accumulates reusable optimization knowledge to guide subsequent evolution, while explicit mechanisms prevent cheating behaviors that bypass kernel-level computation. Experiments on KernelBench show that our method consistently improves both correctness and performance over strong baselines, achieving up to 100% correctness on Level~2.

cs.NE↗

arXiv2026

LH-AVLN: A Benchmark for Long-Horizon Audio-Visual-Language Navigation

Embodied navigation is moving toward long-horizon missions, yet existing long-horizon benchmarks are largely acoustically silent, and audio-visual navigation tasks typically focus on a single goal. We introduce LH-AVLN, a benchmark for Long-Horizon Audio-Visual-Language Navigation that combines multi-goal mission execution, heterogeneous goal specifications, and persistent spatialized acoustic cues. In LH-AVLN, an agent receives a global mission of two to four goals specified by category, language description, or reference image, and navigates with RGB-D observations, pose, and binaural audio in indoor 3D environments. The benchmark supports both ordered and unordered missions, where alternating goal-associated sounds can guide non-line-of-sight search but may also become distractors as mission progress changes. We further develop PAG-Nav, a training-free reference agent that maintains a temporal uniform semantic map and performs progressive goal-state planning, using sound for search while reserving completion for visual-semantic verification. Experiments show that existing vision-language, memory-based, and audio-visual agents struggle to complete full LH-AVLN missions, and that PAG-Nav provides a stronger diagnostic baseline while leaving substantial room for future progress.

cs.RO↗

arXiv2026

TF-Engram: A Train-Free Engram with SSD-Backed Memory for Large Language Models

Large Language Models (LLMs) store factual knowledge and domain-specific patterns implicitly in dense Transformer parameters, making knowledge expansion costly through pretraining, fine-tuning, retrieval augmentation, or longer contexts. Engram-style memory offers a compact hidden-state injection pathway, but existing GPU-resident designs often rely on hash-based compression, causing unrelated phrases to collide in shared slots and weakening phrase-level semantic fidelity. We present TF-Engram, a train-free Engram system that constructs phrase-specific semantic memory offline from external corpora, stores large memory tables across a GPU--DRAM--SSD hierarchy, and uses Early-Exit Guided Predictive Prefetching to hide external-memory latency during autoregressive decoding. On Qwen3-0.6B, TF-Engram improves the average downstream score from 57.6 to 59.4, outperforming both the frozen backbone and a parameter-matched LoRA baseline. System evaluation shows that large TF-Engram tables can be built with moderate offline cost, SSD-backed storage substantially reduces GPU memory demand, and predictive prefetching recovers much of the throughput loss caused by external memory access. These results demonstrate that static phrase memory can be integrated into LLM inference as a scalable, train-free, and low-overhead system component.

cs.CL↗

arXiv2026

SAM-Sode: Towards Faithful Explanations for Tiny Bacteria Detection

Interpretability in object detection provides crucial confidence support for clinical auxiliary diagnosis. However, in tiny bacteria detection, traditional explanation methods often suffer from blurred foreground boundaries and diffuse feature attribution due to the extreme sparsity of target morphological features and severe interference from complex backgrounds. Such limitations hinder the provision of logically coherent morphological evidence. To bridge this gap, we propose a novel eXplainable AI (XAI) framework, SAM-Sode. The framework innovatively transforms initial feature attribution maps into geometry-aware prompts, leveraging the prior knowledge of the foundation model (SAM3) to achieve spatial refinement and morphological reconstruction of the explanatory mappings. Furthermore, we introduce a dual-constraint mechanism based on physical significance and geometric alignment to perform instance-level denoising, generating coherent explanations that better align with human expert intuition. Experimental results on our self-constructed bacteria dataset with complex circuit backgrounds (containing 2,524 images) and other public datasets demonstrate that the proposed method effectively suppresses background redundancy and significantly enhances the decision-making transparency of tiny object detection.

cs.CV↗

arXiv2025

Tangram: Accelerating Serverless LLM Loading through GPU Memory Reuse and Affinity

Serverless Large Language Models (LLMs) have emerged as a cost-effective solution for deploying AI services by enabling a 'pay-as-you-go' pricing model through GPU resource sharing. However, cold-start latency, especially the model loading phase, has become a critical performance bottleneck, as it scales linearly with model size and severely limits the practical deployment of large-scale LLM services. This paper presents Tangram, a novel system that accelerates Serverless LLM loading through efficient GPU memory reuse. By leveraging the unused GPU memory to retain model parameters, Tangram significantly reduces model transfer time and cold-start latency. Its design includes three key components: unified GPU memory pool for tensor-level parameter sharing across models, on-demand KV cache allocation for dynamic memory management, and GPU-affinity-aware scheduling for maximizing resource utilization. These techniques collectively address the critical challenges of inefficient memory usage and the cold-start problem in Serverless LLM platforms. We have implemented a fully functional prototype, and experiments show that Tangram achieves up to 6.2 times faster loading and reduces Time-To-First-Token (TTFT) during cold-start by 23--55% over state-of-the-art methods.

cs.DC↗

arXiv2025

Attention-Aware GNN-based Input Defense against Multi-Turn LLM Jailbreak

Large Language Models (LLMs) have gained significant traction in various applications, yet their capabilities present risks for both constructive and malicious exploitation. Despite extensive training and fine-tuning efforts aimed at enhancing safety, LLMs remain susceptible to jailbreak attacks. Recently, the emergence of multi-turn attacks has intensified this vulnerability. Unlike single-turn attacks, multi-turn attacks incrementally escalate dialogue complexity, rendering them more challenging to detect and mitigate. In this study, we introduce G-Guard, an innovative attention-aware Graph Neural Network (GNN)-based input classifier specifically designed to defend against multi-turn jailbreak attacks targeting LLMs. G-Guard constructs an entity graph for multi-turn queries, which captures the interrelationships between queries and harmful keywords that present in multi-turn queries. Furthermore, we propose an attention-aware augmentation mechanism that retrieves the most relevant single-turn query based on the ongoing multi-turn conversation. The retrieved query is incorporated as a labeled node within the graph, thereby enhancing the GNN's capacity to classify the current query as harmful or benign. Evaluation results show that G-Guard consistently outperforms all baselines across diverse datasets and evaluation metrics, demonstrating its efficacy as a robust defense mechanism against multi-turn jailbreak attacks.

cs.LG↗

arXiv2024

Mera: Memory Reduction and Acceleration for Quantum Circuit Simulation via Redundancy Exploration

With the development of quantum computing, quantum processor demonstrates the potential supremacy in specific applications, such as Grovers database search and popular quantum neural networks (QNNs). For better calibrating the quantum algorithms and machines, quantum circuit simulation on classical computers becomes crucial. However, as the number of quantum bits (qubits) increases, the memory requirement grows exponentially. In order to reduce memory usage and accelerate simulation, we propose a multi-level optimization, namely Mera, by exploring memory and computation redundancy. First, for a large number of sparse quantum gates, we propose two compressed structures for low-level full-state simulation. The corresponding gate operations are designed for practical implementations, which are relieved from the longtime compression and decompression. Second, for the dense Hadamard gate, which is definitely used to construct the superposition, we design a customized structure for significant memory saving as a regularity-oriented simulation. Meanwhile, an ondemand amplitude updating process is optimized for execution acceleration. Experiments show that our compressed structures increase the number of qubits from 17 to 35, and achieve up to 6.9 times acceleration for QNN.

quant-ph↗

arXiv2021

A Survey of Blockchain Data Management Systems

Blockchain has been widely deployed in various sectors, such as finance, education, and public services. Since blockchain runs as an immutable distributed ledger, it has decentralized mechanisms with persistency, anonymity, and auditability, where transactions are jointly performed through cryptocurrency-based consensus algorithms by worldwide distributed nodes. There have been many survey papers reviewing the blockchain technologies from different perspectives, e.g., digital currencies, consensus algorithms, and smart contracts. However, none of them have focused on the blockchain data management systems. To fill in this gap, we have conducted a comprehensive survey on the data management systems, based on three typical types of blockchain, i.e., standard blockchain, hybrid blockchain, and DAG (Directed Acyclic Graph)-based blockchain. We categorize their data management mechanisms into three layers: blockchain architecture, blockchain data structure, and blockchain storage engine, where block architecture indicates how to record transactions on a distributed ledger, blockchain data structure refers to the internal structure of each block, and blockchain storage engine specifies the storage form of data on the blockchain system. For each layer, the works advancing the state-of-the-art are discussed together with technical challenges. Furthermore, we lay out the future research directions for the blockchain data management systems.

cs.CR↗