arXiv subjects

Zeyan Li

Publications and source records attributed to Zeyan Li.

13 recordsLinked to original sources

arXiv2026

E-MIA: Exam-Style Black-Box Membership Inference Attacks against RAG Systems

Retrieval-Augmented Generation (RAG) equips large language models (LLMs) with external evidence by retrieving documents at inference time, but it also turns the retrieval corpusinto a sensitive asset. Under a black-box setting, an adversary given a candidate document can infer whether it has been ingested into the RAG knowledge base (i.e., document-level membership inference) solely from query response interactions, thereby leaking corpus coverage and the existence of sensitive topics. Existing RAG MIA methods either rely on soft signals such as semantic similarity, which often yield overlapping member/non-member score distributions and unstable thresholds, or employ explicit confirmation probes whose intent is conspicuous and thus prone to refusal and detection. We propose E-MIA, which converts verifiable hard evidence in the target document (e.g., fine-grained details, proper nouns/technical terms, definitional statements, metadata cues, and causal/constraint relations) into an exam with four objectively gradable question types (FB/SC/MC/T/F), and uses the aggregated exam score across multiple evidence targeted questions as the membership signal. Experiments across multiple datasets and diverse RAG configurations demonstrate that E-MIA improves member/non-member separability in stringent settings while preserving natural, stealthy queries, and we further analyze the impact of question composition and exam length on attack effectiveness.

cs.CR↗

arXiv2026

KairosVL: Orchestrating Time Series and Semantics for Unified Reasoning

Driven by the increasingly complex and decision-oriented demands of time series analysis, we introduce the Semantic-Conditional Time Series Reasoning task, which extends conventional time series analysis beyond purely numerical modeling to incorporate contextual and semantic understanding. To further enhance the mode's reasoning capabilities on complex time series problems, we propose a two-round reinforcement learning framework: the first round strengthens the mode's perception of fundamental temporal primitives, while the second focuses on semantic-conditioned reasoning. The resulting model, KairosVL, achieves competitive performance across both synthetic and real-world tasks. Extensive experiments and ablation studies demonstrate that our framework not only boosts performance but also preserves intrinsic reasoning ability and significantly improves generalization to unseen scenarios. To summarize, our work highlights the potential of combining semantic reasoning with temporal modeling and provides a practical framework for real-world time series intelligence, which is in urgent demand.

cs.AI↗

arXiv2025

Information Physics of Intelligence: Unifying Logical Depth and Entropy under Thermodynamic Constraints

The rapid scaling of artificial intelligence models has revealed a fundamental tension between model capacity (storage) and inference efficiency (computation). While classical information theory focuses on transmission and storage limits, it lacks a unified physical framework to quantify the thermodynamic costs of generating information from compressed laws versus retrieving it from memory. In this paper, we propose a theoretical framework that treats information processing as an enabling mapping from ontological states to carrier states. We introduce a novel metric, Derivation Entropy, which quantifies the effective work required to compute a target state from a given logical depth. By analyzing the interplay between Shannon entropy (storage) and computational complexity (time/energy), we demonstrate the existence of a critical phase transition point. Below this threshold, memory retrieval is thermodynamically favorable; above it, generative computation becomes the optimal strategy. This "Energy-Time-Space" conservation law provides a physical explanation for the efficiency of generative models and offers a rigorous mathematical bound for designing next-generation, energy-efficient AI architectures. Our findings suggest that the minimization of Derivation Entropy is a governing principle for the evolution of both biological and artificial intelligence.

cs.IT↗

arXiv2025

Does Scaling Law Apply in Time Series Forecasting?

Rapid expansion of model size has emerged as a key challenge in time series forecasting. From early Transformer with tens of megabytes to recent architectures like TimesNet with thousands of megabytes, performance gains have often come at the cost of exponentially increasing parameter counts. But is this scaling truly necessary? To question the applicability of the scaling law in time series forecasting, we propose Alinear, an ultra-lightweight forecasting model that achieves competitive performance using only k-level parameters. We introduce a horizon-aware adaptive decomposition mechanism that dynamically rebalances component emphasis across different forecast lengths, alongside a progressive frequency attenuation strategy that achieves stable prediction in various forecasting horizons without incurring the computational overhead of attention mechanisms. Extensive experiments on seven benchmark datasets demonstrate that Alinear consistently outperforms large-scale models while using less than 1% of their parameters, maintaining strong accuracy across both short and ultra-long forecasting horizons. Moreover, to more fairly evaluate model efficiency, we propose a new parameter-aware evaluation metric that highlights the superiority of ALinear under constrained model budgets. Our analysis reveals that the relative importance of trend and seasonal components varies depending on data characteristics rather than following a fixed pattern, validating the necessity of our adaptive design. This work challenges the prevailing belief that larger models are inherently better and suggests a paradigm shift toward more efficient time series modeling.

cs.LG↗

arXiv2025

Adaptive and Efficient Log Parsing as a Cloud Service

Logs are a critical data source for cloud systems, enabling advanced features like monitoring, alerting, and root cause analysis. However, the massive scale and diverse formats of unstructured logs pose challenges for adaptable, efficient, and accurate parsing methods. This paper introduces ByteBrain-LogParser, an innovative log parsing framework designed specifically for cloud environments. ByteBrain-LogParser employs a hierarchical clustering algorithm to allow real-time precision adjustments, coupled with optimizations such as positional similarity distance, deduplication, and hash encoding to enhance performance. Experiments on large-scale datasets show that it processes 229,000 logs per second on average, achieving an 840% speedup over the fastest baseline while maintaining accuracy comparable to state-of-the-art methods. Real-world evaluations further validate its efficiency and adaptability, demonstrating its potential as a robust cloud-based log parsing solution.

cs.SE↗

arXiv2025

Flow-of-Action: SOP Enhanced LLM-Based Multi-Agent System for Root Cause Analysis

In the realm of microservices architecture, the occurrence of frequent incidents necessitates the employment of Root Cause Analysis (RCA) for swift issue resolution. It is common that a serious incident can take several domain experts hours to identify the root cause. Consequently, a contemporary trend involves harnessing Large Language Models (LLMs) as automated agents for RCA. Though the recent ReAct framework aligns well with the Site Reliability Engineers (SREs) for its thought-action-observation paradigm, its hallucinations often lead to irrelevant actions and directly affect subsequent results. Additionally, the complex and variable clues of the incident can overwhelm the model one step further. To confront these challenges, we propose Flow-of-Action, a pioneering Standard Operation Procedure (SOP) enhanced LLM-based multi-agent system. By explicitly summarizing the diagnosis steps of SREs, SOP imposes constraints on LLMs at crucial junctures, guiding the RCA process towards the correct trajectory. To facilitate the rational and effective utilization of SOPs, we design an SOP-centric framework called SOP flow. SOP flow contains a series of tools, including one for finding relevant SOPs for incidents, another for automatically generating SOPs for incidents without relevant ones, and a tool for converting SOPs into code. This significantly alleviates the hallucination issues of ReAct in RCA tasks. We also design multiple auxiliary agents to assist the main agent by removing useless noise, narrowing the search space, and informing the main agent whether the RCA procedure can stop. Compared to the ReAct method's 35.50% accuracy, our Flow-of-Action method achieves 64.01%, meeting the accuracy requirements for RCA in real-world systems.

cs.SE↗

arXiv2024

ChatTS: Aligning Time Series with LLMs via Synthetic Data for Enhanced Understanding and Reasoning

Understanding time series is crucial for its application in real-world scenarios. Recently, large language models (LLMs) have been increasingly applied to time series tasks, leveraging their strong language capabilities to enhance various applications. However, research on multimodal LLMs (MLLMs) for time series understanding and reasoning remains limited, primarily due to the scarcity of high-quality datasets that align time series with textual information. This paper introduces ChatTS, a novel MLLM designed for time series analysis. ChatTS treats time series as a modality, similar to how vision MLLMs process images, enabling it to perform both understanding and reasoning with time series. To address the scarcity of training data, we propose an attribute-based method for generating synthetic time series with detailed attribute descriptions. We further introduce Time Series Evol-Instruct, a novel approach that generates diverse time series Q&As, enhancing the model's reasoning capabilities. To the best of our knowledge, ChatTS is the first TS-MLLM that takes multivariate time series as input for understanding and reasoning, which is fine-tuned exclusively on synthetic datasets. We evaluate its performance using benchmark datasets with real-world data, including six alignment tasks and four reasoning tasks. Our results show that ChatTS significantly outperforms existing vision-based MLLMs (e.g., GPT-4o) and text/agent-based LLMs, achieving a 46.0% improvement in alignment tasks and a 25.8% improvement in reasoning tasks. We have open-sourced the source code, model checkpoint and datasets at https://github.com/NetManAIOps/ChatTS.

cs.AI↗

arXiv2023

Generic and Robust Root Cause Localization for Multi-Dimensional Data in Online Service Systems

Localizing root causes for multi-dimensional data is critical to ensure online service systems' reliability. When a fault occurs, only the measure values within specific attribute combinations are abnormal. Such attribute combinations are substantial clues to the underlying root causes and thus are called root causes of multidimensional data. This paper proposes a generic and robust root cause localization approach for multi-dimensional data, PSqueeze. We propose a generic property of root cause for multi-dimensional data, generalized ripple effect (GRE). Based on it, we propose a novel probabilistic cluster method and a robust heuristic search method. Moreover, we identify the importance of determining external root causes and propose an effective method for the first time in literature. Our experiments on two real-world datasets with 5400 faults show that the F1-score of PSqueeze outperforms baselines by 32.89%, while the localization time is around 10 seconds across all cases. The F1-score in determining external root causes of PSqueeze achieves 0.90. Furthermore, case studies in several production systems demonstrate that PSqueeze is helpful to fault diagnosis in the real world.

cs.SE↗

arXiv2022

Actionable and Interpretable Fault Localization for Recurring Failures in Online Service Systems

Fault localization is challenging in an online service system due to its monitoring data's large volume and variety and complex dependencies across or within its components (e.g., services or databases). Furthermore, engineers require fault localization solutions to be actionable and interpretable, which existing research approaches cannot satisfy. Therefore, the common industry practice is that, for a specific online service system, its experienced engineers focus on localization for recurring failures based on the knowledge accumulated about the system and historical failures. Although the above common practice is actionable and interpretable, it is largely manual, thus slow and sometimes inaccurate. In this paper, we aim to automate this practice through machine learning. That is, we propose an actionable and interpretable fault localization approach, DejaVu, for recurring failures in online service systems. For a specific online service system, DejaVu takes historical failures and dependencies in the system as input and trains a localization model offline; for an incoming failure, the trained model online recommends where the failure occurs (i.e., the faulty components) and which kind of failure occurs (i.e., the indicative group of metrics) (thus actionable), which are further interpreted by both global and local interpretation methods (thus interpretable). Based on the evaluation on 601 failures from three production systems and one open-source benchmark, in less than one second, DejaVu can on average rank the ground truths at 1.66-th to 5.03-th among a long candidate list, outperforming baselines by at least 51.51%.

cs.SE↗

arXiv2022

Constructing Large-Scale Real-World Benchmark Datasets for AIOps

Recently, AIOps (Artificial Intelligence for IT Operations) has been well studied in academia and industry to enable automated and effective software service management. Plenty of efforts have been dedicated to AIOps, including anomaly detection, root cause localization, incident management, etc. However, most existing works are evaluated on private datasets, so their generality and real performance cannot be guaranteed. The lack of public large-scale real-world datasets has prevented researchers and engineers from enhancing the development of AIOps. To tackle this dilemma, in this work, we introduce three public real-world, large-scale datasets about AIOps, mainly aiming at KPI anomaly detection, root cause localization on multi-dimensional data, and failure discovery and diagnosis. More importantly, we held three competitions in 2018/2019/2020 based on these datasets, attracting thousands of teams to participate. In the future, we will continue to publish more datasets and hold competitions to promote the development of AIOps further.

cs.SE↗

arXiv2022

Causal Inference-Based Root Cause Analysis for Online Service Systems with Intervention Recognition

Fault diagnosis is critical in many domains, as faults may lead to safety threats or economic losses. In the field of online service systems, operators rely on enormous monitoring data to detect and mitigate failures. Quickly recognizing a small set of root cause indicators for the underlying fault can save much time for failure mitigation. In this paper, we formulate the root cause analysis problem as a new causal inference task named intervention recognition. We proposed a novel unsupervised causal inference-based method named Causal Inference-based Root Cause Analysis (CIRCA). The core idea is a sufficient condition for a monitoring variable to be a root cause indicator, i.e., the change of probability distribution conditioned on the parents in the Causal Bayesian Network (CBN). Towards the application in online service systems, CIRCA constructs a graph among monitoring metrics based on the knowledge of system architecture and a set of causal assumptions. The simulation study illustrates the theoretical reliability of CIRCA. The performance on a real-world dataset further shows that CIRCA can improve the recall of the top-1 recommendation by 25% over the best baseline method.

cs.SE↗

arXiv2022

GenAD: General Representations of Multivariate Time Seriesfor Anomaly Detection

The reliability of wireless base stations in China Mobile is of vital importance, because the cell phone users are connected to the stations and the behaviors of the stations are directly related to user experience. Although the monitoring of the station behaviors can be realized by anomaly detection on multivariate time series, due to complex correlations and various temporal patterns of multivariate series in large-scale stations, building a general unsupervised anomaly detection model with a higher F1-score remains a challenging task. In this paper, we propose a General representation of multivariate time series for Anomaly Detection(GenAD). First, we pre-train a general model on large-scale wireless base stations with self-supervision, which can be easily transferred to a specific station anomaly detection with a small amount of training data. Second, we employ Multi-Correlation Attention and Time-Series Attention to represent the correlations and temporal patterns of the stations. With the above innovations, GenAD increases F1-score by total 9% on real-world datasets in China Mobile, while the performance does not significantly degrade on public datasets with only 10% of the training data.

cs.NI↗

arXiv2018

Unsupervised Anomaly Detection via Variational Auto-Encoder for Seasonal KPIs in Web Applications

To ensure undisrupted business, large Internet companies need to closely monitor various KPIs (e.g., Page Views, number of online users, and number of orders) of its Web applications, to accurately detect anomalies and trigger timely troubleshooting/mitigation. However, anomaly detection for these seasonal KPIs with various patterns and data quality has been a great challenge, especially without labels. In this paper, we proposed Donut, an unsupervised anomaly detection algorithm based on VAE. Thanks to a few of our key techniques, Donut greatly outperforms a state-of-arts supervised ensemble approach and a baseline VAE approach, and its best F-scores range from 0.75 to 0.9 for the studied KPIs from a top global Internet company. We come up with a novel KDE interpretation of reconstruction for Donut, making it the first VAE-based anomaly detection algorithm with solid theoretical explanation.

cs.LG↗