arXiv subjects

Tailin Zhou

Publications and source records attributed to Tailin Zhou.

9 recordsLinked to original sources

arXiv2026

ScienceFlow: A long-horizon agent for ML research, scientific discovery and beyond

Enabling LLM agents to sustain productive, stable, and goal-aligned research over extended horizons is a central challenge for autonomous machine learning and scientific discovery, as progress hinges on continuously managing evolving state, exploration decisions, and computational resources. Pioneering autoresearch agents, despite great success, still lack mechanisms for continuity, recovery from dead ends, and value-driven compute allocation, which inherently undermines overall search efficiency, wastes computational resources, and lowers the chance of ultimate success. To bridge this gap, we introduce ScienceFlow, an end-to-end autoresearch agent framework that organizes long-horizon research work into research segments grounded in executable workspaces. It represents research progress as recoverable executable states, enabling efficient exploration, revision, and execution. Transitions between research segments are governed by Executable-State Transition through Re-Anchoring (ESTRA), which selects either the live state or an archived state as the next anchor and determines whether to continue or redirect the research trajectory. An evidence-aware execution controller allocates resources to physical jobs based on resource availability, remaining budget, and validated progress. We evaluate ScienceFlow on tasks spanning machine learning, scientific modeling, and mathematical optimization. Results on diverse long-horizon benchmarks demonstrate its ability to sustain effective research processes, highlighted by a SOTA 70.22 percent Any-Medal score on the full MLE-bench within a 24-hour budget, outperforming prior reported results by 4.92 percentage points. The efficacy of ScienceFlow further demonstrates that efficient state management, adaptive exploration, and objective-aligned execution are critical for scaling autonomous research beyond short-horizon interactions.

cs.AI↗

arXiv2026

Hierarchical Self-Improvement: A Framework for Task-Specific Evolvable Agent Harnesses

Modern LLM agents are often improved by modifying prompts, tools, or workflows manually, while the executable scaffold surrounding the model---the \emph{harness}---is typically treated as a fixed artifact after deployment. This work studies an alternative where the harness is \emph{task-specific and continuously evolvable}: each task family maintains its own harness, which is hot-swapped across iterations through a fixed task-injection seam and rewritten using environment feedback. We introduce \textbf{Hierarchical Self-Improvement (HSI)}, a framework in which a single frozen LLM $M$ operates across three hierarchical scopes: a task harness $H$ that executes tasks, an evolver that rewrites $H$, and a meta-evolver that rewrites the evolver's strategy code under a frozen outer anchor. A thinking-on/off design isolates the contribution of harness evolution by disabling reasoning during task execution while enabling it during self-modification. HSI is bounded by two factors: a \emph{feedback-fidelity bound}, since evolution requires informative reward signals to guide selection, and a \emph{backbone capability bound}, since harness redesign cannot overcome limitations of the frozen model. On BALROG with DeepSeek-V4-Flash-Preview as the frozen backbone, HSI achieves consistent gains over the initial harness on moderate-difficulty tasks ($+39.3$ on BabyAI, $+33.0$ on Crafter, $+25.0$ on TextWorld, and $+15.0$ on MiniHack, all in raw \% Progress), while obtaining strong held-out generalization on BabaIsAI sub-suites ($0.98$ best-test on BreakStop and $1.00$ on GoTo from a $20\%$ unseen split). On tasks beyond the backbone's capability (NLE), harness evolution provides no improvement. These results demonstrate task-specific harness evolution as a viable axis for improving frozen LLM agents under clear empirical limits. Code is available at https://github.com/TailinZhou/hsi.

cs.AI↗

arXiv2025

Learning Design-Score Manifold to Guide Diffusion Models for Offline Optimization

Optimizing complex systems, from discovering therapeutic drugs to designing high-performance materials, remains a fundamental challenge across science and engineering, as the underlying rules are often unknown and costly to evaluate. Offline optimization aims to optimize designs for target scores using pre-collected datasets without system interaction. However, conventional approaches may fail beyond training data, predicting inaccurate scores and generating inferior designs. This paper introduces ManGO, a diffusion-based framework that learns the design-score manifold, capturing the design-score interdependencies holistically. Unlike existing methods that treat design and score spaces in isolation, ManGO unifies forward prediction and backward generation, attaining generalization beyond training data. Key to this is its derivative-free guidance for conditional generation, coupled with adaptive inference-time scaling that dynamically optimizes denoising paths. Extensive evaluations demonstrate that ManGO outperforms 24 single- and 10 multi-objective optimization methods across diverse domains, including synthetic tasks, robot control, material design, DNA sequence, and real-world engineering optimization.

cs.LG↗

arXiv2024

Understanding and Improving Model Averaging in Federated Learning on Heterogeneous Data

Model averaging is a widely adopted technique in federated learning (FL) that aggregates multiple client models to obtain a global model. Remarkably, model averaging in FL yields a superior global model, even when client models are trained with non-convex objective functions and on heterogeneous local datasets. However, the rationale behind its success remains poorly understood. To shed light on this issue, we first visualize the loss landscape of FL over client and global models to illustrate their geometric properties. The visualization shows that the client models encompass the global model within a common basin, and interestingly, the global model may deviate from the basin's center while still outperforming the client models. To gain further insights into model averaging in FL, we decompose the expected loss of the global model into five factors related to the client models. Specifically, our analysis reveals that the global model loss after early training mainly arises from \textit{i)} the client model's loss on non-overlapping data between client datasets and the global dataset and \textit{ii)} the maximum distance between the global and client models. Based on the findings from our loss landscape visualization and loss decomposition, we propose utilizing iterative moving averaging (IMA) on the global model at the late training phase to reduce its deviation from the expected minimum, while constraining client exploration to limit the maximum distance between the global and client models. Our experiments demonstrate that incorporating IMA into existing FL methods significantly improves their accuracy and training speed on various heterogeneous data setups of benchmark datasets. Code is available at \url{https://github.com/TailinZhou/FedIMA}.

cs.LG↗

arXiv2024

A Survey of What to Share in Federated Learning: Perspectives on Model Utility, Privacy Leakage, and Communication Efficiency

Federated learning (FL) has emerged as a secure paradigm for collaborative training among clients. Without data centralization, FL allows clients to share local information in a privacy-preserving manner. This approach has gained considerable attention, promoting numerous surveys to summarize the related works. However, the majority of these surveys concentrate on FL methods that share model parameters during the training process, while overlooking the possibility of sharing local information in other forms. In this paper, we present a systematic survey from a new perspective of what to share in FL, with an emphasis on the model utility, privacy leakage, and communication efficiency. First, we present a new taxonomy of FL methods in terms of three sharing methods, which respectively share model, synthetic data, and knowledge. Second, we analyze the vulnerability of different sharing methods to privacy attacks and review the defense mechanisms. Third, we conduct extensive experiments to compare the learning performance and communication overhead of various sharing methods in FL. Besides, we assess the potential privacy leakage through model inversion and membership inference attacks, while comparing the effectiveness of various defense approaches. Finally, we identify future research directions and conclude the survey.

cs.LG↗

arXiv2024

Federated Prompt-based Decision Transformer for Customized VR Services in Mobile Edge Computing System

This paper investigates resource allocation to provide heterogeneous users with customized virtual reality (VR) services in a mobile edge computing (MEC) system. We first introduce a quality of experience (QoE) metric to measure user experience, which considers the MEC system's latency, user attention levels, and preferred resolutions. Then, a QoE maximization problem is formulated for resource allocation to ensure the highest possible user experience,which is cast as a reinforcement learning problem, aiming to learn a generalized policy applicable across diverse user environments for all MEC servers. To learn the generalized policy, we propose a framework that employs federated learning (FL) and prompt-based sequence modeling to pre-train a common decision model across MEC servers, which is named FedPromptDT. Using FL solves the problem of insufficient local MEC data while protecting user privacy during offline training. The design of prompts integrating user-environment cues and user-preferred allocation improves the model's adaptability to various user environments during online execution.

cs.AI↗

arXiv2023

FedFA: Federated Learning with Feature Anchors to Align Features and Classifiers for Heterogeneous Data

Federated learning allows multiple clients to collaboratively train a model without exchanging their data, thus preserving data privacy. Unfortunately, it suffers significant performance degradation due to heterogeneous data at clients. Common solutions involve designing an auxiliary loss to regularize weight divergence or feature inconsistency during local training. However, we discover that these approaches fall short of the expected performance because they ignore the existence of a vicious cycle between feature inconsistency and classifier divergence across clients. This vicious cycle causes client models to be updated in inconsistent feature spaces with more diverged classifiers. To break the vicious cycle, we propose a novel framework named Federated learning with Feature Anchors (FedFA). FedFA utilizes feature anchors to align features and calibrate classifiers across clients simultaneously. This enables client models to be updated in a shared feature space with consistent classifiers during local training. Theoretically, we analyze the non-convex convergence rate of FedFA. We also demonstrate that the integration of feature alignment and classifier calibration in FedFA brings a virtuous cycle between feature and classifier updates, which breaks the vicious cycle existing in current approaches. Extensive experiments show that FedFA significantly outperforms existing approaches on various classification datasets under label distribution skew and feature distribution skew.

cs.LG↗

arXiv2023

Mode Connectivity and Data Heterogeneity of Federated Learning

Federated learning (FL) enables multiple clients to train a model while keeping their data private collaboratively. Previous studies have shown that data heterogeneity between clients leads to drifts across client updates. However, there are few studies on the relationship between client and global modes, making it unclear where these updates end up drifting. We perform empirical and theoretical studies on this relationship by utilizing mode connectivity, which measures performance change (i.e., connectivity) along parametric paths between different modes. Empirically, reducing data heterogeneity makes the connectivity on different paths more similar, forming more low-error overlaps between client and global modes. We also find that a barrier to connectivity occurs when linearly connecting two global modes, while it disappears with considering non-linear mode connectivity. Theoretically, we establish a quantitative bound on the global-mode connectivity using mean-field theory or dropout stability. The bound demonstrates that the connectivity improves when reducing data heterogeneity and widening trained models. Numerical results further corroborate our analytical findings.

cs.LG↗

arXiv2023

Attention-based QoE-aware Digital Twin Empowered Edge Computing for Immersive Virtual Reality

Metaverse applications such as virtual reality (VR) content streaming, require optimal resource allocation strategies for mobile edge computing (MEC) to ensure a high-quality user experience. In contrast to online reinforcement learning (RL) algorithms, which can incur substantial communication overheads and longer delays, the majority of existing works employ offline-trained RL algorithms for resource allocation decisions in MEC systems. However, they neglect the impact of desynchronization between the physical and digital worlds on the effectiveness of the allocation strategy. In this paper, we tackle this desynchronization using a continual RL framework that facilitates the resource allocation dynamically for MEC-enabled VR content streaming. We first design a digital twin-empowered edge computing (DTEC) system and formulate a quality of experience (QoE) maximization problem based on attention-based resolution perception. This problem optimizes the allocation of computing and bandwidth resources while adapting the attention-based resolution of the VR content. The continual RL framework in DTEC enables adaptive online execution in a time-varying environment. The reward function is defined based on the QoE and horizon-fairness QoE (hfQoE) constraints. Furthermore, we propose freshness prioritized experience replay - continual deep deterministic policy gradient (FPER-CDDPG) to enhance the performance of continual learning in the presence of time-varying DT updates. We test FPER-CDDPG using extensive experiments and evaluation. FPER-CDDPG outperforms the benchmarks in terms of average latency, QoE, and successful delivery rate as well as meeting the hfQoE requirements and performance over long-term execution while ensuring system scalability with the increasing number of users.

eess.SY↗