arXiv subjects

Bo Sun

Publications and source records attributed to Bo Sun.

At least 19 recordsLinked to original sources

arXiv2026

Auto-RecSys: Harnessing Autonomous Research Agents for Industry-Scale Recommender System

Auto-research agents have shown the potential to automate hypothesis generation, experiment execution, and iterative refinement. However, scaling this paradigm to industry-scale recommendation models introduces two challenges: (1) long feedback loops, where model training can take days, making serial iteration prohibitively slow and requiring parallel exploration across multiple research directions; and (2) system complexity, where large configurations, fragile infrastructure dependencies, and multi-day GPU jobs require robust and recoverable execution. We present Auto-RecSys, an autonomous research system for long-horizon experimentation on industry-scale recommendation models. Auto-RecSys addresses these challenges through three harness designs: (1) distributed asynchronous execution for running multiple experiments in parallel across servers, (2) centralized cross-server memory for persistent and recoverable execution across sessions and failures, and (3) cognitive-procedural separation, where natural-language skill files guide LLM reasoning while deterministic scripts enforce operational correctness. Auto-RecSys further employs a dual-loop self-evolving architecture: an Execution Evolution Loop in which model-specific playbooks accumulate operational knowledge by recording failed attempts and crystallizing successful pipelines, and an Idea Evolution Loop in which experimental outcomes inform subsequent ideation. Evaluated on recommendation models, Auto-RecSys significantly reduces the human time required per experiment cycle and improves execution reliability as its playbooks mature.

cs.CL↗

arXiv2026

OpenHarmony Bench: Evaluating LLMs and Coding Agents on OpenHarmony App Development

We present OPENHARMONY BENCH, an app-level coding benchmark for evaluating LLM-based coding agents on OpenHarmony ArkTS applications. Unlike function-level benchmarks, it evaluates complete app-level changes: each task requires an agent to modify a buildable ArkTS project so that a requested behavior works end to end, involving UI state, data persistence, build configuration, and platform APIs. The benchmark installs and drives the delivered application on a device to check whether the behavior is observable. It covers three input sources: natural-language feature requests (new-feature), structured scenario specifications (spec-driven), and bug descriptions (bug-fix). The benchmark contains 153 top-level tasks and 242 Feature points (F-points), where an F-point is one executable behavior check. The snapshot includes 32 new-feature tasks, 50 spec-driven tasks with 139 F-points, and 71 bug-fix tasks. The main leaderboard is scored over top-level tasks rather than independently weighted F-points. We describe the benchmark construction, statistics, and build-and-test evaluation pipeline, and evaluate DevEco Code with eight LLMs across three independent full-suite runs per configuration. Three findings emerge. First, newer generations complete more tasks than their predecessors within evaluated model-family pairs. Second, buildability is close to saturated while behavioral correctness is not: mean Final Build Success Rate is 94.77% to 100.00%, whereas mean Task Completion is 48.36% to 58.39%. Third, spec-driven tasks have the lowest Task Completion under all-checks task scoring, with no configuration exceeding 35%. The code, data, tasks, reference solutions, tests, evaluation scripts, and leaderboard are released through the official OPENHARMONY BENCH website at https://bench.matrix.openharmony.cn/.

cs.SE↗

arXiv2026

Data-Driven Online Slice Admission Control and Resource Allocation in NextG Mobile Networks

Virtualization in 5G and beyond networks enables the creation of virtual networks (i.e., network slices) tailored to the needs of different applications. To maximize revenue under limited infrastructure resources, InPs must decide in real time whether to admit incoming slice requests (SRs) based on their resource demands and offered values, while accounting for the opportunity cost of consuming scarce resources. To address this challenge, we introduce Online Pricing-based Slice Admission Control and Resource Allocation (OPA) framework. This framework dynamically assigns pseudo-prices to resources that capture long-term scarcity and anticipated inter-temporal opportunity costs. The short-term admission and resource allocation decisions for each SR are then guided by these prices. Additionally, we design an exponential pricing strategy that guarantees bounded worst-case performance. To improve practical performance, we further develop a data-driven exponential pricing approach that learns from historical data. Evaluations on a real-world network topology show that it improves mean revenue by 32.2% and 26.7% over state-of-the-art DRL and optimization-based approaches, respectively, while reducing computational cost by an order of magnitude relative to the latter.

cs.NI↗

arXiv2026

Learning Graphs through Continuous Information Entropy Fields

Graph theory is inherently descriptive, capturing what relationships exist but not why they arise, because it treats edges as primitive constructs. This paper proposes a new explanatory framework for graph learning, where relationships emerge from latent continuous information entropy fields, and a graph becomes a discrete instantiation of an underlying field. To formalize this field, we introduce the Field-informed Graph Network (FGN). It learns a scalar field from node features and leverages it to modulate message passing. The information-theoretic objective balances structural fidelity with field smoothness, forming a self-reinforcing loop. In this loop, the field modulates information diffusion through field-modulated weighting, and the updated node representations iteratively refine the field. As a result, FGN learns by simulating its own co-evolution. Extensive experiments on node classification and graph classification benchmarks demonstrate superior performance, robustness to perturbations, and structurally coherent field representations.

cs.LG↗

arXiv2026

Source-Free Detection and Impact Analysis of Compiler Optimization Problems in Mobile Applications

Mobile apps frequently suffer from frame drops, overheating, and excessive power consumption. While developers optimize algorithms and debug code, a critical bottleneck often goes unnoticed: native libraries compiled with low optimization levels (O0/O1 instead of O2/O3). Because these libraries execute without functional errors, the resulting performance degradation remains hidden in production apps. We present \textsc{OptDetect}, a source-free framework that detects compiler optimization problems directly from app binaries. \textsc{OptDetect} handles mixed optimization levels through binary disassembly, chunk-level classification, and weighted score aggregation, achieving 93.0\% accuracy on controlled datasets and 81.9\% on real-world datasets. Applying \textsc{OptDetect} to 21,972 native libraries from 830 top-ranked Google Play apps, we find that 30.5\% of libraries use low optimization levels, affecting 91.7\% of apps. Through case studies on 12 production apps, fixing detected issues reduces CPU instructions by 10-63\% (median: 20.5\%) for commercial apps and 15-58\% (median: 32\%) for open-source apps. Performance complaints decrease in 5 of 6 commercial apps, and ratings increase in 5 of 6. Further investigation reveals that widely-used third-party libraries are themselves distributed at low optimization levels, with 49.7\% of 1,073 libraries in a major repository exhibiting this problem. These findings show that compiler optimization problems are common, source-free detectable, and practically consequential in mobile app ecosystems.

cs.SE↗

arXiv2026

CollabSim: A CSCW-Grounded Methodology for Investigating Collaborative Competence of LLM Agents through Controlled Multi-Agent Experiments

Multi-agent systems (MAS) built on large language models have shown growing promise, with their effectiveness resting on agents' ability to coordinate through text-based channels much as human teams do. Yet recent study suggests that MAS often falter not because agents lack individual task-solving ability, but because they lack collaborative competence: the capacity to establish common ground, maintain shared task understanding, balance individual and collective incentives, and repair misalignment as interaction unfolds. Decades of research in Computer-Supported Cooperative Work have characterized these requirements for human teams coordinating under constrained communication, yet existing MAS evaluations focus mainly on task outcomes or single-agent proficiency in reasoning, planning, and tool use. To enable a systematic analysis of agents' collaborative competence in MAS, we introduce CollabSim, a configurable simulation framework that combines a theory-grounded definition of collaborative capabilities, controlled manipulation of interaction conditions, and action-level probing of agents' internal states. Experiments across four LLMs show that CollabSim can capture condition effects, separate model performance patterns, and reveal task-dependent effects of agent design.

cs.CL↗

arXiv2026

Sensor2Sensor: Cross-Embodiment Sensor Conversion for Autonomous Driving

Robust training and validation of Autonomous Driving Systems (ADS) require massive, diverse datasets. Proprietary data collected by Autonomous Vehicle (AV) fleets, while high-fidelity, are limited in scale, diversity of sensor configurations, as well as geographic and long-tail-behavioral coverage. In contrast, in-the-wild data from sources like dashcams offers immense scale and diversity, capturing critical long-tail scenarios and novel environments. However, this unstructured, in-the-wild video data is incompatible with ADS expecting structured, multi-modal sensor inputs for validation and training. To bridge this data gap, we propose Sensor2Sensor, a novel generative modeling paradigm that translates in-the-wild monocular dashcam videos into a high-fidelity, multi-modal sensor suite (AV logs) comprising multi-view camera images and LiDAR point clouds. A core challenge is the lack of paired training data. We address this by converting real AV logs into dashcam-style videos via 4D Gaussian Splatting (4DGS) reconstruction and novel-view rendering. Sensor2Sensor then utilizes a diffusion architecture to perform the generative conversion. We perform comprehensive quantitative evaluations on the fidelity and realism of the generated sensor data. We demonstrate Sensor2Sensor's practical utility by converting challenging in-the-wild internet and dashcam footage into realistic, multi-modal data formats, further unlocking vast external data sources for AV development.

cs.CV↗

arXiv2026

Fatigue-Related Reaction Time Forecasting via EEG Functional Connectivity in Sustained Attention Task

Mental fatigue related behavioral performance decline precipitates catastrophic accidents in sustained attention tasks. While existing neurophysiological systems effectively detect current behavioral performance, they often lack the capability to forecast behavioral lapses with sufficient temporal lead time for intervention. This study proposes a novel model for the reaction time (RT) forecasting using EEG functional connectivity features. Thirty participants engaged in a sustained Psychomotor Vigilance Test (PVT) with concurrent 30-channel EEG recording. Mutual information (MI) between electrodes was calculated as functional connectivity features. Random Forest regression model (RF) was trained to predict single-trial RTs across forecasting horizons ranging from 0 to 20 seconds. The model demonstrated robust predictive validity, achieving a Root Mean Square Error (RMSE) of 23.75 ms for immediate detection and maintaining high accuracy (RMSE = 24.07 ms) across different forecasting horizons. Interpretability analysis via SHAP and Linear Mixed Effects model further support the validity of the proposed model and revealed distinct temporal biomarkers. This study validates the feasibility of forecasting behavioral performance 20 seconds in advance, offering a promising methodology for proactive fatigue management in safety-critical systems.

cs.HC↗

arXiv2026

HyRA: A Hybrid Resource Allocation Framework for RAN Slicing

The advent of 5G and the emergence of 6G networks demand unprecedented flexibility and efficiency in Radio Access Network (RAN) resource management to satisfy diverse service-level agreements (SLAs). Existing RAN slicing frameworks predominantly rely on per-slice resource reservation, which ensures performance isolation but leads to inefficient utilization, particularly under bursty traffic. We introduce HyRA, a hybrid resource allocation framework for RAN slicing that combines dedicated per-slice allocations with shared resource pooling across slices. HyRA preserves performance isolation while improving resource efficiency by leveraging multiplexing gains in bursty traffic conditions. We formulate this design as a bi-level stochastic optimization problem, where the outer loop determines the dedicated and shared resource budgets and the inner loop performs per-UE scheduling under a novel water-filling approach. By using the sample-average approximation, the Karush-Kuhn-Tucker (KKT) conditions of the inner loop, and Big-M encoding, we transform the problem into a tractable mixed-integer program that standard optimization solvers can solve. Extensive simulations under diverse demand patterns, SLA configurations, and traffic burstiness show that HyRA achieves up to 50-75% spectrum savings compared to dedicated-only and shared-only baselines. These results highlight HyRA as a viable approach for resource-efficient, SLA-compliant RAN slicing in future mobile networks.

cs.NI↗

arXiv2025

Risk-Sensitive Online Selection with Bounded Adaptivity

Designing randomized online algorithms that perform reliably not only in expectation but also under unfavorable realizations of randomness is a fundamental challenge in online decision-making. In this paper, we study this challenge in online adversarial selection, where a decision maker allocates $k$ units of a resource to sequentially arriving buyers through posted prices. We focus on two intertwined considerations that are often overlooked simultaneously: tail-risk sensitivity and bounded adaptivity, where tail risk is measured using conditional value-at-risk (CVaR) and bounded adaptivity limits the number of allowable policy updates over time. Our main contribution is a correlated posted-price mechanism that uses a single random seed to coordinate pricing decisions across time. This correlation induces a monotonic ordering of pricing profiles across sample paths, improving lower-tail performance while respecting the adaptivity constraint. More broadly, our results highlight correlation as a mechanism for controlling tail risk in randomized online algorithms. Using this framework, we derive competitive guarantees for several regimes of the problem under both static and dynamic pricing. Our analysis develops a risk-sensitive randomized online primal-dual framework tailored to CVaR objectives and reveals a systematic trade-off between allowable adaptivity, risk sensitivity, and competitive performance. Experiments on real airline pricing data further illustrate the empirical impact of correlated pricing on welfare concentration and tail behavior.

cs.GT↗

arXiv2025

Electron Tunneling Enhances Thermal Conductance through Metal-Insulator-Semiconductor Junctions

The presence of interfaces in semiconductor devices substantially hinders thermal transport, contributing disproportionately to the overall thermal resistance. However, approaches to enhance interfacial thermal transport remain scarce without changing the interface structure, as the intrinsic electron and phonon properties of constituent materials set an upper limit. Here, we find a new thermal transport pathway, electronic heat tunneling, to enhance interfacial thermal conductance through metal-insulator-semiconductor junctions. By applying photoexcitation or bias voltage, we observe remarkable thermal conductance increases in operando, opening a new channel for efficient interfacial heat dissipation. The electron quantum tunneling pathway is parallel to conventional phonon-mediated interfacial thermal transport, and violates the Wiedemann-Franz law since this pathway deviates from the paradigm of diffusive transport. Moreover, we develop a tunneling mismatch model to describe the enhanced thermal conductance, originating from tunneling heat flux. Our Letter demonstrates a previously unexplored heat transport mechanism to enhance thermal conductance, bypassing the need for interface engineering. These findings emphasize the essential need to understand semiconductor thermal properties under realistic operating conditions.

cond-mat.mes-hall↗

arXiv2025

Make It Long, Keep It Fast: End-to-End 10K Long User Behavior Sequence Modeling for Billion-Scale Douyin Recommendation

Short-video recommenders such as Douyin must exploit extremely long user behavior histories without breaking latency or cost budgets. We present an end-to-end industrial recommender system that scales long-sequence recommendation modeling to 10K-length histories in production. First, we introduce Stacked Target-to-History Cross Attention (STCA), which replaces history self-attention with stacked cross-attention from the target to the history, reducing complexity from quadratic to linear in sequence length and enabling efficient end-to-end training over long user behavior sequences. Second, we propose Request Level Batching (RLB), a user-centric batching scheme that aggregates multiple targets for the same user/request to share the user-side encoding, substantially lowering sequence-related storage, communication, and compute without changing the learning objective. Third, we design a length-extrapolative training strategy -- train on shorter windows, infer on much longer ones -- so the model generalizes to 10K-scale histories without additional training cost. Across offline and online experiments, we observe predictable, monotonic gains as we scale history length and model capacity, mirroring the scaling law behavior observed in large language models. Deployed at full traffic on Douyin, our system delivers significant improvements on key engagement metrics while meeting production latency, demonstrating a practical path to scaling end-to-end ultra-long sequence recommendation to the 10K regime.

cs.LG↗

arXiv2025

Green Bin Packing

The online bin packing problem and its variants are regularly used to model server allocation problems. Modern concerns surrounding sustainability and overcommitment in cloud computing motivate bin packing models that capture costs associated with highly utilized servers. In this work, we introduce the green bin packing problem, an online variant with a linear cost $\beta$ for filling above a fixed level $G$. For a given instance, the goal is to minimize the sum of the number of opened bins and the linear cost. We show that when $\beta G \le 1$, classical online bin packing algorithms such as FirstFit or Harmonic perform well, and can achieve competitive ratios lower than in the classic setting. However, when $\beta G > 1$, new algorithmic solutions can improve both worst-case and typical performance. We introduce variants of classic online bin packing algorithms and establish theoretical bounds, as well as test their empirical performance.

cs.DS↗

arXiv2025

Online Multi-Class Selection with Group Fairness Guarantee

We study the online multi-class selection problem with group fairness guarantees, where limited resources must be allocated to sequentially arriving agents. Our work addresses two key limitations in the existing literature. First, we introduce a novel lossless rounding scheme that ensures the integral algorithm achieves the same expected performance as any fractional solution. Second, we explicitly address the challenges introduced by agents who belong to multiple classes. To this end, we develop a randomized algorithm based on a relax-and-round framework. The algorithm first computes a fractional solution using a resource reservation approach -- referred to as the set-aside mechanism -- to enforce fairness across classes. The subsequent rounding step preserves these fairness guarantees without degrading performance. Additionally, we propose a learning-augmented variant that incorporates untrusted machine-learned predictions to better balance fairness and efficiency in practical settings.

cs.LG↗

arXiv2025

DPRF: A Generalizable Dynamic Persona Refinement Framework for Optimizing Behavior Alignment Between Personalized LLM Role-Playing Agents and Humans

The emerging large language model role-playing agents (LLM RPAs) aim to simulate individual human behaviors, but the persona fidelity is often undermined by manually-created profiles (e.g., cherry-picked information and personality characteristics) without validating the alignment with the target individuals. To address this limitation, our work introduces the Dynamic Persona Refinement Framework (DPRF). DPRF aims to optimize the alignment of LLM RPAs' behaviors with those of target individuals by iteratively identifying the cognitive divergence, either through free-form or theory-grounded, structured analysis, between generated behaviors and human ground truth, and refining the persona profile to mitigate these divergences. We evaluate DPRF with five LLMs on four diverse behavior-prediction scenarios: formal debates, social media posts with mental health issues, public interviews, and movie reviews. DPRF can consistently improve behavioral alignment considerably over baseline personas and generalizes across models and scenarios. Our work provides a robust methodology for creating high-fidelity persona profiles and enhancing the validity of downstream applications, such as user simulation, social studies, and personalized AI.

cs.CL↗

arXiv2025

Drive&Gen: Co-Evaluating End-to-End Driving and Video Generation Models

Recent advances in generative models have sparked exciting new possibilities in the field of autonomous vehicles. Specifically, video generation models are now being explored as controllable virtual testing environments. Simultaneously, end-to-end (E2E) driving models have emerged as a streamlined alternative to conventional modular autonomous driving systems, gaining popularity for their simplicity and scalability. However, the application of these techniques to simulation and planning raises important questions. First, while video generation models can generate increasingly realistic videos, can these videos faithfully adhere to the specified conditions and be realistic enough for E2E autonomous planner evaluation? Second, given that data is crucial for understanding and controlling E2E planners, how can we gain deeper insights into their biases and improve their ability to generalize to out-of-distribution scenarios? In this work, we bridge the gap between the driving models and generative world models (Drive&Gen) to address these questions. We propose novel statistical measures leveraging E2E drivers to evaluate the realism of generated videos. By exploiting the controllability of the video generation model, we conduct targeted experiments to investigate distribution gaps affecting E2E planner performance. Finally, we show that synthetic data produced by the video generation model offers a cost-effective alternative to real-world data collection. This synthetic data effectively improves E2E model generalization beyond existing Operational Design Domains, facilitating the expansion of autonomous vehicle services into new operational contexts.

cs.CV↗

arXiv2025

Consistent Estimation of Numerical Distributions under Local Differential Privacy by Wavelet Expansion

Distribution estimation under local differential privacy (LDP) is a fundamental and challenging task. Significant progresses have been made on categorical data. However, due to different evaluation metrics, these methods do not work well when transferred to numerical data. In particular, we need to prevent the probability mass from being misplaced far away. In this paper, we propose a new approach that express the sample distribution using wavelet expansions. The coefficients of wavelet series are estimated under LDP. Our method prioritizes the estimation of low-order coefficients, in order to ensure accurate estimation at macroscopic level. Therefore, the probability mass is prevented from being misplaced too far away from its ground truth. We establish theoretical guarantees for our methods. Experiments show that our wavelet expansion method significantly outperforms existing solutions under Wasserstein and KS distances.

cs.LG↗

arXiv2025

Online Rounding Schemes for $ k $-Rental Problems

We study two online resource allocation problems with reusability in an adversarial setting, namely kRental-Fixed and kRental-Variable. In both problems, a decision-maker manages $k$ identical reusable units and faces a sequence of rental requests over time. We develop theoretically grounded relax-and-round algorithms with provable competitive ratio guarantees for both settings. For kRental-Fixed, we present an optimal randomized algorithm that achieves the best possible competitive ratio. The algorithm first computes an optimal fractional allocation using a price-based approach, and then applies a novel lossless online rounding scheme to obtain an integral solution. For kRental-Variable, we first establish the impossibility of achieving lossless online rounding. We then introduce a limited-correlation rounding technique that treats each unit independently while introducing controlled dependencies across allocation decisions involving the same unit. Combined with a carefully-crafted price-based method for computing the fractional allocation, this approach yields an order-optimal competitive ratio for the variable-duration setting.

cs.DS↗