arXiv subjects

Jun Chen

Publications and source records attributed to Jun Chen.

At least 19 recordsLinked to original sources

arXiv2026

V$^3$AST-Bench: Verifiable, Valid, and Versatile Safety Evaluation for Coding Agents

Large language models increasingly act through code, modifying repositories, executing programs, and interacting with services in agent form. Their actions introduce risks that cannot be assessed from generated responses alone. Safety evaluation must verify actual consequences, distinguish harm avoidance from task incapability, and cover realistic workflows. We introduce V$^3$AST-Bench, a benchmark for Verifiable results, Valid interpretation, and Versatile risk coverage of Agent Safety. It comprises 212 risk tasks spanning 18 categories. Dedicated runtime environments verify harm and task outcomes through executable checks over states, records, and artifact behavior, while paired capability twin tasks verify workflow competence. Our experiments span 51 combinations of 11 models and five harnesses, spanning 86,496 trials and over 9 billion tokens of model interaction. Results show that V$^3$AST-Bench remains challenging even for frontier coding agents, as Claude Opus 4.8 with Claude Code still causes verified harm on 20.9% of tasks. Across model-harness combinations, agents with comparable competence exhibit markedly different safety outcomes, showing that task capability alone does not guarantee safe execution. More importantly, even explicit recognition of risks does not reliably prevent harmful actions, revealing a critical gap between risk awareness and safe execution. These findings highlight the need to move beyond risk recognition toward enforcing safety throughout execution.

cs.SE↗

arXiv2026

On the Fundamental Limits of Bistatic ISAC Over Vector Channels

This paper studies the fundamental limits of bistatic integrated sensing and communications (ISAC) systems over vector channels. We first consider a class of parallel channels in which, on each component channel, the channel from the transmitter (Tx) to the communication receiver (Rx) is degraded with respect to the channel from the Tx to the sensing Rx, or vice versa, where the direction of degradation may vary across component channels. For this class of channels, we establish matching achievability and converse results, which yield a single-letter characterization of the capacity-distortion function $C(D)$. For parallel Gaussian channels with additive Gaussian states, the evaluation of the capacity-distortion-power function $C(D,P)$ under the squared error distortion measure reduces to a joint power and distortion allocation problem across the component channels. We then extend the analysis to general vector channels and derive single-letter lower and upper bounds on $C(D)$. In particular, for vector Gaussian channels with additive Gaussian states, we prove the optimality of Gaussian superposition coding without requiring a global degradation order between the communication and sensing channels, by combining the channel enhancement technique with a matrix extremal inequality based on Fisher information. This leads to an exact characterization of $C(D,P)$.

cs.IT↗

arXiv2026

FSVL: Taxonomic Fast-Slow Vision-to-Language Collaborative Framework for Interpretable Confusable Fish Species Identification and Knowledge Generation in the Middle Yangtze River

The Yangtze River Basin is a critical region for freshwater biodiversity conservation. The implementation of the Yangtze River Ten-Year Fishing Ban has increased the importance of effective aquatic biodiversity monitoring and species-level identification. However, fine-grained identification of closely related freshwater fish species remains challenging due to subtle morphological differences. Existing deep learning methods often struggle to balance classification accuracy and inference efficiency, while LLMs may generate unsupported information when producing specialized biological knowledge. To address these challenges, this study proposes FSVL, a fast-slow dual-pathway framework integrating visual recognition with LLM-based Retrieval-Augmented Generation (RAG). A lightweight ResNet50 fast path performs preliminary classification across 51 fish species in the Middle Yangtze River Basin and routes predefined groups of easily confused species to a slow path for refined identification. Guided by taxonomic criteria, the slow path uses YOLOv8n to detect key morphological regions, including the head, dorsal fin, and tail, leveraging morphometric and outline traits to improve discrimination among closely related species. The framework further integrates a retrieval-enhanced Qwen3-14B backend, using soft tokens and Low-Rank Adaptation (LoRA) for parameter-efficient domain knowledge generation. On the evaluated test set of 607 images, the fast classifier achieved 99.67% accuracy, whereas the cascade corrected the two errors, yielding 100.00% overall accuracy; further validation on independent specimens and external conditions remains needed. Across 255 knowledge generation samples, scientific name accuracy rises from 91.37% to 100.00%, genus accuracy from 90.20% to 98.82%, and ROUGE-L from 0.335 to 0.655, supporting intelligent biodiversity monitoring and fisheries management.

cs.CV↗

arXiv2026

KunSet: A Corpus and Benchmark for Lyrics Alignment and Rhyme Retrieval in Kunqu Opera

Kunqu is one of the oldest extant forms of Chinese opera, yet its digitization remains scarce and difficult to annotate, with two challenges standing out: first, close analysis of Kunqu performance demands syllable-level lyrics alignment; second, analyzing Kunqu pronunciation requires its phonological attributes (e.g., rhyme) to be linked with each syllable. To address these challenges, we present KunSet, the first Kunqu performance corpus comprising 482,164 fine-grained annotations of syllable-level lyrics and phonological attributes in total media duration of 155.2 h with 106.3 h annotated. To ensure annotation quality, KunSet was preannotated by an iterative ML-assisted pipeline and corrected by over 200 trained annotators via a purpose-built online annotation system under a two-round review process. To assess KunSet as a music information retrieval (MIR) benchmark, we evaluate two tasks: lyrics alignment, by finetuning a forced-alignment model with an acoustic anchor decoder; and same-rhyme retrieval, using frozen Chinese HuBERT features. Models built with KunSet outperform their baselines. We release KunSet and model checkpoints to support MIR research on traditional-music corpora.

cs.SD↗

arXiv2026

Neural Video Coding under Limited Delay: Theoretical Optimality and System Design

Traditional neural video compression (NVC) methods predominantly follow strictly causal predictive-coding paradigm, treating zero-frame delay as a fixed constraint rather than delay as an explicit design parameter. From an information-theoretic perspective, allowing limited frame delay enables a codec to exploit correlations with future frames within the delay window, thereby reducing uncertainty that is inaccessible to strictly causal prediction and potentially improving the rate-distortion tradeoff. In this work, we formalize this intuition through a rigorous rate-distortion analysis builds on the clique factorization of Markov networks, and propose Graph-Factored Sequential Coding (GFSC), a principled framework for sequential video coding under limited delay. To our knowledge, we provide the first theoretical result showing that, for general $K$-order Markov sources with positive distributions, a $K$-frame delay is sufficient to attain joint-coding rate-distortion optimality over a nontrivial distortion region, a guarantee unavailable to zero-delay causal coding in general. In practice, we instantiate GFSC through lightweight neural modules, allowing existing causal predictive codecs to be augmented without modifying their core architectures. Experiments across multiple NVC models and datasets show that GFSC consistently improves rate-distortion performance over the corresponding causal counterparts. Under matched delay budgets, GFSC also achieves competitive performance compared with representative bidirectional codecs.

cs.LG↗

arXiv2026

A Self-Coupling Gaussian Extremal Inequality for Information-Constrained Optimal Transport with Applications to Rate-Distortion-Perception

How much mutual information is needed to achieve a target correlation between two copies of a prescribed distribution? We prove a sharp Gaussian extremal inequality for this self-coupling problem: among distributions with a fixed mean and covariance, the Gaussian has the greatest minimum information cost for meeting a positive semidefinite cross-covariance target. The inequality is strict for non-Gaussian distributions when the target is positive definite. The proof uses a midpoint gluing construction. Two Ornstein--Uhlenbeck half-bridges, each started from the prescribed endpoint law, meet at a common intermediate distribution and are joined by conditional gluing. Changing the Gaussian reference then turns the resulting relative-entropy bound into the desired extremal inequality. As a first application, scaling one endpoint of the self-coupling establishes Gaussian extremality for rate-distortion-perception with unlimited common randomness, under quadratic distortion and squared Wasserstein-2 perception measures. We also establish an analogous result for finite common randomness, where the problem involves two mutual information constraints associated with an auxiliary random variable. In that setting, we retain an intermediate observation on one half-bridge and balance the two information budgets simultaneously using the Poincaré-Miranda theorem.

cs.IT↗

arXiv2026

StepAudio 3 Realtime Technical Report

Realtime spoken interaction demands deep reasoning, prompt responses, and fluid turn-taking. We present StepAudio 3 Realtime, an audio-language foundation model organized around a continuous listen-converse-think-act loop. Deep Perception captures rich acoustic cues to interpret user intent, while Seamless Duplex models synchronized audio streams to handle pauses, backchannels, and interruptions naturally. Crucially, we resolve the tension between deep deliberation and latency via Think-While-Speaking, executing private reasoning in parallel with spoken delivery. In reasoning mode, StepAudio 3 reaches a 73.0 macro average on StepAudioChat. With Think-While-Speaking, it achieves dialogue and reasoning performance comparable to dedicated reasoning models while speaking in real time. Furthermore, an integrated Voice Agent handles asynchronous tool execution without disrupting the dialogue flow. StepAudio 3 Realtime achieves top-tier performance across key dimensions: an exceptional 90.6 on the MMSU benchmark, 98.9 Overall on the Artificial Analysis Full-Duplex Bench, and a 56.0% macro task-success rate on $τ$-Voice.

cs.SD↗

arXiv2026

Learning from Totally Unpaired Data via Minimum Conditional Entropy Coupling

Consider an "unsupervised classification" setting, where the goal is to learn a predictor of the categorical class variable $X$ using the categorical feature variables $Y_1,\ldots,Y_m$, but we can only train the predictor on $m+1$ totally unpaired datasets, where each dataset only consists of unordered samples of one of the variables $X,Y_{1},\ldots,Y_{m}$. Although this problem is highly underdetermined, we demonstrate that, under the assumption that a predictor with small cross entropy loss exists, we can find a good predictor by coupling the marginal distributions of $X,Y_1,\ldots,Y_m$ in a manner that minimizes the conditional entropy $H(X|Y_1,\ldots,Y_m)$. This problem, which we call minimum conditional entropy coupling (MCEC), is a generalization of the widely studied minimum entropy coupling problem (where $m=1$). Although the MCEC problem is NP-hard, we give an algorithm with a time complexity that is polynomial in $N$, which gives a coupling with conditional entropy within $O(\log \log N)$ from the optimum, where $N$ is the dimension of the output coupling. We also evaluate the performance of the MCEC predictor on synthetic and real-world datasets.

cs.IT↗

arXiv2026

A multimodal large language model for evidence-based autism spectrum disorder screening

The clinical management of autism spectrum disorder (ASD) faces a bottleneck in early screening, mainly because trained specialists are scarce and conventional assessment tools are subjective. Here, we introduce ASDchat, a multimodal large language model designed for evidence-based ASD screening, which takes video, audio, and dialogue as input. ASDchat adopts a dual-branch architecture, where the decision branch generates screening probabilities and the evidence branch generates traceable, timestamped behavioral evidence aligned with the Autism Diagnostic Observation Schedule (ADOS-2). The model was trained and evaluated on a dataset of 1,035 participants from 27 sites in China, which covered typically developing (TD) children, children with ASD, and children with other developmental disorders. For ASD versus TD, ASDchat reached an area under the receiver operating characteristic curve (AUC) of 0.953 $\pm$ 0.021. On 6 held-out sites that were not used for training, the mean AUC was 0.961. Furthermore, unsupervised clustering of the behavioral dimensions split the ASD cases into six subtypes with different phenotypic profiles, and ASDchat suggests an intervention for each subtype. ASDchat provides a feasible path for large-scale, evidence-based ASD screening in clinical practice.

cs.CV↗

arXiv2026

NFTR: From Provable Mode-Averaging to Geodesic Subgoal Selection in Offline Goal-Conditioned RL

Hierarchical Implicit Q-Learning (HIQL), an offline goal-conditioned RL method, selects subgoals by value-function advantages alone. This rule has two coupled failure modes. Optimistic bias treats lucky stochastic outcomes as skillful choices, and mode collapse reduces a multi-modal subgoal distribution to a single Gaussian mean that often falls in unreachable regions. We propose NFTR (Normalizing Flow subgoal policies with Triangle-slack Reweighting). A conditional Normalizing Flow replaces the Gaussian policy. A closed-form mode-averaging result identifies the Gaussian limitation, while conditional flows support multi-modal weighted maximum likelihood and direct sampling. A triangle slack score, computed from a jointly trained MRN distance head whose triangle inequality is guaranteed architecturally, corrects the AWR weight multiplicatively and measures a detour inside the learned geometry without requiring exact distance recovery. Triangle-slack vanishes on geodesics in deterministic MDPs and remains a conservative upper bound on composability violation under stochastic dynamics. The RWDR objective preserves AWR's population-level monotonic improvement and admits a three-term suboptimality decomposition. On OGBench the flow carries the larger share of the gain, while the full co-trained configuration adds task-dependent gains. The combined method avoids Gaussian mode averaging and performs well on stochastic tasks. GitHub page: https://github.com/erdemtbao/NFTR

cs.LG↗

arXiv2026

On the Cardinality of Optimal Representations in the Binary-Source Information Bottleneck

The information bottleneck (IB) seeks a representation $U$ of a source $X$ that retains as much information as possible about a target $Y$, subject to a constraint on $I(U;X)$. A classical argument shows that it suffices to consider representations with at most $|\mathcal{X}|+1$ symbols, and this bound is known to be tight whenever $|\mathcal{X}| \geq 3$. We show that the binary case behaves differently: if $X$ is binary and $Y$ is finite, then for every joint distribution of $(X,Y)$ and every rate constraint, the IB optimum is attained by a binary $U$. Hence the bound $|\mathcal{U}| \leq |\mathcal{X}|+1$ sharpens to $|\mathcal{U}| \leq |\mathcal{X}|$ for binary sources. The proof combines a separating hyperplane argument with the observation that, for a binary source, the ratio of the second derivatives of the two entropy functions involved is concave.

cs.IT↗

arXiv2026

FastQSL 2: A Comprehensive Toolkit for Magnetic Connectivity Analysis

We present a new version of FastQSL for locating quasi-separatrix layers (QSLs) -- regions characterized by strong magnetic connectivity gradients, preferential current buildup, and subsequent magnetic reconnection. This version now supports spherical coordinates, utilizing a second spherical coordinate system for tracing magnetic field lines around the polar regions. This approach completely resolves the singularity problem at the two poles. Furthermore, our code accommodates arbitrary mesh shapes for output, can provide both magnetic field and electric current density on the mesh, and can save the traced magnetic field lines. We suggest using $Q_\mathrm{local}$ calculated through a localized mapping to locate (quasi-)separators. By quickly and accurately outputting the footpoint coordinates of magnetic field lines, FastQSL can be used to derive the two key parameters used for modeling solar wind speed and slip-squashing factors for the case of zero boundary flow. Compared with the first version, FastQSL 2 achieves significant improvements in terms of application scope.

astro-ph.SR↗

arXiv2026

VenusRL: A Fully Disaggregated Agentic RL System with Priority Scheduling and Scalable Interaction

Agentic Reinforcement Learning (RL) trains LLM agents through multi-turn interactions with external tool environments. Its multi-turn nature exposes two system-level bottlenecks unaddressed by existing agentic RL frameworks. First, end-to-end training throughput is constrained by the slowest trajectories to complete, yet optimizing per-GPU utilization alone scatters rollout progress across many groups, delaying the completion of enough groups to unblock the next training step. Second, tool sandboxes are statically over-provisioned by their declared memory ceilings, leaving most physical memory stranded while replicating near-identical state across sandboxes launched from the same prompt. We present VenusRL, a fully disaggregated agentic RL system that addresses both bottlenecks. VenusRL's priority-aware action scheduler uses length-prediction heuristics to identify sample groups whose completion is most likely to unblock the next training step, and pushes them ahead of others across batch admission, KV Cache residency, and cross-worker request orchestration. VenusRL's environment resource manager combines a memory-aware admission threshold with a template-keyed page-sharing pool, packing more sandboxes per node while preserving strict memory isolation via write-protected page table entry aliasing and copy-on-write. Across representative agentic RL workloads, VenusRL achieves up to 4.24x end-to-end training speedup over state-of-the-art baselines and reduces environment cost by up to 89%.

cs.DC↗

arXiv2026

Clipped Affine Policy: Low-Complexity Near-Optimal Online Power Control for Energy Harvesting Communications over Fading Channels

This paper studies online power control for battery-limited point-to-point energy harvesting communications over slow block-fading channels. A linear-policy-based approximation is developed for the relative-value function in the Bellman equation of the power control problem. This approximation leads to two fundamental parameterized clipped affine policies: an optimistic policy derived from a certainty-equivalence-type approximation and a robust policy derived from worst-case analysis. For independent and identically distributed energy arrivals and channel states, two families of power control schemes are developed based on the optimistic clipped affine (OCA) and robust clipped affine (RCA) policies, respectively. The proposed adaptive RCA policy based on reinforcement learning (RCA-RL) is further extended to address four scenarios with contextual information: one-step energy lookahead, one-step channel lookahead, one-step joint energy-channel lookahead, and Markov energy arrivals. Extensive simulation results show that the proposed schemes provide a favorable tradeoff between computational complexity and performance. The adaptive RCA policy based on the maximin optimal linear-policy-slope approximation (RCA-OLA-A) and the RCA-RL scheme achieve the best overall performance, while the RCA policy based on the maximin optimal linear policy (RCA-OL) is the best-performing closed-form policy. In particular, RCA-OLA-A, RCA-RL, and the aforementioned RCA-RL extensions achieve less than 2% performance loss relative to the optimal policy across a range of scenarios, consistently outperforming the considered benchmark approaches, including generic reinforcement learning baselines. The RCA-OL policy also performs well with less than 4% performance loss.

cs.IT↗

arXiv2026

Semantic Projection for Continual Self-Evolution of Language Agents

Language-model agents increasingly rely on persistent natural-language skills to adapt beyond their frozen model parameters. When a shared skill is repeatedly revised from a non-stationary, heterogeneous task stream, however, improvements for new tasks can overwrite procedures needed for earlier ones. In continual learning, Orthogonal Gradient Descent (OGD) addresses analogous interference by projecting a new-task gradient onto a subspace that locally preserves prior predictions. Natural-language skill revisions, however, have neither gradients nor a canonical vector space in which such a projection can be performed. We introduce \emph{Semantic-Scope Projected Evolution} (SSPE), which transfers the functional principle of gradient projection from parameter space to behavior space. SSPE treats an unconstrained skill revision as a proposed update, identifies acquired capabilities with which it may interfere, and uses the observed gains and regressions to construct a compatible revision rather than merely rejecting the update. This enables one shared skill to evolve across latent and recurring task contexts without exposing semantic domain identities to the evolution model. Across controlled synthetic streams and heterogeneous real-agent benchmarks, SSPE improves final cross-domain competence and mitigates forgetting relative to strong skill-evolution baselines. The evolved skill also retains the strongest average performance after transfer to a different executor model. These results establish semantic projection as a promising principle for stable and adaptive self evolution of language agents.

cs.AI↗

arXiv2026

Integrated Thermal and Power Management for Wave-Powered Subsea Data Centers via Nonlinear Model Predictive Control

This paper develops an integrated modeling and nonlinear model predictive control (NMPC) framework for coordinating thermal management, flexible workload scheduling, wave-power utilization, and battery operation in a wave-powered subsea data center. Realistic data center workloads are constructed from job-level CPU, memory, and GPU measurements from the MIT Supercloud dataset and divided into interactive and delay-tolerant flexible jobs. Thermal behavior is represented by a three-node lumped model of the IT equipment, recirculating nitrogen, and pressure hull with surrounding seawater as the thermal boundary. The NMPC jointly optimizes the flexible workload power budget and cooling command subject to thermal, battery, and workload constraints. Closed-loop simulations under different workload, thermal, battery, and renewable-generation conditions demonstrate that the proposed framework maintains thermal safety while adapting cooling operation and flexible workload execution to wave-power availability and battery state-of-charge. The parametric studies show that battery capacity and wave-generation capacity strongly affect battery availability and flexible-workload queue accumulation, while excessive renewable generation capacity may lead to increased energy curtailment. Monte Carlo and distance-correlation analyses further show that flexible-job delay is relatively insensitive to the investigated system parameters, whereas terminal battery state-of-charge is primarily influenced by battery energy capacity and wave generation capacity.

eess.SY↗

arXiv2026

Local stability of direct Mach configurations

We investigate the local stability of direct Mach configurations for the two-dimensional steady compressible Euler equations. Given a piecewise constant Mach configuration that consists of an incident shock, a reflected shock, a Mach stem and a slip line, with subsonic downstream states, we prove the following: under a transversality condition on the shock polar loops and a suitable condition on the truncation segment, every sufficiently small perturbation of the incoming flow gives rise to a Mach configuration that is a small perturbation of the background one, in the sense that the reflected shock, the Mach stem and the slip line, together with the downstream subsonic flow, stay close to their background counterparts; moreover, the solution is unique in the class of solutions satisfying the a priori estimate (for incoming flows that are small in a slightly stronger norm). Lagrangian coordinates are employed to transform the unknown contact discontinuity curve into a fixed boundary and to reduce the full Euler system to a first-order elliptic system for the flow direction and the pressure. The key idea in dealing with the contact discontinuity is to solve a mixed boundary value problem, in divergence form with discontinuous coefficients, for a single elliptic equation, so that the contact discontinuity conditions are naturally preserved as compatibility conditions for the solutions of the elliptic problem.

math.AP↗

arXiv2026

RLVR is a Kernel, Not a Function: Statistical Inference for pass@$k$ Crossovers

Reinforcement learning with verifiable rewards (RLVR) often improves pass@1 while falling behind its base model at larger sampling budgets $k$, a crossover read as evidence that RLVR only sharpens existing capability. We identify two limits to this reading. First, a visible crossing need not be statistically established: comparing models on the same prompts, we build confidence bands across sampling budgets $k$ that require evidence of both an early gain and a later loss. Across five public RLVR pairs no crossing is statistically established in the initial evaluations, while a 32k-token evaluation on fresh prompts locates a reversal with first loss between 11 and 61 samples; power analysis shows why failure to detect a crossing need not mean no crossing, and why more prompts can help more than more answers per prompt. Second, base success alone does not determine what RLVR does to a prompt: prompts with the same base success rate have different post-RL success rates, and these differences repeat across independent generation halves. The relationship is a conditional distribution---a Markov kernel---rather than a single curve, and fitting it predicts crossings in independent generations for the same prompts and corrects the simple model's power estimates. Theory further shows how losses on a minority of the hardest prompts can overturn an early lead even when training improves other prompts, separating evidence that a crossover exists from claims about what it means for capability.

stat.ML↗