Search arXiv⌕ Search

arXiv subjects

Hanwen Zhang

Publications and source records attributed to Hanwen Zhang.

At least 19 recordsLinked to original sources
arXiv2026

Stacking Stars: Minimum-Area Packing with and without Rotation

We study the problem of packing a given set of polygons into a minimum-area convex container, that is, placing them without overlap so as to minimize the area of their convex hull. Every previously known polynomial-time algorithm with a nontrivial approximation guarantee for packing polygons with arbitrary edge orientations required the polygons to be convex. In fact, an algorithm was not even known when every input polygon was a quadrilateral. We obtain such algorithms for the entire class of star-shaped polygons with unrestricted combinatorial complexity. The stars may be just translated or also rotated. In either case, we give an $O(\log n)$-approximation algorithm, where $n$ is the number of stars. Likewise, we get $O(\log n)$-approximation algorithms for bin and strip packing, assuming that the diameter of the stars is at most some (small) constant. Previous approximation algorithms for polygon packing focused on convex objects, and the analysis always relied on a density argument, i.e., if the area of the packed objects is at least a fraction $ρ$ of the area of the container, then we have a $(1/ρ)$-approximation. However, star-shaped polygons may need an arbitrarily large area to be packed, even when the total area of the stars approaches $0$ and each star has bounded diameter. This phenomenon prevents the use of density arguments. We instead establish a strong structural property about any packing: Even though the stars can fit together in complicated ways, we can reorganize them into simple stacks, replace these stacks by rectangles, and consider a packing of these rectangles. The rearrangement costs only a factor of $O(\log n)$, which stems from the number of size groups. We complement our algorithms by showing that both the translational and the rotational versions are APX-hard, even for constant complexity stars.

cs.CG↗
arXiv2026

Task-Performance Routing for Multi-Teacher Distillation in CT Medical Image Segmentation

Pre-trained models offer complementary strengths in CT organ segmentation, but their relative accuracy varies across organs and spatial regions. Multi-teacher knowledge distillation can combine these strengths in a single student, yet global or instance-level teacher weights overlook variation within a 3D volume. We propose Task-Performance Routing (TPR), which uses region-wise segmentation errors to determine teacher weights during training without learned routing parameters. In hard regions, TPR favors teachers with lower prediction errors; in easy regions, it uses uniform weights to combine their knowledge. A hierarchical spatial partition provides fine-grained routing in shallow encoder stages and coarse-grained routing in deep stages. Experiments on BTCV, MSD-Liver, and MSD-Pancreas with two teacher pairs and two backbone architectures show higher Dice scores than the individual teachers, adaptive ensembles, and evaluated multi-teacher distillation baselines. On MSD-Pancreas, TPR achieves Dice scores of 65.40% with SwinUNETR and 68.25% with U-Net, compared with 64.38% and 67.45% for the corresponding adaptive ensembles. Inference uses the student alone, with no routing or teacher overhead. Code is available at https://github.com/EdvinCecilia/TPR_MultiTeacher.

cs.CV↗
arXiv2026

Second-Order Steps, Spectral Accuracy: Discrete Parallel Transport of Invariant Subspaces

Eigensolvers return orthonormal bases of invariant subspaces, but each basis is determined only up to a unitary factor, so bases computed at different parameter values cannot, in general, be interpolated directly. Discrete parallel transport resolves this ambiguity by aligning consecutive bases through the polar factors of their overlaps, but approximates the continuous parallel-transported frame only to second order. We show that the resulting frames can nevertheless be interpolated with spectral accuracy. For analytic families of invariant subspaces sampled at $N+1$ Chebyshev nodes, the computed frames are exact samples of an analytic frame related to the parallel-transported frame by a gauge that remains within $O(N^{-2})$ of the identity throughout a fixed complex neighborhood, uniformly in $N$. Consequently, Chebyshev interpolation approximates this analytic frame with geometric convergence. Moreover, the continuous parallel-transported frame can be recovered to spectral accuracy by solving a small matrix differential equation, without additional eigensolves or derivatives of the underlying matrix family. The results do not require the family to be Hermitian and remain valid through eigenvalue crossings and branch points, provided the selected invariant subspace depends analytically on the parameter. Numerical examples confirm the predicted rates.

math.NA↗
arXiv2026

REACT: Representational Early Analysis and Control of Chain-of-Thought Trajectories

Chain-of-thought (CoT) generation offers an opportunity to intervene before a reasoning model completes its answer. But does the point at which risk is easiest to detect also offer the most effective control? We study this question with REACT, a conditional activation intervention that learns a safety direction from paired safe and unsafe trajectories and applies it once to an early CoT hidden state. The model weights remain frozen, and all subsequent tokens are generated without further edits. To study timing, we train directions separately at each of the first five CoT positions and evaluate ungated edits with full post-edit regeneration. Across three distilled reasoning models and three generation seeds, the strongest gains concentrate in a two- to three-token early window; on DeepSeek-R1-Distill-Llama-8B, the effect falls to approximately zero at the fourth and fifth tokens, even though later states can be more predictive of risk. This separates risk readability from effective control: intervention timing must be evaluated through its effect on subsequent generation. On the same 8B model, REACT raises the rate of fully safe CoT-answer outputs on StrongREJECT-273 from 29.06% to 82.05%, while keeping benign refusal and usefulness near baseline. An anonymized implementation is available at https://anonymous.4open.science/r/REACT_official2-333B.

cs.LG↗
arXiv2026

EvoSteer: Online Self-Evolving Graph Orchestration via Reference-Anchored Credit Assignment

In recent years, LLM-based multi-agent systems have been widely applied to orchestrate tool-using agents into executable communication graphs. However, existing self-evolving orchestration still faces key challenges, including post-hoc evolution that revises the team only after the trajectory ends, credit diffusion that gives every action the same terminal advantage under confounded baselines, and skill admission that is uncalibrated and never retired. To address these challenges, we propose EvoSteer, a new paradigm of Online Self-Evolving Graph Orchestration -- the orchestrator builds a running team and repairs its plausible but failing steps from execution features and a learned value estimate. To support this paradigm, we introduce Anchored Trajectory Balance (AnchorTB), a regression-style flow-matching loss that assigns each orchestration action a coefficient by balancing subtrajectories against a frozen reference. Built on the learned flow, we further propose Validated Skill Admission, in which a candidate skill is tried before promotion and promoted only if paired evidence passes a sequential test under a shared nominal testing budget. Moreover, AnchorTB combines measured task-level reference reward statistics with prefix-dependent corrections. Experimental results on twelve datasets show that EvoSteer significantly outperforms baselines across question answering, mathematical reasoning, code generation, and interactive decision making. Our code is available at https://github.com/beita6969/evosteer.

cs.AI↗
arXiv2026

CollabFlow: Recursive Self-Improvement of Agent Collaboration

Recursive self-improvement (RSI) lets a system improve from its own outcomes; in LLM-based multi-agent systems, Agents refine one another within a task, and outcomes improve how they collaborate across tasks. However, existing multi-agent collaboration leaves this loop open: collaboration is pre-defined at the operator level, topology-only learning keeps verbatim exchange that propagates errors, and reward maximization on a system's own outcomes concentrates on a few teams. To address these challenges, we propose CollabFlow, an RSI system of Learned Agent Collaboration: a trainable Collab-Director constructs teams of complete Agents, a frozen executor runs them, and each round's outcomes retrain the director. Within each round, the edges of a collaboration graph carry protocols of Evidence-Conditioned Communication: a receiver adopts a differing answer only when the sender's evidence is stronger by a margin, so the director learns who communicates and how. Across rounds, we further propose Collaborative Trajectory Balance (CTB), a flow-based objective that credits each team once across its construction orders and targets a reward-proportional distribution over teams, so several good teams stay in play. We also bound how far this self-generated target moves between rounds, which shrinks as records accumulate. On twelve datasets, CollabFlow outperforms all baselines and keeps improving across rounds. Code is available at https://anonymous.4open.science/r/CollabFlow-631E.

cs.MA↗
arXiv2026

Retrievable Gradients: Continual Post-Training Without Cumulative Weight Drift

Continual post-training enables models to absorb emerging knowledge after deployment, but repeatedly updating shared parameters can accumulate weight drift, potentially causing catastrophic forgetting and degrading general capabilities. Retrieval-augmented generation avoids such parameter drift, yet often lacks the depth of parametric knowledge integration. In this paper, we propose ReGrad (Retrievable Gradients), a new paradigm that treats gradients as retrievable units of knowledge. ReGrad pre-computes document-specific gradients offline, stores them in an indexed Gradient Bank, and retrieves only query-relevant gradients at inference time for temporary weight adaptation. However, raw language-modeling gradients are optimized for token-level document reconstruction rather than for query-driven knowledge use. We therefore introduce a bi-level meta-learning objective that reshapes document-derived gradients into generalizable adaptation signals for downstream tasks. Experiments across general and domain-specific settings show that \textsc{ReGrad} outperforms CPT and RAG baselines, enabling scalable and reversible parametric knowledge injection without accumulating weight drift.

cs.CL↗
arXiv2026

Understanding Verification Vulnerabilities in Cross-chain Bridges: An Empirical Study

With the rapid development of blockchain technology, numerous heterogeneous blockchains have emerged, each featuring distinct architectures, consensus mechanisms, and application ecosystems. Cross-chain bridges have consequently become a critical infrastructure component, typically relying on smart contracts to facilitate message passing and asset transfers across heterogeneous blockchains. Unlike single-chain contracts, bridge contracts introduce additional verification logic to validate external-chain data, such as state proofs, message sender identities, and signatures. However, correctly implementing such logic is challenging. We refer to vulnerabilities that violate the security guarantees of cross-chain verification as Cross-chain Verification Vulnerabilities (CVVs). To characterize CVVs, we conduct the first systematic empirical study of this vulnerability class. By analyzing 588 security audit reports from 36 real-world cross-chain bridge projects, we derive 11 CVV patterns and organize them into 5 high-level categories according to the affected verification properties. Following this taxonomy, we construct a CVV dataset containing 830 vulnerability instances across 786 contract files. The dataset establishes mappings among audit reports, audited source-code versions, affected source files, and CVV-pattern labels. We then assess existing detection approaches. Current tools provide limited documented coverage of CVV patterns, while only two of four evaluated LLMs outperform an all-pattern constant baseline of 29.3% Macro-F1; the best reaches 51.2%. Together, the limited tool coverage and inconsistent LLM gains motivate detection methods tailored to cross-chain verification semantics.

cs.CR↗
arXiv2026

A Unified Score Matching Paradigm for Video Anomaly Detection and Anticipation

Video anomaly detection (VAD) is a fundamental and safety-critical task in computer vision. Recent generative approaches detect anomalies from a distributional perspective, but remain limited by local anomaly modes. Meanwhile, video anomaly anticipation (VAA), as a proactive extension beyond post-hoc detection, introduces additional challenges. In particular, the contrastive inference paradigm in VAD, which relies on ground-truth frames, is not applicable to VAA, hindering its development. To address these challenges, we propose a unified score-driven framework, termed Uni-DSM, based on denoising score matching (DSM), which models anomaly patterns through likelihood estimation and score functions over the learned data distribution. Within this unified framework, we adopt a shared noise-conditioned score transformer backbone with scene-dependent embeddings and motion-aware weighting for distribution-level modeling. Instead of introducing separate architectures, Uni-DSM unifies VAD and VAA through different inference and supervision paradigms built upon the same score-based formulation. For VAD, we instantiate an autoregressive denoising score matching (ADSM) mechanism, which progressively accumulates anomalous evidence via autoregressive denoising, enabling enhanced perception of local modes beyond visual cues. For VAA, we extend the same architecture by incorporating a lightweight auxiliary decoder and a novel self-distilled denoising score matching (SDSM) mechanism. By constructing supervision from output discrepancies instead of relying on unavailable future ground truth, our method achieves efficient training suitable or early anomaly anticipation. Extensive experiments on multiple benchmark datasets demonstrate state-of-the-art performance in both VAD and VAA while maintaining high efficiency, establishing a unified and scalable pipeline from anomaly detection to anticipation.

cs.CV↗
arXiv2026

An Agentic AI Framework with Large Language Models and Chain-of-Thought for UAV-Assisted Logistics Scheduling with Mobile Edge Computing

In cloud manufacturing, unmanned aerial vehicles (UAVs) can support both product collection and mobile edge computing (MEC). This joint operation forms a hybrid scheduling problem, where physical logistics decisions are coupled with computational task scheduling. In this paper, UAVs collect finished products from manufacturing stations and transport them back to a central depot. Meanwhile, computational tasks generated by industrial sensor devices at these stations are processed locally, at UAVs, or offloaded via UAVs to the cloud. This coupling makes the problem challenging. A UAV can provide MEC services only during its service window at a station, so routing decisions directly determine when UAV-assisted offloading is available. Routing decisions also affect the UAV energy budget and the availability of onboard computing and communication resources for computational task execution under task deadline constraints. To address this, we propose an agentic-AI-assisted optimization framework with two components. First, we develop an agentic AI that combines large language models, retrieval-augmented generation, and chain-of-thought reasoning to translate user input into an interpretable mathematical formulation for the hybrid scheduling problem. Second, we design a hierarchical deep reinforcement learning approach based on proximal policy optimization (PPO), where the upper layer learns UAV routing and the lower layer optimizes per-slot task execution and resource allocation. Simulation results show that the proposed framework yields more consistent formulations, while the hierarchical PPO achieves full product collection in 99.6% of the last 500 episodes and maintains a 100% deadline satisfaction rate, with more stable performance than the advantage actor-critic approach.

cs.AI↗
arXiv2026

Fast Evaluation of the Azimuthal Fourier Modes of the 3D Helmholtz Green's Function and Their Derivatives

We introduce an $O(M)$ algorithm for evaluating the azimuthal Fourier modes $G_{k,m}$, $m = 0, 1, ..., M$, of the three-dimensional Helmholtz Green's function with real wavenumber $k$, together with all their first- and second-order derivatives with respect to the cylindrical source and target coordinates. The cost is independent of both the wavenumber and the source-target separation, and high relative accuracy is retained even for modes whose magnitude is exponentially small. The method combines contour deformation at a few boundary modes with a boundary-value formulation of the five-term recurrence in the mode index. Derivative quantities are obtained from stable recurrences, adding only a small constant factor to the cost of $G_{k,m}$ alone. Numerical experiments demonstrate high relative accuracy, linear scaling in $M$, and applications to modal boundary integral equation solvers for axisymmetric acoustic scattering, where the $k$-independent kernel evaluator makes dense per-mode linear algebra the dominant cost.

math.NA↗
arXiv2026

PhysBrain 1.5: From Vision-Language Models to Physical Foundation Models

We present PhysBrain 1.5, a unified model for understanding physical environments, generating actions, and predicting future states. Motivated by the physical loop of observation, interaction, and environmental change, we bring these capabilities into a common learning framework. Starting from a general vision--language model, we encode language responses, end-effector motion, and dense visual targets as discrete sequences and jointly optimize them with autoregressive next-token prediction. Pre-training draws its embodied supervision entirely from human interaction videos, using task-centered episodes to pair semantic and spatial context with recovered motion and subsequent observations. We then adapt the model through supervised fine-tuning on a mixture of human demonstrations, robot trajectories, and simulated experience. Across 28 embodied understanding benchmarks, our 8B model achieves an average score of 72.5, setting a new open-source state of the art and performing on par with leading proprietary models such as GPT-6-Astra and Gemini 3.6 Flash. It achieves the best open-source results on 14 benchmarks while retaining general multimodal capabilities. Beyond these understanding evaluations, qualitative examples show the model's ability to produce end-effector trajectories and predict future scenes through spatially aligned RGB, depth, and robot-mask outputs.

cs.CV↗
arXiv2026

NemoSplat: Feed-Forward 4D Gaussian Splatting for Media-Aware Underwater Reconstruction

Reconstructing photorealistic scenes in unconstrained underwater environments remains challenging due to severe media-induced light scattering and unpredictable dynamic objects. Recent feed-forward visual foundation models have demonstrated remarkable capabilities in generalized novel view synthesis and tracking. However, when directly applied to aquatic videos, optical attenuation and motion interference fatally corrupt their feature aggregation, leading to severe tracking and reconstruction failures. To overcome these limitations, we present NemoSplat, the first feed-forward 4D Gaussian Splatting framework tailored for media-aware dynamic reconstruction directly from uncalibrated marine videos. Beyond providing robust estimations of camera poses and dense scene depth, we devise a Promptable Dynamic Disentangler that utilizes a confidence-aware fusion strategy of learned dynamic probabilities and optional semantic text priors, effectively isolating massive transient entities. Furthermore, to counteract visual degradation, a Media-Aware Gaussian Predictor is formulated to jointly estimate intrinsic 3D Gaussian attributes alongside physical media parameters, rendering pristine scene appearance in a single forward pass. Additionally, we introduce a large-scale underwater dataset with massive dynamic elements to facilitate training and evaluation. Extensive experiments on our dataset demonstrate that NemoSplat achieves state-of-the-art tracking accuracy and high-fidelity rendering. Homepage: https://nemosplat.hkustvgd.com

cs.CV↗
arXiv2026

Physiological World Models for Human State Transitions

Continuous multimodal sensing now allows human physiology to be observed throughout daily life rather than only during occasional clinical visits. However, most health artificial intelligence systems are designed to recognize current states, estimate risks or analyse individual biomarkers. They do not directly model how physiological states change in response to real-world events, behaviours, contexts and interventions. Here we propose the Physiological World Model (PWM), an event-conditioned framework for learning these changes at the level of the whole person. We introduce the HumanState Transition Token, a structured, quality-scored unit that connects the physiological state before an event with the event or action, relevant context and intervention information, the physiological trajectory after the event, observed outcomes and data quality. We describe four capability levels, from state representation to bounded intervention planning, together with four data acquisition and validation protocols. We also propose six benchmark tasks covering HumanState representation, forecasting across multiple timescales, individualized response prediction, simulation of alternative interventions, bounded planning and reliability under distribution shift. Together, this framework provides a practical path towards personalized health management, behavioural intervention design and clinician-supervised decision support, while clearly separating prediction from causal inference and making uncertainty, safety, governance and limits of use explicit.

cs.AI↗
arXiv2026

Reifying Research Logic: AI-Assisted Workflow Construction and Incremental Refinement for Quantitative Syntax

Quantitative language research often depends on long chains of computational steps, yet the logic connecting those steps usually remains buried in scripts. This makes analyses harder to inspect, share, and revise than they need to be. Focusing on quantitative syntax, we present QLWF, a visual workflow platform that turns natural-language research descriptions into executable workflows through an AI assisted five-stage pipeline. In this setting, reification makes the research logic visible as a workflow, while formalization gives that workflow deterministic execution semantics. The language model is used only during construction. Execution is handled by a fixed node library and engine, which keeps the resulting workflows reproducible. QLWF also supports incremental refinement, so saved workflows can be revised by changing only the parts that need to change rather than being rebuilt from scratch. To evaluate the approach, we build a 64-task benchmark called QL-Bench from the quantitative-syntax literature. Across three runs, QLWF produces structurally valid and executable workflows for every task and reaches a mean output-plausibility rate of 98.4%, well above the prompt-based baselines. On a separate 12-task lifecycle benchmark, this refinement process succeeds in every case and uses roughly one-third of the tokens required by full regeneration. The paper also releases the node library, benchmark, workflow templates, and platform as reusable resources for quantitative-syntax research.

cs.MA↗
arXiv2026

Constraint-First Reasoning: A Training-Free Protocol for Exploiting Answer-Space Constraints in Mathematical Problem Solving

Large language models can derive a plausible mathematical object yet still violate explicit requirements--for example, by omitting a modular reduction, returning a non-integer, or using the wrong encoded answer form. We introduce Constraint-First Reasoning (CFR), a training-free two-stage prompting protocol: Stage 1 extracts and summarizes constraints entailed by the problem, and Stage 2 solves while checking intermediate and final results against that summary. Routed-CFR activates the two-stage protocol only when a text-only regex router detects restrictive cues; otherwise it uses direct chain-of-thought (CoT). Across AIME, CMIMC, BRUMO, and AIMO_AMC, the method improves direct CoT on multiple backbones. We further report convention-controlled routing experiments, matched prompting baselines, problem-level paired tests, decoding robustness, constraint-quality audits, total-token accounting, and an OlympiadBench evaluation. These analyses position CFR as a targeted test-time intervention whose benefit depends on recoverable constraints and reliable Stage 1 extraction, rather than as a general-purpose replacement for mathematical reasoning.

cs.CL↗
arXiv2026

Nautilus: A Verifiable Hierarchical Federated Learning Framework for Vehicular-Edge-Cloud Systems

Federated Learning (FL) enables privacy-preserving collaborative learning for Internet of Vehicles (IoV) scenarios, but extreme heterogeneity of vehicular-edge-cloud resources severely limits system efficiency. Dynamic scheduling strategies mitigate this issue but introduce new trust concerns: verifying fair scheduling decisions and faithful client execution of compression instructions without privacy leakage remains an open challenge. We propose Nautilus, a verifiable efficient federated learning framework. First, a multi-dimensional resource-aware scheduling algorithm dynamically allocates compression ratios and training tasks based on vehicle bandwidth, latency and computing power, improving training efficiency. Second, a Zero-Knowledge Proof (ZKP) mechanism ensures scheduling fairness and execution compliance while preserving privacy. Experiments show the framework reduces communication overhead and accelerates convergence with guaranteed system integrity.

cs.DC↗
arXiv2026

Weak-$A_\infty$ packing and estimates for fractional sparse forms on spaces of homogeneous type

We prove two-weight estimates for fractional sparse forms on spaces of homogeneous type. The relevant transformed weights are assumed to belong to weak-$A_\infty$. The weak reverse Hölder inequality is normalized on enlarged balls, while sparse testing is normalized on dyadic cubes. To measure this loss, we introduce $\mathfrak d_Q^w:=w(Q)/w(2ΛB_Q)$. A packing estimate for the defect levels, combined with two-weight testing, gives the required mixed weak-$A_\infty$ bounds.

math.CA↗