Search arXiv⌕ Search

arXiv subjects

Haoxiang Zhang

Publications and source records attributed to Haoxiang Zhang.

At least 19 recordsLinked to original sources
arXiv2026

Flux-Matched Poisson Iteration for Generated Jacobian Equations: Local Convergence and Spectral Galerkin Approximation

We develop a flux-matched Poisson iteration for solving generated Jacobian equations with second boundary conditions. In each step, it solves a fixed shifted Poisson problem with a Neumann-type boundary condition. Near a smooth regular solution with positive $C^{1,1}$ densities, we prove that the iteration has local geometric convergence in $C^{2,α}$ on smooth connected Riemannian domains. The iteration is then discretized using a spectral Galerkin method, with local convergence proved for the Fourier-Jacobi scheme on spherical caps. Optical numerical experiments support these estimates and demonstrate useful imaging performance beyond the local convergence theory.

math.NA↗
arXiv2026

Robot Navigation and Canine Welfare: A Pilot Study Surfacing Challenges for Cross-Species Legibility and Proxy Validity in Animal-Machine Interaction

Autonomous delivery robots increasingly share residential spaces with companion animals, yet navigation behavior is rarely designed with animal welfare in mind. This paper presents a pilot study evaluating a curve-out maneuver, a laterally arcing pass-by trajectory triggered at the dog's alert distance, proposed as a candidate welfare-informed navigation intervention grounded in canine social communication research and legible motion theory. Following a research through design approach, we developed and evaluated a first design iteration of the artifact in a naturalistic home environment with a companion dog. Contrary to our hypothesis, the curve-out produced a more acute stress response than the straight-line baseline. We argue that discontinuous execution limitation, the robot stopping and reorienting multiple times to approximate the arc, undermined legibility, which in itself became a stressor. The pilot surfaces the legibility of robot behavior to non-humans' eyes and proxy stimulus validity as structural challenges for AMI and repositions the design problem for future work.

cs.HC↗
arXiv2026

GTSM: Gaussian Splatting with Thermal State-Space Dynamics for Spatio-Temporal Thermal Field Reconstruction and Prediction

Temperature-field reconstruction and prediction are important for condition monitoring, process analysis, and state assessment in industrial and dynamic environments. This work aims to recover a four-dimensional spatio-temporal thermal state field directly from two-dimensional observations, without requiring a predefined 3D model, prescribed thermodynamic equation, or predefined heating/cooling trajectory. To this end, we propose the Gaussian Thermal State-Space Model (GTSM), which represents thermal evolution through a shared scene-level latent process and Gaussian-specific local thermal states. The resulting continuous dynamics are propagated using differentiable RK2 integration, allowing thermal evolution to be inferred from multi-view and multi-temporal observations and continuously extrapolated beyond the observed interval. GTSM is designed as a backbone-agnostic thermal dynamics module for Gaussian representations and can be integrated with both static 3D Gaussian Splatting and dynamic 4D Gaussian Splatting, supporting thermal reconstruction and prediction under both static and time-varying target geometry. The reconstructed representation further enables novel-view thermal rendering and future-time thermal-state prediction. Experiments on real infrared sequences and synthetic dynamic thermal data demonstrate improved reconstruction accuracy and substantially stronger temporal extrapolation than representative baselines, while also validating the applicability of GTSM to deforming targets.

cs.CV↗
arXiv2026

Self-Retrospection Distillation: Turning Post-hoc Experiences into Prior Foresight

Reinforcement learning with verifiable rewards (RLVR) turns agent experience into learning signals primarily through scalar outcome rewards after interaction. For group-relative objectives, however, this signal vanishes when all rollouts receive the same reward, even though their trajectories may reveal useful information about what the task requires and how the agent fails. We ask a complementary question: can hindsight teach an agent what it could have anticipated before acting? We introduce prospective learning, which uses post-hoc experience to supervise foresight predictions from the pre-interaction view, and instantiate it with Self-Retrospection Distillation (SRD). Intuitively, a completed trajectory reveals knowledge that would have been useful and pitfalls that should be avoided; SRD distills this privileged hindsight into trajectory-blind foresight of the same policy. Foresight serves only as a training target and need not be explicitly generated at inference time. Across 10 tool-integrated reasoning and long-horizon agentic tasks, SRD complements RLVR and self-distillation baselines with gains of up to 24.2 pp. Its advantage is especially pronounced when reward contrast is scarce: when 37--98% of rollout groups are reward-uniform across model scales, yet SRD can still exploit learning signal from sampled trajectories. In the 2B setting, where 98% of groups are all-failure, the RLVR training ends up at 0.0% success, while adding SRD reaches 60.6% under the same rollout budget. Our results suggest that post-hoc agent experience is useful not only for evaluating or improving behavior, but also for shaping predictive representations before available interaction. Training code can be found at https://github.com/SalesforceAIResearch/SRD.

cs.AI↗
arXiv2026

Information Gain-based Rollout Policy Optimization: An Adaptive Tree-Structured Rollout Approach for Multi-Turn Search Agents

Reinforcement learning (RL) improves large language model (LLM) agents on long-horizon search tasks that require multiple intermediate decisions before a final outcome. However, rollout budgets are often allocated without assessing intermediate-state utility, which can waste computation on unpromising branches. We propose Information Gain-based Rollout Policy Optimization (IGRPO), a framework that organizes rollout collection around intermediate-state informativeness. Specifically, IGRPO performs budget-aware tree-structured rollouts in which expansion probabilities depend on node-level informativeness, allowing informative branches to receive more computation while less informative branches are expanded less frequently within a fixed rollout budget. By directly shaping how training trajectories are generated, IGRPO induces a limiting teacher distribution over search trajectories that favors higher cumulative informativeness. The resulting distribution provides an explicit policy optimization target, connecting adaptive rollout collection with principled policy learning. Experiments on seven search-augmented question answering benchmarks show that IGRPO achieves higher average accuracy than strong baselines on both 3B and 7B backbones under comparable rollout budgets, supporting informativeness-guided trajectory generation for training search agents. Code is available at https://github.com/e3trange/IGRPO.

cs.AI↗
arXiv2026

CoTrace: Data Recipes for Training Terminal Agents with Harness-Model Co-Evolution

Terminal-agent capability depends jointly on model weights and the runtime harness that formats prompts, binds tools, and handles error recovery. Existing harness-model co-evolution approaches improve both components, yet often treat trajectories produced during harness search as an undifferentiated replay buffer. This practice overlooks that a trajectory's value for model training depends on the harness under which it was generated. To systematically analyze this interface, we establish an alternating co-evolution framework that decouples harness search and policy training through component-wise promotion decisions. Within this framework, we introduce CoTrace, a harness-aware data recipe that explicitly governs trajectory routing, provenance matching, and curriculum refresh. Under CoTrace, recurring execution failures guide harness synthesis, while policy training is strictly conditioned on verified rollouts matched to the adopted runtime for supervised fine-tuning (SFT) or fresh online interactions for reinforcement learning (RL). On the Tmax promotion split, CoTrace advances Qwen3.5-9B from 78 to 88 solved tasks under supervised fine-tuning while an online reinforcement variant reaches 90. Specifically, a compact harness-matched corpus produces steady model gains at substantially lower compute than much larger corpora pooled across sibling harnesses. Furthermore, evaluations on Terminal-Bench 2.1 and SWE-bench Lite show that out-of-distribution transfer depends fundamentally on harness compatibility, where maintaining consistency between training and evaluation runtimes prevents procedural execution breakdowns observed under foreign scaffolds.

cs.CL↗
arXiv2026

Spurious Tool Use: When RL Agents Learn the Wrong Reason to Act

Large language model (LLM) agents increasingly interleave natural language reasoning with external tools such as web search and code execution. These tool-use policies are often optimized via reinforcement learning (RL), which can amplify spurious correlations in the training data. In this work, we study when and why RL-trained agents learn shortcut tool-selection policies: invoking tools based on superficial prompt cues rather than genuine task requirements. We construct controlled synthetic environments combining factual question answering and mathematical reasoning tasks, and inject cues that are strongly correlated with specific tools during training but causally irrelevant to tool necessity. Across counterfactual evaluations where cues are present but the associated tools are not required, agents exhibit substantial shortcut behavior, with spurious tool invocation rates increasing by up to 39 percent. However, shortcut formation is not universal: across the conditions we test, it arises only when the agent has already learned to use the target tool reliably, suggesting that task competence, rather than dataset imbalance alone, is a key factor in shortcut learning. A swapped-cue analysis further shows that semantic alignment between cues and tools substantially amplifies this effect. To mitigate these failures, we introduce a dense, decision-level reward in which an LLM judge evaluates the necessity of each tool call. This tool-necessity reward effectively suppresses cue-driven tool use while preserving task performance, providing a practical approach to improving the robustness of LLM agent tool-use policies.

cs.CL↗
arXiv2026

OpenResearcher: A Fully Open Pipeline for Long-Horizon Deep Research Trajectory Synthesis

Training deep research agents requires long-horizon trajectories that interleave search, evidence aggregation, and multi-step reasoning. However, existing data collection pipelines typically rely on proprietary web APIs, making large-scale trajectory synthesis costly, unstable, and difficult to reproduce. We present OpenResearcher, a reproducible pipeline that decouples one-time corpus bootstrapping from multi-turn trajectory synthesis and executes the search-and-browse loop entirely offline using three explicit browser primitives: search, open, and find, over a 15M-document corpus. Using GPT-OSS-120B as the teacher model, we synthesize over 97K trajectories, including a substantial long-horizon tail with 100+ tool calls. Supervised fine-tuning a 30B-A3B backbone on these trajectories achieves 54.8\% accuracy on BrowseComp-Plus, a +34.0 point improvement over the base model, while remaining competitive on BrowseComp, GAIA, and xbench-DeepSearch. Because the environment is offline and fully instrumented, it also enables controlled analysis, where our study reveals practical insights into deep research pipeline design, including data filtering strategies, agent configuration choices, and how retrieval success relates to final answer accuracy. We release the pipeline, synthesized trajectories, model checkpoints, and the offline search environment at https://github.com/TIGER-AI-Lab/OpenResearcher.

cs.IR↗
arXiv2026

Secure Coverage Enhancement in Aerial Reconfigurable Intelligent Surface-Assisted High-Speed Train Communication Systems

High-speed trains (HSTs) have become a prominent means of transportation, requiring high data rates and reliable communication services for HST passengers. However, the wireless channels in HST communication systems are susceptible to various security threats, including eavesdropping. Addressing these security concerns is therefore of critical importance. One promising technology for enhancing security is the integration of a reconfigurable intelligent surface (RIS) on an unmanned aerial vehicle, referred to as an aerial reconfigurable intelligent surface (ARIS). This technology offers significant potential for improving wireless network performance, though it also introduces unique challenges in terms of physical layer security (PLS). This paper investigates the PLS of ARIS-aided HST communication systems. A problem of maximizing the weighted sum secrecy rate is formulated by jointly optimizing the active beamforming at the base station (BS) and the phase shift at the ARIS, subject to constrains on the BS transmit power and the unit modulus of the ARIS reflecting coefficient. To address this problem, a joint optimization algorithm is proposed using the block coordinate descent method. Specifically, the problem is decomposed into two subproblems: active beamforming design and ARIS phase shift optimization. The active beamforming is optimally designed via the successive convex approximation technique, while the ARIS phase shift is efficiently updated using the alternating direction method of multipliers technique. Simulation results demonstrate the rapid convergence of the proposed algorithm, which achieves a higher secrecy rate compared to existing methods in the literature.

cs.IT↗
arXiv2026

Beyond the Mean: Multi-Moment Policy Optimization for LLM Reasoning

Reinforcement learning has become a central paradigm for improving the reasoning capabilities of large language models. Existing methods generally aim to reduce the failure probabilities induced across problems. In this paper, we introduce a moment-based perspective on policy optimization for LLM reasoning by treating the failure probability of a randomly sampled problem as a random variable and characterizing optimization objectives through its moments. Under this perspective, many existing methods optimize only a single moment of the failure-probability distribution, leaving its broader distributional structure largely uncharacterized. We propose \textbf{M}ulti-\textbf{M}oment \textbf{P}olicy \textbf{O}ptimization (MMPO), a novel policy optimization framework that jointly minimizes multiple moments of the failure-probability distribution. MMPO admits a direct operational interpretation as minimizing the expected truncated time required to obtain the first successful response. Beyond MMPO, we further develop a general moment-transformation framework that systematically induces different moment profiles and provides a unified view of a broader family of policy optimization objectives. Experiments across five mathematical reasoning benchmarks and models of different scales demonstrate that MMPO consistently outperforms strong baselines. We hope this moment-based perspective offers new insights into the design of policy optimization objectives for LLM reasoning.

cs.AI↗
arXiv2026

In-the-Flow Agentic System Optimization for Effective Planning and Tool Use

Outcome-driven reinforcement learning has advanced reasoning in large language models (LLMs), but prevailing tool-augmented approaches train a single, monolithic policy that interleaves thoughts and tool calls under full context; this scales poorly with long horizons and diverse tools and generalizes weakly to new scenarios. Agentic systems offer a promising alternative by decomposing work across specialized modules, yet most remain training-free or rely on offline training decoupled from the live dynamics of multi-turn interaction. We introduce AgentFlow, a trainable, in-the-flow agentic framework that coordinates four modules (planner, executor, verifier, generator) through an evolving memory and directly optimizes its planner inside the multi-turn loop. To train on-policy in live environments, we propose Flow-based Group Refined Policy Optimization (Flow-GRPO), which tackles long-horizon, sparse-reward credit assignment by converting multi-turn optimization into a sequence of tractable single-turn policy updates. It broadcasts a single, verifiable trajectory-level outcome to every turn to align local planner decisions with global success and stabilizes learning with group-normalized advantages. Across ten benchmarks, AgentFlow with a 7B-scale backbone outperforms top-performing baselines with average accuracy gains of 14.9% on search, 14.0% on agentic, 14.5% on mathematical, and 4.1% on scientific tasks, even surpassing larger proprietary models like GPT-4o. Further analyses confirm the benefits of in-the-flow optimization, showing improved planning, enhanced tool-calling reliability, and positive scaling with model size and reasoning turns.

cs.AI↗
arXiv2026

Dr-DCI: Scaling Direct Corpus Interaction via Dynamic Workspace Expansion

Agentic search over large corpora relies on retriever-mediated interfaces (e.g., BM25 or ColBERT) for scalable candidate discovery. While effective at ranking relevant documents, these interfaces expose evidence only as ranked results or bounded document views, limiting agents' ability to reorganize material and verify constraints across documents. Direct Corpus Interaction (DCI) addresses this limitation by exposing shell-executable corpus operations for flexible search, filtering, comparison, and verification. However, full-corpus terminal commands become slow and unstable as the corpus grows, degrading performance and efficiency. We introduce DR-DCI, a retriever-steered DCI framework that treats retrieval as an agent-callable action for expanding a local workspace. Rather than operating directly over the full corpus, the agent dynamically pulls relevant documents into an evolving workspace and conducts DCI operations within it. This design combines retriever-level recall with DCI-style precision: retrieval keeps exploration scalable, while DCI preserves the local operations needed for effective evidence resolution. Experiments show that DR-DCI is both effective and efficient across scales. On Browsecomp-Plus, DR-DCI reaches 71.2\% accuracy, improving over raw DCI and ablated variants by up to 8.3 points while reducing tool usage, wall time, and estimated cost. With workspace-preserving context reset, accuracy further improves to 73.3\%. In corpus-scaling experiments, DR-DCI remains effective from 100K to 10M documents, whereas raw DCI becomes unstable and BM25 performs substantially worse. DR-DCI also scales to a 20M-scale file-per-document Wiki-18 QA setting, achieving an average score of 63.0 across six benchmarks and outperforming retrieval-based and trained search-agent baselines. Ablation analysis further shows that ranked previews and inter-document DCI are key to performance.

cs.AI↗
arXiv2026

PRISMR: Overcoming Parse Collapse in Multimodal Listwise Ranking via Parameterized Representation Internalization

Generative listwise ranking with Large Multimodal Models (LMMs) aims to capture global list context in a single forward pass, but its effectiveness degrades in long-context multimodal scenarios. We identify a recurring failure mode, parse collapse, where the autoregressive decoder produces fluent yet incomplete rankings by silently omitting candidates and terminating early. This failure stems from limited context utilization rather than simple formatting mistakes, making prompt engineering and constrained decoding insufficient. We propose PRISMR (Parameterized Representation Internalization for Semantic Multimodal Ranking), a framework that replaces transient in-context list processing with parametric structural conditioning. PRISMR uses a lightweight hypernetwork to encode multimodal candidates in parallel and generate item-specific LoRA weights, which are synthesized into an instance-specific adapter for a LMM. This paradigm enables more robust internalization of list structure while preserving the base model. We further introduce a large-scale multimodal review-ranking benchmark for evaluation. Experiments demonstrate that PRISMR substantially reduces parse collapse, improves listwise ranking performance, and transfers effectively across domains and instruction-tuned backbones.

cs.AI↗
arXiv2026

Masking Stale Observations Helps Search Agents -- Until It Doesn't: A Regime Map and Its Mechanism

Long-horizon search agents accumulate large amounts of retrieved content across many tool calls, making context-budget efficiency increasingly important. A minimal intervention is to mask stale observations from the context as the trajectory progresses, but it remains unclear when this form of context management helps and why. We study observation masking through a systematic sweep over various agent backbones (4B to 284B parameters) and three retrievers on offline and live-web agentic search benchmarks. We find that the accuracy gain from masking follows an asymmetric inverted-U shape when plotted against the model's accuracy without context management: a plateau under weak retrievers, a peak when a strong retriever meets a mid-capacity model, and a sharp collapse when the model is saturated. This pattern reflects the interaction between retriever recall and the model's implicit filtering capacity, rather than either factor in isolation. Mechanistically, masking implements a token-for-turn trade-off: it removes observations the model has largely stopped attending to and pages the agent rarely re-opens. The added turns help when they convert failures into successes, but they fail when masking removes evidence the model would otherwise have used. We therefore reframe context management as a regime-dependent intervention and provide a holistic perspective for analyzing context use in agentic deep search. We release our scaffold and trajectories here (https://github.com/i-DeepSearch/observation-masking) to support future research.

cs.CL↗
arXiv2026

PointNeRT: A Physics Aware Neural Ray Tracing Surrogate for Propagation Channel Modeling

Ray tracing (RT) has emerged as a key tool for propagation channel modeling and network planning. Conventional RT is based on electromagnetic (EM) wave theory and its application relies on detailed mesh-based environment representations and material properties. In realistic environments, limited environmental geometry and material uncertainties hinder its scalability to complex scenarios. In this paper, we propose a novel physics aware neural RT surrogate named PointNeRT to address these limitations. The proposed model directly takes point clouds as environmental input, and efficiently reconstruct multipath without explicitly constructing mesh models or manually defining EM interaction rules. PointNeRT adopts a hop-by-hop modeling strategy guided by physical interaction constraints. It supports sequential prediction of multipath propagation and power attenuation. Numerical results and experiments demonstrate that the proposed method implicitly captures surface normal characteristics and EM material effects. It further achieves robust generalization in mobility scenarios and provides a physics-guided neural modeling of multipath propagation.

eess.SP↗
arXiv2026

Beyond Semantic Similarity: Rethinking Retrieval for Agentic Search via Direct Corpus Interaction

Modern retrieval systems, whether lexical or semantic, expose a corpus through a fixed similarity interface that compresses access into a single top-k retrieval step before reasoning. This abstraction is efficient, but for agentic search, it becomes a bottleneck: exact lexical constraints, sparse clue conjunctions, local context checks, and multi-step hypothesis refinement are difficult to implement by calling a conventional off-the-shelf retriever, and evidence filtered out early cannot be recovered by stronger downstream reasoning. Agentic tasks further exacerbate this limitation because they require agents to orchestrate multiple steps, including discovering intermediate entities, combining weak clues, and revising the plan after observing partial evidence. To tackle the limitation, we study direct corpus interaction (DCI), where an agent searches the raw corpus directly with general-purpose terminal tools (e.g., grep, file reads, shell commands, lightweight scripts), without any embedding model, vector index, or retrieval API. This approach requires no offline indexing and adapts naturally to evolving local corpora. Across IR benchmarks and end-to-end agentic search tasks, this simple setup substantially outperforms strong sparse, dense, and reranking baselines on several BRIGHT and BEIR datasets, and attains strong accuracy on BrowseComp-Plus and multi-hop QA without relying on any conventional semantic retriever. Our results indicate that as language agents become stronger, retrieval quality depends not only on reasoning ability but also on the resolution of the interface through which the model interacts with the corpus, with which DCI opens a broader interface-design space for agentic search.

cs.IR↗
arXiv2026

Robust Transmission Design for RIS-Assisted High-Speed Train Communication Coverage Enhancement With Imperfect Cascaded Channels

Reconfigurable intelligent surface (RIS) has recently been gained attention as an effective technique improving the coverage and performance of communication systems by creating additional communication links. Deployment of RIS is crucial for overcoming signal coverage limitations, especially in high-speed train (HST) scenarios. Considerable research has been performed assuming perfect channel state information (CSI). However, due to the rapidly time-varying fading channels and feedback delays, achieving perfect CSI at the base station (BS) is not feasible in the HST scenarios. To tackle this problem, this paper investigates a robust design strategy for RIS-aided HST communication coverage enhancement, particularly focusing on cascaded BS-RIS-user channels at BS (CBRUB). The study explores the optimization problem under two types distinct of models: centered on minimizing transmit power subject to worst-case rate constraints within the bounded CSI error (BCSIE) model, and the other focusing on outage probability (OP) constraints under the statistical CSI error (SCSIE) model. We use the S-procedure to approximate the non-convex (NC) constraints, converting the worst-case rate constraints into linear matrix inequalities. Additionally, the Bernstein-type inequality is applied to transform the OP constraints into second-order cone constraints and linear inequalities. The simulation analysis results show that CBRUB errors have a significant effect on system performance compared to direct CSI errors.

cs.IT↗
arXiv2026

ReviewGrounder: Improving Review Substantiveness with Rubric-Guided, Tool-Integrated Agents

The rapid rise in AI conference submissions has driven increasing exploration of large language models (LLMs) for peer review support. However, LLM-based reviewers often generate superficial, formulaic comments lacking substantive, evidence-grounded feedback. We attribute this to the underutilization of two key components of human reviewing: explicit rubrics and contextual grounding in existing work. To address this, we introduce REVIEWBENCH, a benchmark evaluating review text according to paper-specific rubrics derived from official guidelines, the paper's content, and human-written reviews. We further propose REVIEWGROUNDER, a rubric-guided, tool-integrated multi-agent framework that decomposes reviewing into drafting and grounding stages, enriching shallow drafts via targeted evidence consolidation. Experiments on REVIEWBENCH show that REVIEWGROUNDER, using a Phi-4-14B-based drafter and a GPT-OSS-120B-based grounding stage, consistently outperforms baselines with substantially stronger/larger backbones (e.g., GPT-4.1 and DeepSeek-R1-670B) in both alignment with human judgments and rubric-based review quality across 8 dimensions. The code is available \href{https://github.com/EigenTom/ReviewGrounder}{here}.

cs.CL↗