Searcharxiv⌕ Search

arXiv subjects

Xinyu Ma

Publications and source records attributed to Xinyu Ma.

At least 19 recordsLinked to original sources
arXiv2026

Token-Level Credit Assignment Optimization for Generative Document Retrieval

Generative retrieval models perform document retrieval by autoregressively generating document identifiers (DocIDs). This process naturally forms a sequential decision problem, i.e., the model makes a sequence of token-level decisions, selecting a DocID token at each decoding step, with the resulting complete sequence identifying the retrieved document. However, relevance feedback is available only after the complete DocID has been generated and mapped to a document, resulting in a granularity mismatch between token-level generation decisions and document-level retrieval supervision. Consequently, existing reinforcement learning methods for generative retrieval rely on sequence-level rewards, assigning the same document-level relevance signal to every decoding step. Such uniform credit assignment obscures the contribution of individual token decisions, making it difficult to identify which decisions contribute to retrieval success or failure. In this paper, we propose Token-Level Credit Assignment for Generative Retrieval (TCA), a fine-grained reinforcement learning framework that aligns the granularity of credit assignment with that of autoregressive DocID generation. Unlike assigning a single reward to an entire generated DocID, TCA derives fine-grained rewards by comparing the hidden-state trajectory of each generated DocID with the gold DocID trajectory obtained from a frozen reference model. These trajectory-based rewards provide differentiated feedback across decoding steps, allowing the policy to reinforce generation paths that remain aligned with the target DocID. Moreover, TCA decouples token-level credit assignment from policy optimization and can be instantiated with both GRPO and PPO. Experiments on benchmarks show that our method consistently outperforms baselines, demonstrating the effectiveness of fine-grained supervision for aligning DocID generation.

cs.IR↗
arXiv2026

RaMark: Radioactive Watermarking for Generated Tabular Data

Recent advances in generative modeling have made generated tabular data a practical solution for privacy-sensitive data sharing, where watermarking enables ownership verification. However, existing watermarking methods fundamentally fail under retraining attacks, in which an adversary retrains a generative model on a watermarked dataset and regenerates high-utility data that no longer carries the watermark. We address this challenge by introducing radioactivity, the property that a watermark remains detectable after generative model retraining, and propose RaMark, a radioactive watermarking method that embeds a sinusoidal dependency as an intrinsic component of the data distribution. By coupling the watermark with the underlying distribution, RaMark ensures that any generative model preserving data utility also has to preserve the watermark. We theoretically show that with high probability removing watermark degrades utility and alters data distribution. Extensive experiments on two real-world tabular datasets, under a large-scale ownership verification setting with $10^5$ independent data owners, demonstrate that RaMark achieves substantially stronger radioactivity than seven state-of-the-art methods and consistently outperforms them against both retraining and data modification attacks.

cs.CR↗
arXiv2026

Ion-Implanted Silicon Nanoregions Enable Ultra-Low-Loss Trimming of Cladded Photonic Integrated Circuits

Photonic integrated circuits (PICs) have emerged as a key platform for information processing, including optical communication and computing. As PIC complexity increases, fabrication-induced response deviations accumulate, making post-fabrication trimming critical for unlocking their full potential. Here, we demonstrate that silicon ion implantation enables scalable, ultra-low-loss trimming of cladded PICs by locally forming a high-index silicon-rich region. Structural characterization confirms that the silicon-implanted nanoregion is confined to the cladding without observable damage to the underlying waveguide. The implantation-induced excess loss is below 0.001 dB per {\pi} phase shift, while the optical response remains stable over a four-month observation period. Automated trimming is demonstrated on a photonic crossbar array, reducing the average channel output variation from 78.5% to 5.9%. These results establish a practical route towards automated, ultra-low-loss post-fabrication trimming for large-scale cladded PICs.

physics.optics↗
arXiv2026

Search for Truth from Reasoning: A Dynamic Representation Editing Framework for Steering LLM Trajectories

Current approaches to enhance Large Language Model (LLM) reasoning, such as Chain-of-Thought and "Wait" prompts, primarily encourage models to think more, yet often fail to guide them toward Truth. While Representation Editing (RepE) offers a intrinsic control, its application to dynamic reasoning trajectories remains underexplored. In this work, we bridge this gap by investigating the geometry of truth within unfolding reasoning chains. We uncover three critical insights: (1) Truth is encoded at the sentence level and is entangled with latent reasoning patterns; (2) Effective intervention follows an Uncertainty Principle and a Decay Effect, requiring localization to early, high-entropy forks; (3) Naive steering vectors suffer from noise, risking collateral damage to correct trajectories. Based on these findings, we propose DynaSteer, a dynamic RepE framework. DynaSteer employs pattern clustering to disentangle reasoning manifolds and utilizes Fisher-LDA to project purified truth. By dynamically monitoring lookahead entropy, it selectively steers and rolls back trajectories only when necessary. Comprehensive experimental results on several MATH benchmark verify the effectiveness of DynaSteer, and experiments on out-of-domain coding tasks further confirm its generalization ability. Our code is publicly available at https://github.com/tianlwang/DynaSteer.

cs.AI↗
arXiv2026

Empowering Polymeric Materials Discovery by Artificial Intelligence

Polymeric materials underpin modern technologies spanning energy storage, microelectronics, healthcare and sustainable manufacturing. Yet their rational design remains exceptionally challenging because material performance emerges from complex interactions among molecular composition, chain architecture, processing history and hierarchical structural evolution across multiple length and time scales. Consequently, polymer research has long relied on labor-intensive experimentation and fragmented modeling approaches, limiting both mechanistic understanding and innovation efficiency. Recent advances in data infrastructure, machine learning, large artificial intelligence (AI) models and laboratory automation are beginning to reshape this landscape. Rather than functioning as isolated tools, polymer databases, predictive models, AI agents and automated laboratories are increasingly converging into interconnected discovery ecosystems. As a result, the central challenge is shifting from improving predictive accuracy alone to enabling reliable decision-making, adaptive learning and seamless integration across computation, experimentation and scientific reasoning. We argue that polymer science is entering an era of autonomous discovery, in which data, simulation, reasoning and experimentation operate within self-improving feedback loops that continuously generate hypotheses, design materials, execute experiments and refine predictive models. By unifying molecular design, process optimization, experimental validation and industrial translation, such autonomous ecosystems establish a more predictive, reproducible and scalable paradigm for polymer innovation, fundamentally transforming how polymer research is conducted.

physics.chem-ph↗
arXiv2026

Beyond Domains: Reusing Web Skills via Transferable Interaction Patterns

Large language model (LLM) web agents are usually deployed as tool callers: each turn, the model reads a fresh page observation and emits one structured tool action. When every action is a low-level primitive, horizons grow quickly and so do policy-facing LLM completions, dominating latency and cost on benchmarks such as Mind2Web and WebArena. Recent systems therefore wrap repeated interaction fragments as web skills: callable tools built from successful trajectories or induced programs, so one call can replace several primitives. However, prior skill libraries are still triggered mainly by instruction similarity or coarse site metadata, which yields low skill reuse on held-out sites and leaves much of the potential step and token reduction on the table. We present SkillMigrator, an agent that learns reusable web skills and transfers them across sites by matching layout structure rather than specific element references. Each induced skill is stored as a transferable interaction pattern (TIP): the skill paired with a structural sketch of the snapshot at induction time. At test time, SkillMigrator retrieves TIPs by layout similarity and grounds their references on the live page. The rest of the stack is standard: accessibility-snapshot observations with stable references, and fixed tool calling over primitives plus skill invocations. Compared with the state-of-the-art approaches, SkillMigrator reduces the average LLM-action count on successful trajectories by 8-10% across both WebArena and Mind2Web at matched success rate.

cs.AI↗
arXiv2026

Octave bandwidth 3D-Printed Couplers for Low-Loss Thin-Film Lithium Tantalate Circuits

Low-loss, broadband photonic integrated circuits (PICs) are critical enablers for optical communications, photonic computing, and quantum applications. Lithium tantalate on insulator (LTOI) is an emerging photonic platform offering a wide transparency window and strong Pockels effect, and thereby enabling efficient electro-optic modulation and high data rates. Here, we present the first implementation of efficient out-of-plane polymer coupling interfaces fabricated via 3D direct laser writing for both fully etched strip and partially etched rib LTOI waveguides, achieving ultra-low coupling losses of 0.9 dB (strip) and 1.25 dB (rib) per interface. Both coupler types exhibit a 3 dB optical bandwidth spanning more than an octave from 850 nm to 1740 nm and maintain stable operation under 1 W optical input power. Combined with on-chip waveguides exhibiting propagation losses below 0.1 dB/cm, these characteristics represent a key step toward unlocking the full potential of LTOI for high-speed optical signal processing with an unprecedented degree of parallelism. In addition, the octave-spanning bandwidth enables efficient interfacing of both the fundamental and second-harmonic signals, making the platform highly attractive for second harmonic generation based quantum squeezing applications.

physics.optics↗
arXiv2026

ROSD: Reflective On-Policy Self-Distillation for Language Model Reasoning across Domains

On-policy self-distillation (OPSD) improves the reasoning performance of large language models (LLMs) by providing dense token-level supervision for on-policy rollouts. However, existing OPSD methods often yield limited gains on in-domain reasoning and generalize poorly to out-of-domain problems. We identify two key causes: conditioning the self-teacher on a verified solution encourages imitation of training-domain reference trajectories rather than error-specific correction, and applying distillation to the full response can overwrite valid reasoning prefixes and reinforce overfitting. We propose Reflective On-policy Self-Distillation (ROSD), a framework that turns reference-solution imitation into targeted reasoning correction through reflection-guided, error-localized distillation. For each rollout, ROSD uses a self-reflector to extract a corrective idea and locate the first erroneous span. The corrective idea guides the self-teacher toward targeted supervision, while the localized error span restricts distillation to where correction is needed. This design corrects flawed reasoning while preserving valid prefixes. Experiments on multiple in-domain and out-of-domain reasoning benchmarks show that ROSD yields stronger in-domain reasoning performance overall and substantially better out-of-domain generalization than standard OPSD. Code is available at https://github.com/ZiqiZhao1/ROSD.

cs.CL↗
arXiv2026

Indefinite Stochastic LQ Optimal Control for Jump-Diffusion Systems with Random Coefficients

This paper studies indefinite stochastic linear-quadratic (LQ) optimal control for jump-diffusion systems with random coefficients. We construct an algebraic inverse flow from the zero-control base system, extract the semimartingale kernel of the value function, and prove that it satisfies a generalized stochastic Riccati equation with jumps (SREJ). Under a uniform convexity condition, we establish the existence and uniqueness of open-loop optimal controls for any initial pair and show that the associated matrix $\mathscr{N}(t)$ is uniformly positive definite, yielding an exact closed-loop feedback representation of the optimal control via the SREJ. A distinguishing feature of our approach is that it requires neither relaxation techniques (as in the compensator method) nor additional invertibility assumptions on the optimal state process, and it accommodates the general case where the control enters the jump part ($F \neq 0$). As an application, we analyze a financial portfolio problem with a jump-diffusion risky asset whose excess return is zero, where the investor minimizes a cost functional with a negative terminal wealth weight. The uniform convexity condition reduces to an explicit inequality among the risk aversion coefficient, volatility, jump magnitude, and risk-free rate, thereby delineating the parametric region in which an optimal strategy exists. These results extend classical indefinite LQ theory to jump-diffusion systems with random coefficients.

math.OC↗
arXiv2026

OGER: A Robust Offline-Guided Exploration Reward for Hybrid Reinforcement Learning

Recent advancements in Reinforcement Learning with Verifiable Rewards (RLVR) have significantly improved Large Language Model (LLM) reasoning, yet models often struggle to explore novel trajectories beyond their initial policy distribution. While offline teacher guidance and entropy-driven strategies have been proposed to address this, they often lack deep integration or are constrained by the model's inherent capacity. In this paper, we propose OGER (Offline-Guided Exploration Reward), a novel framework that unifies offline teacher guidance and online reinforcement learning through a specialized reward modeling lens. OGER employs multi-teacher collaborative training and constructs an auxiliary exploration reward that leverages both offline trajectories and the model's own entropy to incentivize autonomous exploration. Extensive experiments across mathematical and general reasoning benchmarks demonstrate that OGER consistently outperforms competitive baselines, achieving substantial gains in mathematical reasoning while maintaining robust generalization to out-of-domain tasks. We provide a comprehensive analysis of training dynamics and conduct detailed ablation studies to validate the effectiveness of our entropy-aware reward modulation. Our code is available at https://github.com/ecoli-hit/OGER.git.

cs.AI↗
arXiv2026

Voltage-Programmable Photon Statistics Using a High-Extinction Thin-film Lithium Niobate Modulator

Controlling the statistical properties of light, namely the fluctuations in photon arrival, entropy and number, is essential for both classical and quantum photonics. While integrated systems provide tunable control over amplitude, phase, and wavelength, real-time modulation of photon statistics has remained a long-standing challenge. Herein, we introduce the concept and experimental realization of a photon statistics transducer: a high-extinction, broadband electro-optic device capable of deterministically shaping photon-number distributions at nanosecond timescales. Our approach employs a cascaded thin-film lithium niobate (TFLN) Mach-Zehnder amplitude modulator delivering more than 50 dB extinction, enabling precise suppression and release of coherent seed light from an integrated InP laser. By exploiting the interplay between seed suppression and erbium-doped fiber amplifier dynamics, we demonstrate smooth, voltage-controlled switching between Poissonian and super-Poissonian photon statistics, with second-order coherence g2(0) tunable from 1.0 to 1.7. Complementary measurements with superconducting nanowire single-photon detectors further show photon-flux control down to sub-photon levels, highlighting the potential for future operation with non-classical sources. The photon statistics transducer thus establishes statistical modulation as a new functional primitive in integrated photonics. Applications range from entropy generation and secure communication to neuromorphic and hybrid quantum-classical processing, where controlled randomness and entropy are essential resources. By enabling programmable transitions between statistical regimes using only electronic drive signals, our work lays the foundation for adaptive, entropy-aware photonic systems that bridge classical and quantum domains.

physics.optics↗
arXiv2026

LLM-based Listwise Reranking under the Effect of Positional Bias

LLM-based listwise passage reranking has attracted attention for its effectiveness in ranking candidate passages. However, these models suffer from positional bias, where passages positioned towards the end of the input are less likely to be moved to top positions in the ranking. We hypothesize that there are two primary sources of positional bias: (1) architectural bias inherent in LLMs and (2) the imbalanced positioning of relevant documents. To address this, we propose DebiasFirst, a method that integrates positional calibration and position-aware data augmentation during fine-tuning. Positional calibration uses inverse propensity scoring to adjust for positional bias by re-weighting the contributions of different positions in the loss function when training. Position-aware augmentation augments training data to ensure that each passage appears equally across varied positions in the input list. This approach markedly enhances both effectiveness and robustness to the original ranking across diverse first-stage retrievers, reducing the dependence of NDCG@10 performance on the position of relevant documents. DebiasFirst also complements the inference-stage debiasing methods, offering a practical solution for mitigating positional bias in reranking.

cs.IR↗
arXiv2026

Cold-Starts in Generative Recommendation: A Reproducibility Study

Cold-start recommendation remains a central challenge in dynamic, open-world platforms, requiring models to recommend for newly registered users (user cold-start) and to recommend newly introduced items to existing users (item cold-start) under sparse or missing interaction signals. Recent generative recommenders built on pre-trained language models (PLMs) are often expected to mitigate cold-start by using item semantic information (e.g., titles and descriptions) and test-time conditioning on limited user context. However, cold-start is rarely treated as a primary evaluation setting in existing studies, and reported gains are difficult to interpret because key design choices, such as model scale, identifier design, and training strategy, are frequently changed together. In this work, we present a systematic reproducibility study of generative recommendation under a unified suite of cold-start protocols.

cs.IR↗
arXiv2026

Model Editing for New Document Integration in Generative Information Retrieval

Generative retrieval (GR) reformulates the Information Retrieval (IR) task as the generation of document identifiers (docIDs). Despite its promise, existing GR models exhibit poor generalization to newly added documents, often failing to generate the correct docIDs. While incremental training offers a straightforward remedy, it is computationally expensive, resource-intensive, and prone to catastrophic forgetting, thereby limiting the scalability and practicality of GR. In this paper, we identify the core bottleneck as the decoder's ability to map hidden states to the correct docIDs of newly added documents. Model editing, which enables targeted parameter modifications for docID mapping, represents a promising solution. However, applying model editing to current GR models is not trivial, which is severely hindered by indistinguishable edit vectors across queries, due to the high overlap of shared docIDs in retrieval results. To address this, we propose DOME (docID-oriented model editing), a novel method that effectively and efficiently adapts GR models to unseen documents. DOME comprises three stages: (1) identification of critical layers, (2) optimization of edit vectors, and (3) construction and application of updates. At its core, DOME employs a hybrid-label adaptive training strategy that learns discriminative edit vectors by combining soft labels, which preserve query-specific semantics for distinguishable updates, with hard labels that enforce precise mapping modifications. Experiments on widely used benchmarks, including NQ and MS MARCO, show that our method significantly improves retrieval performance on new documents while maintaining effectiveness on the original collection. Moreover, DOME achieves this with only about 60% of the training time required by incremental training, considerably reducing computational cost and enabling efficient, frequent model updates.

cs.IR↗
arXiv2026

DeepResearch-9K: A Challenging Benchmark Dataset of Deep-Research Agent

Deep-research agents are capable of executing multi-step web exploration, targeted retrieval, and sophisticated question answering. Despite their powerful capabilities, deep-research agents face two critical bottlenecks: (1) the lack of large-scale, challenging datasets with real-world difficulty, and (2) the absence of accessible, open-source frameworks for data synthesis and agent training. To bridge these gaps, we first construct DeepResearch-9K, a large-scale challenging dataset specifically designed for deep-research scenarios built from open-source multi-hop question-answering (QA) datasets via a low-cost autonomous pipeline. Notably, it consists of (1) 9000 questions spanning three difficulty levels from L1 to L3 (2) high-quality search trajectories with reasoning chains from Tongyi-DeepResearch-30B-A3B, a state-of-the-art deep-research agent, and (3) verifiable answers. Furthermore, we develop an open-source training framework DeepResearch-R1 that supports (1) multi-turn web interactions, (2) different reinforcement learning (RL) approaches, and (3) different reward models such as rule-based outcome reward and LLM-as-judge feedback. Finally, empirical results demonstrate that agents trained on DeepResearch-9K under our DeepResearch-R1 achieve state-of-the-art results on challenging deep-research benchmarks. We release the DeepResearch-9K dataset on https://huggingface.co/datasets/artillerywu/DeepResearch-9K and the code of DeepResearch-R1 on https://github.com/Applied-Machine-Learning-Lab/DeepResearch-R1.

cs.AI↗
arXiv2026

SPARKLING: Balancing Signal Preservation and Symmetry Breaking for Width-Progressive Learning

Progressive Learning (PL) reduces pre-training computational overhead by gradually increasing model scale. While prior work has extensively explored depth expansion, width expansion remains significantly understudied, with the few existing methods limited to the early stages of training. However, expanding width during the mid-stage is essential for maximizing computational savings, yet it remains a formidable challenge due to severe training instabilities. Empirically, we show that naive initialization at this stage disrupts activation statistics, triggering loss spikes, while copy-based initialization introduces gradient symmetry that hinders feature diversity. To address these issues, we propose SPARKLING (balancing {S}ignal {P}reservation {A}nd symmet{R}y brea{K}ing for width-progressive {L}earn{ING}), a novel framework for mid-stage width expansion. Our method achieves signal preservation via RMS-scale consistency, stabilizing activation statistics during expansion. Symmetry breaking is ensured through asymmetric optimizer state reset and asymmetric learning rate re-warmup. Extensive experiments on dense and Mixture-of-Experts (MoE) models demonstrate that, across multiple width axes and optimizer families, SPARKLING consistently outperforms training from scratch and reduces training cost by up to 35% under $2\times$ width expansion.

cs.LG↗
arXiv2026

Mobility-as-a-service (MaaS) system as a multi-leader-multi-follower game: A single-level variational inequality (VI) formulation

This study models a Mobility-as-a-Service (MaaS) system as a multi-leader-multi-follower game that captures the complex interactions among the MaaS platform, service operators, and travelers. We consider a coopetitive setting where the MaaS platform purchases service capacity from service operators and sells multi-modal trips to travelers following an origin-destination-based pricing scheme; meanwhile, service operators use their remaining capacities to serve single-modal trips. As followers, travelers make both mode choices, including whether to use MaaS, and route choices in the multi-modal transportation network, subject to prices and congestion. Inspired by the dual formulation for traffic assignment problems, we propose a novel single-level variational inequality (VI) formulation by introducing a virtual traffic operator, along with the MaaS platform and multiple service operators. A key advantage of the proposed VI formulation is that it supports parallel solution procedures and thus enables large-scale applications. We prove that an equilibrium solution always exists given the negotiated wholesale price of service capacity. Numerical experiments on a small network further demonstrate that the wholesale price can be tailored to align with varying system-wide objectives. The proposed MaaS system demonstrates potential for creating a "win-win-win" outcome -- service operators and travelers are better off compared to the "without MaaS" scenario, meanwhile the MaaS platform remains profitable. Such a Pareto-improving regime can be explicitly specified with the wholesale capacity price. Similar conclusions are drawn from the experiment of an extended multi-modal Sioux Falls network, which also validates the scalability of the proposed model and solution algorithm.

econ.GN↗
arXiv2026

Agentic-R: Learning to Retrieve for Agentic Search

Agentic search has recently emerged as a powerful paradigm, where an agent interleaves multi-step reasoning with on-demand retrieval to solve complex questions. Despite its success, how to design a retriever for agentic search remains largely underexplored. Existing search agents typically rely on similarity-based retrievers, while similar passages are not always useful for final answer generation. In this paper, we propose a novel retriever training framework tailored for agentic search. Unlike retrievers designed for single-turn retrieval-augmented generation (RAG) that only rely on local passage utility, we propose to use both local query-passage relevance and global answer correctness to measure passage utility in a multi-turn agentic search. We further introduce an iterative training strategy, where the search agent and the retriever are optimized bidirectionally and iteratively. Different from RAG retrievers that are only trained once with fixed questions, our retriever is continuously improved using evolving and higher-quality queries from the agent. Extensive experiments on seven single-hop and multi-hop QA benchmarks demonstrate that our retriever, termed \ours{}, consistently outperforms strong baselines across different search agents. Our codes are available at: https://github.com/8421BCD/Agentic-R.

cs.IR↗