Searcharxiv⌕ Search

arXiv subjects

Xin Zhang

Publications and source records attributed to Xin Zhang.

At least 55 records · Page 3Linked to original sources
arXiv2026

BENCHCOMPASS: From Scores to Signals for Training and Harness Decisions in Payment-Domain LLMs

Payment operations are a critical financial infrastructure, but the value of large language models in this domain remains unclear because payment rules change quickly, evidence is fragmented, and decisions depend on transaction state, participant role, region, and payment rail. Existing benchmarks do not isolate whether failures come from missing payment-rule knowledge, poor use of supplied evidence, or brittleness under imperfect harness inputs. We introduce BENCHCOMPASS, a payment-domain benchmark whose construction pipeline builds scenario-grounded tasks from typed evidence packs, applies LLM-based quality checks, creates task-input attack variants, and reserves final item admission for domain experts. The release contains an expert-reviewed Pro benchmark covering payment knowledge, context-grounded scenario reasoning, and Attacked Open robustness, plus a lower-assurance Normal pool for inspection and future curation. Across 16 model variants, BENCHCOMPASS shows qualitatively different failure modes: missing parametric payment knowledge, incomplete reasoning over supplied rules, and failure to reject plausible but invalid workflows. The benchmark remains unsaturated: the best frontier model reaches 89.6% on Open Context-Grounded Reasoning and 81.7% under attacked inputs, while a representative 32B open-weight model reaches 69.8% and 42.6%. Benchmark data and code are available at https://github.com/ant-intl/BenchCompass.

cs.AI↗
arXiv2026

Bioinfoysis Technical Report

Large language model agents have shown promise in bioinformatics, but most existing systems focus primarily on producing final answers, treating planning, tool use, and code execution as transient interactions. This design is poorly suited to long-horizon bioinformatics tasks, where conclusions must remain connected to the data, computations, and intermediate evidence that support them. We introduce \textbf{Bioinfoysis}, a multi-agent harness that represents each request as a persistent, artifact-grounded analysis run. Bioinfoysis combines global planning with step-wise, evidence-driven replanning: the planner maintains an executable checklist and revises pending steps using structured handoffs returned after each worker execution. These handoffs bind intermediate results to their responsible agent, checklist step, and plan generation, preventing stale evidence from being silently reused after replanning. A controlled runtime validates generated scripts, tables, and figures before they are used in downstream analysis or reporting, while role-specific context, persistent memory, and governed bioinformatics skills support reliable execution over long analysis trajectories. We evaluate Bioinfoysis on BixBench and two question-answering tracks of LAB-Bench 2. On BixBench, Bioinfoysis achieves state-of-the-art accuracy of 82.4\%. Across four underlying language models, Bioinfoysis increases average accuracy from 27.81\% to 64.13\% on SeqQA2 and from 3.13\% to 31.25\% on DbQA2. These results demonstrate that reliable bioinformatics automation depends not only on model capability, but also on the harness that governs planning, execution, memory, and evidence flow. We hope that the emergence of Bioinfoysis will play a driving and leading role in the development of the bioinformatics community. Our demo website can be seen in https://report.bioinfoysis.com/.

cs.AI↗
arXiv2026

MLLMRec: A Preference Reasoning Paradigm with Graph Refinement for Multimodal Recommendation

Multimodal recommendation combines the user historical behaviors with the modal features of items to capture the tangible user preferences, presenting superior performance compared to the conventional ID-based recommender systems. However, existing methods still encounter two key problems in the representation learning of users and items, respectively: (1) the initialization of multimodal user representations is either agnostic to historical behaviors or contaminated by irrelevant modal noise, and (2) the widely used KNN-based item-item graph contains noisy edges with low similarities and lacks audience co-occurrence relationships. To address such issues, we propose MLLMRec, a novel preference reasoning paradigm with graph refinement for multimodal recommendation. Specifically, on the one hand, the item images are first converted into high-quality semantic descriptions using a multimodal large language model (MLLM), thereby bridging the semantic gap between visual and textual modalities. Then, we construct a behavioral description list for each user and feed it into the MLLM to reason about the purified user preference profiles that contain the latent interaction intents. The reasoned profiles and the multimodal descriptions of items, together with their ID embeddings, are propagated over the user-item interaction graph to absorb the high-order collaborative signals. On the other hand, we develop the threshold-controlled denoising and topology-aware enhancement strategies to refine the suboptimal item-item graph, which are applied to both the multimodal and ID item representations to improve the accuracy of item representation learning. Extensive experiments on three publicly available datasets demonstrate that MLLMRec achieves the state-of-the-art performance. The source code is provided at https://github.com/Yuzhuo-Dang/MLLMRec.git.

cs.IR↗
arXiv2026

Uncertainty DMD: Restoring Diversity in Few-Step Autoregressive Video Distillation

Few-step distillation improves the efficiency of autoregressive (AR) video generation, but often causes diversity collapse: under the same prompt, different noise samples tend to produce highly similar videos with weakened motion dynamics. We analyze this degradation in Distribution Matching Distillation (DMD)-distilled AR video generators and find that, in the autoregressive setting, it takes the form of a structured uncertainty collapse: the mode-seeking bias of DMD maps different noise samples to nearly identical first chunks, and the deterministic AR cache then propagates this collapsed state to all subsequent chunks, turning a local loss of stochasticity at the rollout root into a global suppression of temporal variation. Based on this analysis, we propose Uncertainty DMD, a simple uncertainty-injection framework that restores stochasticity at two key stages of AR generation: a timestep perturbation for the first chunk to increase first-chunk diversity, and a stochastic cache-writing mechanism for later chunks to preserve uncertainty in autoregressive conditioning. The method requires no architectural changes and introduces only lightweight perturbation operations. The same perturbation mechanisms are used during both training and inference. Experiments show that Uncertainty DMD consistently improves diversity and motion dynamics while maintaining comparable per-sample visual quality.

cs.CV↗
arXiv2026

Multi-band cross-correlation dark sirens: enhancing cosmological parameter and gravitational-wave bias constraints

Multi-band gravitational-wave (GW) observation, combining space-borne and ground-based detectors across different frequency bands, can improve the sky localization of compact binary sources by two to three orders of magnitude compared to single-band detection. This enhancement is crucial for cross-correlation dark siren analyses, since the sky localization uncertainty directly determines the noise level of the GW angular power spectrum. In this work, we present the first Fisher forecast for cross-correlation dark siren cosmology with multi-band GW observations, cross-correlating GW events from the Einstein Telescope (ET), Cosmic Explorer (CE), and B-DECIGO with the Chinese Space-station Survey Telescope photometric galaxy survey. We compare three network configurations: the multi-band B-DECIGO+ET+2CE (BDET2CE), the ground-only ET+2CE (ET2CE), and B-DECIGO alone. In the $Λ$CDM model, BDET2CE achieves $σ(h)/h = 0.35\%$, improving by $37\%$ over the ground-only ET2CE ($0.55\%$) and by $86\%$ over B-DECIGO alone ($2.45\%$). Extending to the $w_0w_a$CDM framework, the multi-band advantage on cosmological parameters becomes more moderate, with BDET2CE improving $σ(h)/h$ by $\sim 4\%$ over ET2CE and $\sim 22\%$ over B-DECIGO. The most striking advantage of multi-band observation lies in the per-bin measurement of the GW clustering bias $b_{\rm GW}(z)$: at $z \sim 1-2$, BDET2CE constrains the bias to $\sim 3\%$ precision, compared to $\sim 8-60\%$ for ET2CE and $\sim 20-33\%$ for B-DECIGO. These precise, redshift-resolved bias measurements open a new avenue for probing the astrophysics of compact binary mergers, enabling constraints on formation channels such as isolated binary evolution and dynamical assembly that predict distinct clustering signatures.

astro-ph.CO↗
arXiv2026

TourPhysics: Bringing Physics to World Models for Exploration and Manipulation from a Single Image

Interactive visual world models must distinguish observation from physical intervention. Camera motion reveals new surfaces, whereas intervention changes object motion, contact, and deformation. Current video world models are largely driven by appearance priors and often lose physical or spatial consistency over long horizons. We present TourPhysics, an online framework initialized from a single image and a declarative physical configuration. TourPhysics extends PhysOmni, our ACM Multimedia 2026 work, from finite physics-grounded video synthesis to persistent exploration and manipulation. TourPhysics combines deterministic simulation with video generation while assigning separate roles to simulator state, geometric evidence, generator controls, and appearance memory. For each action, the simulator computes a finite physical and camera trajectory before the corresponding observation is generated. Accepted observations publish the terminal state and update the appearance memory and subsequent generator controls, while the committed state and simulator geometry remain fixed throughout synthesis and retry. We further separate the simulator geometry used for projection and visibility from the relative depth used to condition the generator. A reference-anchored memory retrieves accepted static appearance through geometric cross-view correspondence and incorporates it through a bounded residual that reverts to the native path when no valid correspondence exists. On simulator-defined camera tours and object manipulations, TourPhysics follows prescribed camera and object trajectories more closely than the evaluated baselines, preserves the input scene, and reduces appearance drift during long-horizon revisits.

cs.CV↗
arXiv2026

Separating Stream Stability from Long-Term Recall in Language Models

Methods for streaming language models are often discussed alongside long-context and memory systems, although they solve different problems. An attention sink can stabilize autoregressive generation over an indefinitely long stream while the model remains unable to use content that has left its recent-token cache. We argue that this distinction should be explicit in system claims and evaluation. We introduce three horizons: the stability horizon, over which predictive behavior remains well behaved; the access horizon, over which past content can still causally affect the output; and the utility horizon, over which a task retains acceptable performance. We show constructively that the stability horizon can be infinite while the access and utility horizons are finite. We then propose ThreeH, an evaluation contract that measures all three horizons under a common state and compute budget. Applying the framework to attention-sink streaming clarifies its strength, constant-memory, stable generation, without treating anchor tokens as semantic memory. The framework exposes roles for cache policies, recurrent state, retrieval, and external memory. Experiments on 128K-token streams, delayed binding recall, and delayed decisions show that attention sinks preserve local modeling but not content beyond the active cache; recurrent and retrieval state extend the semantic horizon.

cs.CL↗
arXiv2026

Histopathological Spectrum-Guided Prostate Stratification via Segmentation-Assisted Diagnostic Transformer

Prostate cancer diagnosis with multiparametric MRI (mpMRI) is commonly based on PI-RADS assessment or binary classification, which suffer from subjectivity and fail to capture clinically relevant pathological heterogeneity. To address this limitation, we construct a Prostate Cancer Histopathology Spectrum Dataset (PCa-HSD) and formulate a clinically meaningful four-class classification task, addressing the underrepresentation of benign lesions that are easily confounded with prostate cancer in existing datasets. We propose Language-guided Segmentation-assisted Diagnostic Transformer model (LSDT), which leverages zero-shot segmentation to provide anatomical priors and performs effective multi-modal slice fusion for classification. Our proposed method consistently improves accuracy across backbones, achieving the best average accuracy of 0.633 and JointRecall of 0.768 in five-fold cross-validation on a cohort of 344 patients. These results demonstrate that integrating pathology supervision and anatomical priors significantly enhances fine-grained prostate MRI classification and provides a more clinically relevant paradigm for risk stratification. Code will be made publicly available in a future revision.

cs.CV↗
arXiv2026

Grounded Skill Synthesis from Code at Scale for Agentic Intelligence

Reusable skills give agents transferable procedural knowledge, making scalable acquisition essential for extending agents beyond prior experience. Existing methods face two limitations: trajectory-based synthesis requires interactions with specific environments, while document-derived skills may lack executable evidence and verification. Source code offers a complementary path: it requires no prior agent experience yet provides executable evidence for grounding abstractions. We present Code2Skill, a fully automated pipeline that transforms selected code units into implementation-anchored records of atomic operations, composite workflows, and recurring patterns, then verifies each record through source-body-blind reconstruction and source-aware comparison. Applied to 19,769 popular, actively maintained GitHub repositories, Code2Skill produces CodeSkillBank, a grounded bank of 1,006,822 accepted records with workflow, boundary, provenance, and source-evidence metadata. Across 72 protocol-matched evaluations covering nine model settings and eight benchmarks, models augmented with retrieved CodeSkillBank skills improve by 11.7% on average over matched baselines and outperform them in 57 cases. Under a unified downstream interface, Code2Skill also outperforms trajectory-derived skill banks on all seven shared benchmarks, showing that repository-derived skills can provide useful procedural knowledge before agents accumulate sufficient interaction experience. Skills synthesized from tested AI-generated code achieve a 93.50% pass rate, compared with 93.00% for human-written code, providing initial evidence that the pipeline can expand with the growing volume of AI-generated software. Overall, Code2Skill transforms procedural knowledge embedded in repositories into grounded, verifiable, and transferable agent skills.

cs.SE↗
arXiv2026

KaLM-Embedding-V2: Superior Training Techniques and Data Inspire A Versatile Embedding Model

Recent advancements in Large Language Models (LLMs)-based text embedding models primarily focus on data scaling or synthesis, yet limited exploration of training techniques and data quality, thereby constraining performance. In this work, we propose KaLM-Embedding-V2 from the Lychee-KaLM team, a series of versatile and compact embedding models, systematically incentivizing advanced embedding capability in LLMs by superior training techniques and high-quality data. For model architecture, we implement the models in a 0.5B compact size with simple mean-pooling to produce fixed-length embeddings and remove the causal attention mask to enable fully bidirectional representation learning. For training techniques, we propose a progressive multi-stage training pipeline: pre-training on weakly supervised large-scale datasets, fine-tuning with supervised high-quality datasets, and contrastive distillation with fine-grained soft signals, integrated with focal-style reweighting and online hard-negative mixing to emphasize difficult samples and enrich hard negatives, respectively. For training data, we curate over 20 categories for pre-training and 100 categories for fine-tuning and contrastive distillation to improve both performance and generalization, leveraging task-specific instructions, hard-negative mining, and example-based multi-class labeling to ensure high quality. Combining these techniques, our KaLM-Embedding-V2 series achieves state-of-the-art performance on the Massive Text Embedding Benchmark, outperforming models of comparable size and rivaling models 3--26x larger, setting a new standard for versatile and compact embedding models under 1B parameters. The code, data, and models are available at https://kalm-embedding.github.io/.

cs.CL↗
arXiv2026

Continuous-time q-learning for Markov regime switching system under Tsallis entropy

This paper studies continuous-time q-learning (the continuous-time counterpart of Q-learning) for a Markov regime-switching system under Tsallis entropy regularization. The Tsallis entropy regularization yields an optimal policy distribution that may not necessarily be a Gibbs measure, thereby complicating algorithm design. Furthermore, to address the limited universality of current continuous-time regime-switching reinforcement learning algorithms (often restricted to the exploratory mean-variance framework), this study focuses on continuous-time q-learning for Markov regime-switching systems based on Tsallis entropy, aiming for a more universally applicable continuous-time reinforcement learning method. We establish the martingale characterization of the q-function under Tsallis entropy for continuous-time Markov regime-switching systems. We further design two q-learning algorithms that differ based on whether the Lagrange multiplier can be explicitly derived. We apply these algorithms to the continuous-time exploratory mean-variance portfolio optimization problem in a regime-switching market. Numerical experiments demonstrate the satisfactory performance of our q-learning algorithms.

math.OC↗
arXiv2026

Actor as Its Own Critic: Unifying Region Understanding and Localization via CycleGRPO

This paper introduces Actor as Its Own Critic, a unified reinforcement learning framework, Cycle Group Relative Policy Optimization (CycleGRPO), that jointly optimizes region understanding and localization for Multimodal Large Language Models (MLLMs). Unlike existing separate pipelines, we leverage the inherent duality between the two tasks to construct a self-evaluating reinforcement learning paradigm: "region $\to$ text $\to$ region''. Specifically, a single MLLM first acts as the actor to generate region captions, then immediately transitions to a critic to ground its generated text back in the spatial domain. Therefore, CycleGRPO requires only region inputs, e.g., masks or bounding boxes, entirely bypassing the need for textual ground truths. A quality-aware token-level cycle-consistency reward is employed to assess the semantic discriminability of text captions via their physical localization accuracy. Empirically, built upon SAMTok, our CycleGRPO framework successfully bootstraps both capabilities simultaneously. Without any task-specific fine-tuning, the framework yields consistent performance gains across a wide range of benchmarks, including region captioning, region VQA, grounded dialogue, and referring segmentation. Overall, CycleGRPO offers a straightforward and scalable way to advance pixel-level capabilities in MLLMs. Code and models are released at https://github.com/devinxzhang/CycleGRPO.

cs.CV↗
arXiv2026

SkillGLoW: Procedural-Family Skill Consolidation for Self-Improving Agents on Long-Horizon Task Streams

LLM agents increasingly self-improve by writing and reusing textual skills, kept either as one global document or as a flat pool of per-task entries, though most of the evidence comes from domains with structurally similar tasks. On long-horizon workloads where each task demands a different solution, the two forms fail in opposite ways: the document collapses into generic discipline, while the pool inflates and its entries stay bound to the instance that wrote them. We argue the missing unit of reuse is the solving procedure shared by a cluster of related tasks, and build SkillGLoW (Global-Local Weave) around it: the local skills a task writes from its own execution are aggregated into procedural families and compressed into de-instantiated global priors, while the instance detail they hold is regenerated per task rather than stored; a commit gate admits a prior only when real execution shows it does not degrade the deployed library. Across four benchmarks (mathematical reasoning, terminal automation, software repair, and embodied control) and three models, the priors gain 17.2 points (hard) over the no-skill baseline on average, with positive gains in all 12 continual-improvement runs, and 18.0 with local regeneration, while the library holds one prior per procedural family, 3.6x more compact than the per-task pool. Under the same protocol GLoW leads a published single-document optimizer on 15 of 21 cells. Unmodified, the library lifts success on unseen ALFWorld tasks from 73.9% to 83.9%, evidence that what transfers is procedure rather than task memory.

cs.AI↗
arXiv2026

Universal fingerprint of topological defect cores

Topological defects are ubiquitous in physics, arising from condensed matter physics to the early universe. Although there exist many universal scaling laws for correlations between topological defects, such as Porod scaling, the fingerprint of topological defect cores has remained largely unexplored. Here, we discover a universal scaling law in the region $k>1/ξ$, where $ξ$ is the healing length of the topological defects, taking the scaling of the form factor $S_f \propto k^{-(d+p+2)}$, where $d$ is the spatial dimension and $p$ is the defect codimension. We analytically prove that this exponent originates from a universal V-shaped cusp at the defect core and is independent of the underlying system and dynamics. Numerical simulations verify this scaling law in four typical frameworks: the time-dependent Ginzburg-Landau and Gross-Pitaevskii equations in the weak-coupling regime, the gauge/gravity duality model in the strong-coupling regime, and the Klein-Gordon equation in the Friedmann-Robertson-Walker background in cosmology. Our work provides a new probe for studying topological defects in systems ranging from superconductors to cosmological phase transitions.

hep-th↗
arXiv2026

BiMTokenizer: Preserving Semantic-Acoustic Balance in Low-Bitrate Speech Tokenization via Bidirectional State-Space Modeling

Speech codecs serve as bridges between continuous speech signals and large language models, yet face an inherent conflict between acoustic fidelity and semantic preservation. To mitigate this conflict, recent works increasingly adopt dual-tower architectures to decouple semantic and acoustic modeling with separate encoders. However, these dual-tower designs incur substantial architectural overhead. To avoid such complexity, we revisit the single-tower paradigm and propose BiMTokenizer, a low-bitrate speech codec (around 1.1 kbps) combining a bidirectional state-space backbone with Residual Spherical Leech Quantization (RSLQ). The bidirectional backbone strengthens temporal modeling, while RSLQ offers a fixed, well-separated lattice bottleneck for robust semantic and acoustic tokenization without learned-codebook collapse. Experiments show that BiMTokenizer achieves superior acoustic reconstruction and the lowest WER among low-bitrate codec baselines across both clean and noisy environments, while using less than half the parameters of recent dual-tower baselines. Furthermore, its robust semantic representations yield strong performance on downstream speech understanding tasks, confirming that a well-designed single-tower codec can preserve the semantic-acoustic balance at low bitrates. The code and model weights are available at https://github.com/ZhangXinWhut/BiMTokenizer.

cs.SD↗
arXiv2026

X-Coder: Advancing Competitive Programming with Synthetic Tasks, Solutions, and Tests

Competitive programming remains challenging for code LLMs. Despite recent progress, many training pipelines still depend on scarce real-world data, raising concerns about scalability and near-duplicate benchmark contamination. In this paper, we examine whether synthetic training artifacts can support the complete SFT-to-RL cycle for competitive programming: no real-world tasks, solutions, or test cases are directly used for post-training. To this end, we synthesize tasks, verified solutions, and reliable test cases that serve as reward signals for reinforcement learning. To improve reward reliability, we introduce a dual-verification strategy that reduces noise in both selected solutions and test outputs. Using this high-quality data, we train the X-Coder model series. X-Coder-14B achieves 67.5\% avg@8 on LiveCodeBench v5 and 63.4\% on v6, outperforming its base model by over 40 points. Further analysis provides practical insights into synthetic post-training, highlighting the value of diverse tasks, verified long-CoT supervision, and code-centric reinforcement. Our data and models are released at https://github.com/JieWu02/X-Coder

cs.CL↗
arXiv2026

Three-band dark-siren cosmology with intermediate mass black hole binaries: Synergy of Taiji, LGWA, and the Einstein Telescope

Gravitational-wave (GW) dark sirens provide an independent probe of the cosmic expansion history. Their cosmological constraining power, however, depends critically on precise luminosity-distance measurements and sky localizations for cross-matching with galaxy catalogs. Multiband GW observations can track GW events across different frequency bands and thus improve both. Motivated by this, we forecast the cosmological potential of intermediate-mass black hole binaries (IMBHBs) observed by a three-band GW detector network composed of Taiji (TJ), the Lunar Gravitational-wave Antenna (LGWA), and the Einstein Telescope (ET). We simulate detectable IMBHB populations and analyze them with a hierarchical Bayesian dark-siren framework that includes galaxy-catalog completeness and redshift uncertainties. We find that the TJ-LGWA-ET network outperforms all two-detector configurations considered here. In the $Λ$CDM model, it constrains the Hubble constant and matter density to $\sim 0.12\%$ and $\sim 0.6\%$, respectively. In the $w$CDM model, a 4-year dark-siren sample alone constrains the dark-energy equation-of-state parameter $w$ to $\sim 2.7\%$. Adding baryon acoustic oscillation (BAO) and Type Ia supernova (SNe Ia) data improves the $w$ constraint to $\sim 2.1\%$, slightly better than that from the current CMB+BAO+SNe Ia combination. We also show that the final constraints remain sensitive to IMBHB population assumptions and galaxy-catalog limitations, which highlights the need for deep galaxy surveys with precise redshift measurements.

astro-ph.CO↗
arXiv2026

CAER: Causal Action Effect Reweighting for World Model Training

World models are becoming core infrastructure for embodied intelligence, with action-conditioned video generation providing controllable predictions of how scenes evolve after agent interventions. Yet existing models are commonly trained with space-time-uniform mean squared error, allowing abundant background tokens to dominate the gradient while sparse interaction dynamics remain under-optimized; such uniform fitting rewards reconstructing appearance rather than learning how actions change the world. We introduce Causal Action Effect Reweighting (CAER), a general training paradigm that redistributes supervision toward the tokens whose predicted future is causally affected by the action. CAER contrasts the model's own predictions with and without action conditioning to localize these tokens online, then normalizes the resulting effect map into a weight that preserves the total coefficient mass and changes only where it is spent. This online signal requires no external annotations or offline preprocessing, avoids additional data-processing time, and scales naturally with model and dataset size. Experiments across heterogeneous action-conditioned world-model tasks show that CAER converges to better solutions than uniform MSE training, with consistent improvements in the physical consistency, controllability, and visual quality of generated videos.

cs.AI↗