Searcharxiv⌕ Search

arXiv subjects

Xinyi Wang

Publications and source records attributed to Xinyi Wang.

At least 37 records · Page 2Linked to original sources
arXiv2026

LEDOM: Reverse Language Model

Autoregressive language models are trained exclusively left-to-right. We explore the complementary factorization, training right-to-left at scale, and ask what reasoning patterns emerge when a model conditions on future context to predict the past. We train LEDOM, an open-source purely reverse autoregressive language model (2B/7B parameters, 435B tokens), and find it develops capabilities distinct from forward models, including abductive inference, question synthesis, and natural resolution of the reversal curse. We then explore one application of the reverse model: combining forward likelihood $P(y \mid x)$ with reverse posterior $P(x \mid y)$ through noisy channel duality. We propose Reverse Reward, which reranks forward outputs using reverse posterior estimates, and prove that bidirectional scoring penalizes hallucinated reasoning chains whose backward reconstruction degrades. Reverse Reward yields gains of up to 6.6\% on AIME 2024 and 15\% on AMC 2023 across multiple strong baselines. We release all models, code, and data here: https://github.com/Arvid-pku/LEDOM.

cs.CL↗
arXiv2026

SNR-Dependent Mismatched Filtering for Bistatic OFDM Ranging

This paper investigates the ranging performance of a bistatic integrated sensing and communications (ISAC) system employing orthogonal frequency-division multiplexing (OFDM), in which an ISAC transmitter emits a communication waveform carrying random data symbols, and a separate receiver performs ranging by correlating the received signal with a locally demodulated symbol sequence. Owing to inevitable demodulation errors, the ranging processor operates under mismatched filtering rather than ideal matched filtering, resulting in a delay-domain correlation response whose sidelobe structure explicitly depends on the signal-to-noise ratio (SNR). Focusing on frequency-flat fading channels, we derive closed-form expressions for the expected sidelobe level (ESL) and the average mainlobe level of the resulting mismatched ranging response for BPSK, QPSK, and general square QAM constellations. The analysis quantitatively characterizes how SNR-driven symbol decision errors reshape the delay-domain sidelobe behavior, thereby providing analytical insight into the SNR-dependent scaling behavior of ranging performance in bistatic OFDM-based ISAC systems. Simulation results validate the theoretical derivations and confirm the accuracy of the proposed analysis.

eess.SP↗
arXiv2026

Omni-Perception Policy Optimization for Multimodal Emotion Reasoning

We find that current emotion-oriented Omni-MLLMs still lack reliable omni-modal perception: they (i) underutilize multimodal cues in their reasoning trajectories and (ii) exhibit unfaithful behavior, often hallucinating modality-specific statements from other modalities. Building on these insights, we propose OPPO (Omni-Perception Policy Optimization), a reinforcement learning framework that explicitly optimizes multimodal perception. First, an Omni-Perception Reward decomposes ground-truth reasoning into fine-grained visual, acoustic, and emotion cues and rewards trajectories that semantically recover these cues. Second, an Omni-Perception Loss compares the policy under full and unimodally masked inputs, applying a KL penalty only to modality-specific evidence tokens to suppress cross-modal hallucination. We further introduce MEP-Bench, a diagnostic benchmark that quantifies utilization and faithfulness. Experiments show that OPPO achieves state-of-the-art performance on MER-UniBench and MME-Emotion, while substantially improving utilization and faithfulness scores on MEP-Bench, highlighting the importance of sufficient and faithful omni perception for multimodal emotion reasoning.

cs.AI↗
arXiv2026

Parker Solar Probe Observations of Preferential Heating of Protons over Alpha Particles near Turbulent Coherent Structures

Solar wind alpha particles exhibit preferential heating and acceleration relative to protons; however, their behavior in the vicinity of turbulent coherent structures remains less understood. We report the first evidence of localized alpha particle and proton heating within coherent structures identified using the Partial Variance of Increments (PVI) method, based on Parker Solar Probe (PSP) observations. Our results show that high-PVI events are associated with significant, species-dependent temperature enhancements: protons undergo a relative larger temperature increase than alpha particles. This preferential proton heating produces a localized decrease in the alpha-to-proton temperature ratio, indicating that the plasma is driven toward thermal equilibration between species. The heating is also anisotropic, being dominated by enhancements in the perpendicular temperature. These temperature-signatures coincide with a pronounced reduction in the normalized alpha-proton differential flow speed and a localized minimum in the Coulomb collision age, suggesting that the relaxation is affected primarily by collisionless kinetic effects. These findings provide new insight into the intermittent energy conversion and ion thermodynamics in the solar wind.

astro-ph.SR↗
arXiv2026

Selective Expert Guidance for Effective and Diverse Exploration in Reinforcement Learning of LLMs

Reinforcement Learning with Verifiable Rewards (RLVR) has become a widely adopted technique for enhancing the reasoning ability of Large Language Models (LLMs). However, the effectiveness of RLVR strongly depends on the capability of base models. This issue arises because it requires the model to have sufficient capability to perform high-quality exploration, which involves both effectiveness and diversity. Unfortunately, existing methods address this issue by imitating expert trajectories, which improve effectiveness but neglect diversity. To address this, we argue that the expert only needs to provide guidance only at critical decision points rather than the entire reasoning path. Based on this insight, we propose MENTOR: Mixed-policy Expert Navigation for Token-level Optimization of Reasoning, a framework that provides expert guidance only at critical decision points to perform effective and diverse exploration in RLVR. Extensive experiments show that MENTOR enables models capture the essence of expert strategies rather than surface imitation, thereby performing high-quality exploration and achieving superior overall performance. Our code is available online.

cs.AI↗
arXiv2026

AMBUSH: Collaborative Capture in Complex Environments with Neural Acceleration

Collaborative capture of dynamic targets is common in nature as an essential strategy for weaker species against the strong. Similar concepts have shown to be useful for numerous robotic applications, such as security and surveillance, search and rescue. However, most existing works focus on analytical and geometric solutions or end-to-end reinforcement learning methods, which are largely constrained to obstacle-free environments or scenarios with sparse, regularly distributed obstacles. This work tackles the problem from a unique perspective: the renowned strategy of``ambush'' alone would suffice for multiple slower pursuers to capture one faster evader with different levels of intelligence efficiently in complex environments. A parameterized strategy of ambush (including discrete and continuous parameters) is designed first, which takes into account the topological properties of the workspace, the truncated line-of-sight visibility, the relative speed ratio and the limited capture range. Then, a Hybrid Monte Carlo Tree Search (H-MCTS) algorithm is proposed to optimize the associated parameters through long-term planning, enabling the identification of highly promising parameters for future capture. Lastly, the neural acceleration is trained offline to learn the ranking of different choices of parameters across various environments, and to directly predict scores, replacing the rollout process in H-MCTS. The neural acceleration is adopted during online H-MCTS to accelerate the planning procedure while guaranteeing the planning quality. Its efficiency and effectiveness are validated in extensive simulations and hardware experiments, against evaders with different capabilities and intelligence levels, including two-times higher velocity and human-controlled behavior.

cs.RO↗
arXiv2026

RoboTacDex: A Dexterous Visual-Tactile-Action Dataset for Humanoid Manipulation

In the field of robot learning, large-scale and diverse demonstration trajectories provide the fundamental basis for enhancing robotic manipulation ability. We introduce RoboTacDex, a large, multi-modal, and diverse dataset of dexterous manipulation behaviors performed with a humanoid robot. Built on the publicly accessible humanoid robot Unitree G1, RoboTacDex consists of 6k trajectories covering 19 tasks, 23 skills, and interactions with 22 objects. RoboTacDex provides comprehensive records including multi-view RGB and depth information, tactile feedback, and detailed semantic annotations. Furthermore, the dataset features a variety of relatively challenging tasks that can only be completed by dual arms and dexterous hands, aiming to mimic human-like operational logic and simulate real-world manipulation complexity. To ensure data collection quality, we develop an improved multi-camera synchronization system to enable millisecond data synchronization and recording of modalities. In our experiments, we evaluate three representative imitation learning models on our dataset, analyzing their performance as well as their respective strengths and limitations across different task categories. Successful trial results and a moderate level of generalization capabilities across a suite of tasks indicate the effectiveness and diversity of the collected dataset. Our dataset will be open-sourced soon.

cs.RO↗
arXiv2026

OASIF: An Efficient Obfuscation-Aware Self-Improving Framework for LLM-Based Assembly Code Instruction Following and Comprehension

Large Language Models (LLMs) have recently shown promise in automated binary analysis, yet they remain brittle under commercial-grade obfuscation. We present OASIF, an Obfuscation-Aware Self-evolving Instruction-Following framework for obfuscated assembly comprehension. OASIF couples a token-efficient assembly encoder with a lightweight projector to expose long obfuscated code to a pretrained code LLM under a bounded context budget and follows a three-phase training: (i) feature-space alignment, (ii) supervised instruction fine-tuning, and (iii) online self-evolving reinforcement learning with hybrid rewards, enabling continual adaptation with minimal manual verification. On VMISA-Bench, a challenging out-of-distribution suite featuring three commercial VM-based obfuscators, OASIF consistently improves open-source backbones; Qwen2.5-Coder-Instruct-14B attains Success Rate gains of +15.9, +5.8, and +16.9 percentage points (pp) on Code Virtualizer, Themida (v3.0.7), and VMProtect (v3.5), respectively, and improves the OASIF-Bench average by +9.8. OASIF further delivers stable gains across seven standard BCSD benchmarks while preserving general and domain-relevant capabilities on HumanEval, VulBench, and HumanEval-Decompile.

cs.SE↗
arXiv2026

From Outcomes to Actions: Leveraging Hindsight for Long-Horizon Language Agent Training

Reinforcement learning (RL) has become a widely adopted technique for improving large language models (LLMs) on complex tasks. Despite this progress, existing RL methods still face challenges in training agents with longer-horizon interactions. One major bottleneck is distinguishing the contribution of different actions in long-horizon interaction, leading to high optimization variance. To address this, we introduce a novel policy gradient method, Hindsight Policy Optimization (HPO), that projects both the current policy distribution and the hindsight distribution into an intent space and extracts low-variance learning signals from the Wasserstein distance between them. We theoretically and empirically show that aggregating semantically similar states and actions in the intent space yields a bounded-variance estimator and improves policy performance stably. Our code is available online.

cs.LG↗
arXiv2026

MER-R1: Multimodal Emotion Reasoning via Slow-Fast Thinking Synergy

We find that explicit reasoning does not necessarily translate into better multimodal emotion recognition (MER) accuracy, even though it makes predictions more interpretable. Specifically, for reasoning-based MLLMs, fast thinking by triggering direct answers often outperforms slow thinking after deliberative reasoning. Our empirical analyses show that fast thinking improves recall with broader and more confident predictions, whereas slow thinking favors precision through conservative filtering of incorrect categories. Building on these insights, we propose MER-R1, a reinforcement learning framework that turns slow-fast complementarity into explicit optimization. Dual-objective disentanglement separates recall and precision into two optimization signals, allowing them to be jointly optimized rather than traded off against each other. Slow-fast confidence calibration further aligns the final slow-thinking answer with fast-thinking intuition, strengthening correct emotions while suppressing incorrect ones. In this way, MER-R1 unifies the recall-oriented intuition of fast thinking with the precision-oriented selectivity of slow thinking. We further provide theoretical justification for this synergy, showing that it mitigates variance-induced interference during optimization. Extensive experiments on MER-UniBench and MME-Emotion show that MER-R1 achieves state-of-the-art performance and makes reasoning genuinely benefit emotion recognition.

cs.AI↗
arXiv2026

Performance Analysis for Heterogeneous Air-Ground ISAC in Coordinated Multipoint Networks

The emergence of the \textit{low-altitude economy} (LAE) calls for highly integrated and reliable wireless systems that can simultaneously support \textit{communication and sensing} (C\&S) functions. Although \textit{integrated sensing and communication} (ISAC) has been widely studied, most existing works focused on link-level or single-cell architectures in terrestrial environments, leaving the potential of network-level cooperative air-ground ISAC largely unexplored. To bridge this gap, a heterogeneous air-ground ISAC network architecture based on \textit{coordinated multipoint} (CoMP) is proposed, which incorporates a cooperative hybrid mono/bi-static sensing scheme to enhance spatial diversity and sensing capability. In the proposed architecture, a two-tier \textit{base station} (BS) deployment is adopted: master BSs are arranged in a hexagonal lattice, while slave BSs follow a Poisson point process distribution. This structure concurrently supports communication for terrestrial users and sensing for aerial targets. A holistic performance analysis framework for both C\&S is further developed, accounting for key channel and network parameters. Simulation results reveal inherent trade-offs between C\&S performance, especially under multi-BS cooperation and varying network density. These findings provide practical guidance for the deployment of scalable and efficient ISAC networks in LAE scenarios.

cs.IT↗
arXiv2026

Don't Tell the Answer, Truly Guide the Reasoning During RL Rollouts

Reinforcement Learning (RL) has become a key driver for enhancing the long chain-of-thought (CoT) reasoning capabilities of Large Language Models (LLMs). However, prevalent methods like GRPO often fail when task difficulty exceeds model capacity, leading to reward sparsity and inefficient training. Prior work attempts to mitigate this with off-policy data, but such methods often induce severe distributional mismatches that destabilize policy updates. In this work, we identify a core issue underlying these failures, which we term low training affinity, and introduce Affinity, the first quantitative metric for monitoring the compatibility between external guidance and the model's intrinsic policy. To address this, we propose HINT, an adaptive framework designed to enhance reasoning capabilities while explicitly preserving high Affinity. First, instead of revealing partial answers, HINT supplies Meta-Hints, which act as abstract cognitive scaffolding to guide the model in articulating solutions independently. Second, to ensure stability, we integrate Affinity-Aware Policy Optimization (AAPO), which dynamically modulates the learning objective based on the Affinity. Extensive experiments across diverse benchmarks demonstrate that HINT consistently outperforms strong baselines, while exhibiting superior stability and robust generalization to out-of-distribution tasks. Code is available at https://github.com/ViviqwerAsd/HINT.

cs.LG↗
arXiv2026

ADaPT: Token-Level Decoupling for Efficient Large Reasoning Models

Large reasoning models rely on long chain-of-thought to achieve strong performance, but applying such reasoning uniformly incurs high computational cost. Existing efficiency-oriented methods attempt to shorten or mix reasoning strategies, yet often degrade reasoning capability. We identify the root cause as sequence-level coupling between efficiency incentives and correctness optimization, which implicitly penalizes long but correct reasoning trajectories. To address this issue, we propose Adaptive Dual-Process Thinking (ADaPT), a token-level dual-process framework that explicitly decouples efficiency and correctness signals during training. ADaPT introduces a mode-selection token to control fast and slow reasoning, applying efficiency-related rewards exclusively to this token to avoid penalizing correct long reasoning while encouraging efficiency when appropriate. Moreover, ADaPT enables precise and continuous control over the efficiency-performance trade-off at inference time: by adjusting the generation probability of the mode-selection token, a single trained model can smoothly move along the efficiency-performance Pareto frontier. Extensive experiments demonstrate that ADaPT significantly reduces inference cost while maintaining strong reasoning performance across multiple benchmarks.

cs.LG↗
arXiv2026

Hubs or Fringes: Pretraining Data Selection via Web Graph Centrality

The performance of modern language models depends critically on pretraining data composition. Yet existing data selection methods rely on auxiliary classifiers for document scoring or mixture optimization, adding computational overhead and dependence on labeled data. We propose WebGraphMix, a lightweight data selection framework that computes structural centrality scores over the Common Crawl host-level web graph and uses them to vary the proportion of central versus peripheral documents in the pretraining mixture. We hypothesize that central hosts expose models to reusable abstractions, while peripheral hosts encode specialized, long-tail knowledge. WebGraphMix computes centrality scores efficiently at web scale, requiring no model training, labeled data, or downstream supervision. We integrate WebGraphMix into the DataComp-LM pipeline and train models at 400M and 1B parameter scales with 8B and 28B tokens respectively, evaluating on 23 tasks ranging from factual knowledge to symbolic reasoning. Our experiments show that central and peripheral web regions encode complementary capabilities. Mixture combining both at a ratio of 1:1 achieves 41.4% on average, compared to 39.8% for uniform sampling. Combining structural scores with document-level quality classifier scores further improves performance to 43.8%. These findings demonstrate that web graph topology is a meaningful axis for pretraining data curation, capturing information that is largely orthogonal to existing content-based approaches.

cs.CL↗
arXiv2026

Finding the Minimal Parameter Budget for Implicit Reasoning: A Data Complexity Driven Scaling Law for Language Models

Reasoning is a core capability of language models (LMs), yet it remains unclear how much model capacity is necessary to support reasoning during pretraining. In this work, we study the minimal parameter budget required for implicit reasoning, defined as the ability to infer new facts from learned knowledge without explicit chain-of-thought supervision. To isolate this phenomenon, we pretrain LMs from scratch in a controlled synthetic environment that mimics the structure and distribution of real-world knowledge graphs, and evaluate their ability to complete missing edges via multi-hop inference. From both a theoretical and an empirical perspective, we identify a scaling law linking this optimal parameter budget to a graph search entropy measure. Across a wide range of model sizes, training steps, and graph complexities, we show that an optimally sized language model can reliably reason over approximately 0.008 bits of information per parameter at most. Our results characterize the minimal sufficient capacity for implicit reasoning during pretraining. Our findings provide principled guidance for matching model size to data complexity and offer new insights into the scaling behavior of reasoning in large language models.

cs.AI↗
arXiv2026

Distributed MoE-based Uplink Detection for Cell-Free Communication Systems

Cell-free Massive multiple input and multiple output (MIMO) is recognized as a key technology for beyond-5G networks, where distributed access points (APs) jointly serve user equipments (UEs) to address the inherent inter-cell interference issue inherent in cellular systems. While conventional distributed signal detection methods offer a practical balance between performance and fronthaul load, they are fundamentally limited by linear processing constraints. In this paper, we propose a novel deep learning based uplink detection framework by introducing the distributed mixture of experts detection network (DMoE-DetNet). In this architecture, each AP acts as a local expert employing convolutional neural networks (CNNs) for non-linear feature extraction, and transmits the local minimum mean square error (MMSE) detection results and statistical channel information to the central processing unit (CPU). In the CPU, an attention-based encoder module captures complex spatio-temporal dependencies among users for global feature fusion, with a gating network at the central processor dynamically weighting the contributions from different APs. At last, a linear detector outputs the symbol probability. Simulation results demonstrate that the proposed DMoE-DetNet significantly outperforms conventional linear processing based cell-free signal detection methods in terms of symbol error rate, showcasing the potential of artificial intelligence-enabled communication systems.

eess.SP↗
arXiv2026

JMed48k: A Multi-Profession Japanese Medical Licensing Benchmark for Vision-Language Model Evaluation

We introduce JMed48k, a multi-profession Japanese healthcare licensing benchmark for evaluating vision-language models. Built from official PDF materials released by the Japanese Ministry of Health, Labour and Welfare, JMed48k contains 48,862 exam questions and 20,142 images from 11 national licensing examinations between 2005 and 2025, with visual content annotated under an 8-type taxonomy. From this corpus, we derive JMed48k-Eval, a recent five-year evaluation subset with 12,484 scored questions, including 9,905 text-only questions and 2,579 questions with images. We evaluate 21 proprietary, open-source, and medical-specific models, reporting text-only and with-image performance separately. Because these subsets contain different questions, we further introduce a paired image-removal audit that evaluates questions with images before and after removing visual content to explore four answer-transition states. The audit shows that proprietary and open source models gain substantially from images, whereas medical-specific systems show limited observable use of visual evidence, with many correct answers persisting after image removal. Even among proprietary models, the net image-removal effect varies sevenfold across professions, from +5.7 points on Physician questions to +39.8 points on Public Health Nurse questions. We release JMed48k to support reproducible, profession-stratified evaluation of vision-language models in medical licensing settings.

cs.CV↗
arXiv2026

Atomic Skills are the Prerequisite: When Reinforcement Learning Synthesizes Compositional Reasoning, and When It Only Amplifies

Does Reinforcement Learning (RL) merely amplify existing skills, or synthesize novel skills? We investigate this question through the lens of Complementary Reasoning: the critical practical capability of integrating internal knowledge with external context, a prerequisite for reliable Continual Learning and Retrieval-Augmented Generation. To avoid pre-training contamination, we construct a controlled semanticsynthetic dataset of biographies and decompose this capability into two atomic skills: Parametric Reasoning (retrieving facts encoded in model weights) and Contextual Reasoning (processing novel in-context information). We present two findings. First, models supervised directly on the composite task reach high accuracy on seen facts and reasoning paths (90%) but collapse on novel facts and reasoning paths (18%), indicating that Supervised Fine-Tuning (SFT) relies on rote memorization rather than genuine skill integration. Second, RL bridges this generalization gap, acting as a skill synthesizer rather than a mere amplifier--but only under a strict prerequisite: it synthesizes new composite strategies only when the base model has first mastered the independent atomic skills via SFT. These results suggest that decoupled atomic training followed by RL offers a scalable path to complex novel reasoning.

cs.AI↗