arXiv subjects

Yang Wang

Publications and source records attributed to Yang Wang.

At least 19 recordsLinked to original sources

arXiv2026

DocSkill: Skill-Integrated Agentic Reasoning for Multilingual Multi-Hop Document Question Answering

Multilingual multi-hop document question answering (MMHop-DocQA) requires locating and integrating evidence across complex layouts, document pages, and diverse languages. Existing approaches include single-pass processing and adaptive tool use, but learning to coordinate evidence acquisition, cross-page integration, and answer verification remains challenging. To address this challenge, we present DocSkill, an agentic framework with a learned skill-orchestration policy for adaptive evidence acquisition and reasoning. The policy selects, revisits, or skips skills according to the evolving evidence state. To construct supervision for skill orchestration, bounded Monte Carlo Tree Search (MCTS) explores alternative skill trajectories offline. These trajectories provide demonstrations for supervised fine-tuning (SFT) and preference pairs for Direct Preference Optimization (DPO). We further introduce MMHop-DocBench, a multilingual multi-hop document QA dataset with training and test splits, containing 60K QA pairs across 10 languages and eight document domains. Its HopGen pipeline builds hierarchical page-element graphs to derive traceable cross-page evidence paths and generate grounded QA pairs with page- and element-level evidence annotations. Experiments show that DocSkill with Qwen3-VL-8B achieves 41.0 accuracy and 49.2 F1 on MMHop-DocBench, exceeding the strongest evaluated baseline by 5.5 and 6.3 points, respectively, and achieving competitive performance on MMLongBench-Doc and DocBench.

cs.CL↗

arXiv2026

Rethinking Scientific Discovery in the Agentic Era

Artificial intelligence has advanced scientific discovery, but most AI4Science systems remain fragmented tools that rely on humans to coordinate problem formulation, literature grounding, model use, simulation, validation, and knowledge reuse. This paper presents \textbf{SCION (Scientific Collaborative Innovation with Agentic Organizational Nexus)}, an agentic scientific operating system that acts as an \textbf{organizational nexus}. Through a Science Agent serving as a \textbf{Meta-Harness}, SCION connects scientific tasks, tools, agents, artifacts, and memory, transforming research into an executable, auditable, and reusable operational process. At its core is the \textbf{Research Execution Plan (REP)}, which compiles high-level scientific intent into staged objectives, dependencies, verification checkpoints, tool requirements, expected artifacts, and fallback conditions. SCION further integrates hierarchical multi-agent execution, profile-driven specialization, selective context construction, governed delegation, and layered epistemic memory to support long-horizon scientific work. We formulate discovery under SCION as \textbf{Target-conditioned Inverse Search} and extend it to hidden-target settings through batch active search under finite experimental budgets. Applications in materials analysis, molecule design, and protein or antibody screening, together with experiments on scientific reading, idea generation, molecule generation, and antibody screening, show that SCION outperforms existing autonomous research-agent baselines, especially in decomposition, verification, refinement, and memory reuse. Overall, SCION shifts AI from isolated tools toward a coordinated operational layer for traceable and reusable scientific innovation.

cs.CL↗

arXiv2026

Clustering Templates: A Data-Driven Framework for Dimensionality Reduction and Compression of Brain Strain in Head Impact

Purpose: Finite element (FE) models are indispensable for studying the biomechanics of traumatic brain injury (TBI), and modern head FE models typically comprise thousands to millions of elements. The resulting brain strain is therefore high-dimensional, and when combined with dense temporal sampling and large numbers of head impact cases, this produces datasets of substantial data volume. Accordingly, this study investigates the effective dimensionality required to accurately represent brain strain and proposes a data-driven framework to reduce the dimensionality of FE simulation data while maintaining high spatiotemporal fidelity. Methods: We developed a series of "clustering templates" by applying K-means clustering to feature vectors derived from the maximum principal strain (MPS) time-history curves of 118,840 brain elements over 3,693 impact simulations using the THUMS head model. The performance of these templates, with cluster counts (K) ranging from 8 to 8192, was rigorously evaluated by quantifying the fidelity of data reconstruction. Results: The clustering templates demonstrated high-fidelity reconstruction of the original FE data. A template with K = 512 achieved a compression ratio of over 200:1 while maintaining a low mean element-wise peak MPS root mean square error (RMSE) of 0.027, and a template with K = 4096 (30:1 compression ratio) achieved an RMSE of 0.016. Conclusion: The proposed clustering templates provide an effective and automated method for the dimensionality reduction and compression of large-scale brain strain data. This framework mitigates critical data management bottlenecks and enables the development of subsequent data analysis and machine learning models, thereby accelerating research in TBI biomechanics.

physics.bio-ph↗

arXiv2026

LeanTTS: Efficient Full-Stack Bilingual TTS for Zero-Shot Cloning and Voice Customization

While recent advances in text-to-speech (TTS) synthesis have substantially improved speech quality and modeling flexibility, the development of lightweight and computationally efficient systems remains comparatively underexplored. We present LeanTTS, a bilingual complete synthesis stack that supports two complementary regimes: reference-prompt zero-shot cloning for unseen speakers and token-conditioned customization for recurring voices. LeanTTS combines a compact Mandarin polyphone front end, a structurally compressed flow-matching acoustic model, and a lightweight Vocos decoder with four-step MeanFlow generation. The resulting system contains 18.9M learned parameters, 86.12% fewer than the 136.2M-parameter ZipVoice-Vocos pipeline. On Seed-TTS-Eval, its zero-shot mode obtains 2.02 CER/62.3 SIM on Mandarin and 2.32 WER/48.3 SIM on English, exposing the quality-size trade-off under aggressive compression. With token-conditioned customized voices, the same system reaches 1.57 CER/77.2 SIM on Mandarin and 1.93 WER/76.6 SIM on English, with UTMOS scores of 3.03 and 4.33, respectively. It further achieves mean/P90 CPU RTFs of 0.049/0.061, establishing a sub-20M operating point for bilingual personalization and real-time CPU deployment. Audio demos are available at https://github.com/notlu/leantts.

cs.SD↗

arXiv2026

Broken Symmetry in LLM Refusal: Answer Release Is More Local Than Refusal Restoration

When a language model refuses to answer a prompt, it is unclear whether the correct answer is erased from its internal representations, or merely suppressed at the output layer. We investigate this mechanism using a controlled withhold setting, which yields perfectly matched answering and refusal trajectories for bidirectional activation patching. We uncover a causal asymmetry in intervention locality under matched causal interventions, which we term broken symmetry. Even when a model generates a clean refusal, the correct answer remains linearly recoverable from its hidden states. Furthermore, releasing this withheld answer is a highly local operation, requiring only a single-position patch. Conversely, the reverse operation is not equally local: reimposing suppression requires broader interventions across multiple positions, and assembling a coherent refusal sequence is more difficult still. We further demonstrate that while an average answer-to-refusal displacement vector marks the geometric difference between these states, it fails to act as a reliable, reversible linear control toggle between behaviours. Taken together, our findings show that refusal does not function as a simple symmetric switch. For safety and auditing, this implies that probe recoverability can overestimate true behavioural control, and locating refusal-relevant directions does not reliably grant the ability to steer a model from answering to coherent refusal.

cs.AI↗

arXiv2026

SimSkip: StrongARM-Comparator Similarity Sensing with SAR-Skipping-Aware Training for Analog Compute-in-Memory

Analog-to-digital conversion dominates the energy and area of analog compute-in-memory (CIM) accelerators. This paper presents SimSkip, a successive-approximation-register (SAR) skipping scheme that combines a lightweight StrongARM-comparator similarity detector with SAR-skipping-aware training. When two held voltages are similar, the upper bits of the preceding code are reused to skip SAR comparisons. Detector-induced disturbance extracted from 55-nm CMOS simulations and bit-exact prefix saturation are embedded in the forward pass, and the network learns to tolerate them through quantization-aware training and knowledge distillation. On an ISAAC-based W4A4 WRN-28-10, SimSkip achieves 77.23% CIFAR-100 top-1 accuracy (mean of three seeds) versus 78.4% for the baseline, while reducing SAR comparisons by 48.2% across three hardware-modeled layers. A first-order model estimates a 27.2% net ADC energy reduction after detector overhead.

cs.AR↗

arXiv2026

EMMA: Enhancing Robotic Control Robustness through a Generative Visual Model

The generalization of vision-language-action (VLA) models heavily relies on diverse training data. However, acquiring large-scale robot manipulation data across varied object appearances is costly and labor-intensive. To address this, we introduce Embodied Manipulation Media Adaptation (EMMA), a framework for augmenting VLA policies that combines a generative data engine with an effective training pipeline. We introduce DreamTransfer, a diffusion Transformer-based architecture for generating multi-view consistent and geometrically plausible embodied manipulation videos. It supports text-driven video editing on foreground, background and lighting while preserving 3D structure and geometric fidelity. We also utilize a hybrid training set of real and generated data and propose AdaMix to enhance the training process. AdaMix is a training strategy that adaptively weights samples according to policy performance to emphasize challenging samples. Evaluations show DreamTransfer outperforms previous video generation models in multi-view consistency, geometric fidelity, and text-conditioning precision. In real-world robotic tasks with zero-shot visual settings, our framework achieves a relative performance increase of over 92% compared to training with real data alone, and improves by an additional 17% with AdaMix.

cs.AI↗

arXiv2026

Opening Knowledge Gaps Drives Scientific Progress

Knowledge production is often viewed as an endogenous process in which discovery arises through the recombination of existing theories, findings, and concepts. Yet not all recombinations are equally valuable, and identifying those that may prove most generative remains challenging. We propose that a crucial form of recombination occurs when linking concepts creates knowledge gaps---empty regions in the conceptual landscape that focus scientific attention on unexplored connections. Using computational topology, we develop a method to identify knowledge gaps in science at scale. In this framework, gaps form when new connections complete cycles in evolving networks of scientific concepts. We apply the method to 34,363,623 papers in Microsoft Academic Graph spanning 1900--2020. The main analyses use samples ranging from 8,643,571 to 31,983,201 papers. Compared with papers that introduce novel concept pairs without opening gaps, gap-opening papers have greater long-term citation impact, are more disruptive of prior work, and exhibit greater delayed recognition. Gap-opening papers receive fewer citations within five years of publication but more over twenty years, consistent with scientific communities needing time to recognize and develop newly opened research territories. Evidence for the generative capacity of gap-opening papers---as indicated by high citation impact and disruption---replicates in five additional databases (OpenAlex, the PubMed Knowledge Graph, Dimensions, American Physical Society journals, and United States patents) and across eight concept vocabularies, including several derived directly from text. Together, these results reveal a distinctive generative role for gap opening in scientific discovery.

cs.CY↗

arXiv2026

Which Agent's Skill Should Change? Group-Relative Credit Assignment for Evolving Multi-Agent Teams

Cooperative language-model agents can improve without weight updates by revising reusable skills, but only if a shared team reward is traced to the responsible agent and skill. Existing methods credit the skills present in each trajectory, yet presence is not responsibility: a harmful skill active in successful runs still earns positive credit. We present \textbf{GRSE}, \textbf{G}roup-\textbf{R}elative \textbf{S}kill \textbf{E}volution, which assigns behavioral credit by comparison: inspired by GRPO, it executes each question several times under the same skill libraries and contrasts the executions to credit agent--skill pairs, with structured diagnosis of failures. On HotpotQA and LoCoMo with two open-weight actors, GRSE outperforms MASkills, the closest prior method, on HotpotQA and every LoCoMo category except open-domain, with the largest gains on multi-hop questions ($+6.7$ and $+3.2$ F1). Its gains rest on better-grounded retrieval rather than answers given without evidence, and its learned skills transfer between the multi-hop tasks. GRSE is robust to corrupted libraries: starting from planted harmful skills, it credits every planted skill with the correct sign and recovers the score of a clean library ($+6.3$ points), whereas MASkills misattributes two and falls below its starting point.

cs.CL↗

arXiv2026

Evolution and breakdown of chiral anomaly in Weyl semiconductor tellurium

Chiral anomaly in topological materials manifests as negative longitudinal magnetoresistance, yet the tunability and evolution of this effect remain elusive. Here we show that doped tellurium serves as a versatile platform to realize a giant negative longitudinal magnetoresistance that is highly tunable via the Fermi level and magnetic field. By moving the Fermi level towards the Weyl point, we observe a pronounced enhancement of negative magnetoresistance, which arises from chiral anomaly and robustly persists up to ambient temperature. Beyond the quantum limit, the system exhibits positive linear magnetoconductance, serving as a compelling transport signature of the chiral anomaly in the lowest Landau level. Notably, in the extreme quantum limit at cryogenic temperatures, the magnetoresistance undergoes a dramatic reversal from negative to positive. This reversal points to a potential field-induced phase transition into a collective electronic state driven by enhanced electron-electron interactions. Our findings provide valuable insights into the unconventional chiral transport and emergent correlated quantum states under extreme magnetic fields.

cond-mat.mtrl-sci↗

arXiv2026

Thinking-Free Policy Initialization Makes Distilled Reasoning Models More Effective and Efficient Reasoners

Reinforcement Learning with Verifiable Reward (RLVR) effectively solves complex tasks but demands extremely long context lengths during training, leading to substantial computational costs. While multi-stage training can partially mitigate this, starting with overly short contexts often causes irreversible performance degradation, ultimately failing to reduce overall training compute significantly. In this paper, we introduce Thinking-Free Policy Initialization (TFPI), a simple yet effective adaptation to RLVR that bridges long Chain-of-Thought (CoT) distillation and standard RLVR. TFPI employs a simple ThinkingFree operation, explicitly discarding the thinking content via a direct append, to reduce token usage during inference. Training with ThinkingFree-adapted inputs improves performance and lowers token consumption, even in the original slow-thinking mode. Extensive experiments across various benchmarks have shown that TFPI accelerates RL convergence, achieves a higher performance ceiling, and yields more token-efficient reasoning models without specialized rewards or complex training designs. With TFPI only, we train a 4B model to reach 89.0% accuracy on AIME24 and 65.5% on LiveCodeBench using less than 4K H20 hours.

cs.LG↗

arXiv2026

Composition-RL: Compose Your Verifiable Prompts for Reinforcement Learning of Large Language Models

Large-scale verifiable prompts underpin the success of Reinforcement Learning with Verifiable Rewards (RLVR), but they contain many uninformative examples and are costly to expand further. Recent studies focus on better exploiting limited training data by prioritizing hard prompts whose rollout pass rate is 0. However, easy prompts with a pass rate of 1 also become increasingly prevalent as training progresses, thereby reducing the effective data size. To mitigate this, we propose Composition-RL, an offline data-augmentation approach that composes existing prompts before RL to address the pass-rate-1 bottleneck without modifying the RL optimizer. The composed dataset is generated and verified once from the fixed source pool and then used for RL training. Extensive experiments across model sizes from 4B to 30B show that Composition-RL consistently improves reasoning capability over RL trained on the original dataset. Performance can be further boosted with a curriculum variant of Composition-RL that gradually increases compositional depth over training. Additionally, Composition-RL enables more effective cross-domain RL by composing prompts drawn from different domains. All our code, data, and models are publicly available to facilitate future research.

cs.CL↗

arXiv2026

Q-Reflector: Learning GUI Action Reflection with Generated Conditions

Mobile GUI agents require accurate action reflection for reliable long-horizon execution, assessing whether each action was successfully executed and whether its outcome constitutes a valid step toward the task goal. When evaluated as standalone reflectors under a common prompt, frontier VLMs and specialized GUI models exhibit imbalanced recognition of valid and invalid transitions. We propose Q-Reflector, a standalone reflector based on Qwen3-VL-8B that learns GUI action reflection as supervised structured prediction. Given task context, the executed action, and paired screenshots, it generates pre- and post-conditions specifying the action's prerequisites and intended local effect, then assesses execution success and task validity through a structured rationale and binary verdict. Training combines supervised fine-tuning, teacher-student distillation, and reward- and preference-based refinement. On 1,082 manually verified AndroidWorld transitions, Q-Reflector achieves 82.16% overall accuracy, exceeding all evaluated baselines and showing more balanced performance than GPT-5.2 and GUI-Owl-1.5-8B. On 117 MobileWorld tasks, Q-Reflector achieves task success rates of 46.15% with GUI-Owl-1.5-8B-Instruct and 35.90% with MAI-UI-8B, improving over no external reflection by 17.09 and 14.53 percentage points, respectively. Under the same condition-generation setting, it outperforms the strongest reflection baselines, GPT-5.2 and Claude Haiku 4.5, by 3.42 and 5.98 percentage points, respectively.

cs.AI↗

arXiv2026

Not All Graphs Are Maps: A Task-Agnostic Exploration Framework for Mobile GUI Navigation

Modern mobile GUI agents can perceive visible screens and execute local actions, but they still lack persistent app-specific navigation knowledge: where functions live, which pages contain them, and which screens are equivalent across repeated visits. We introduce TAG-Nav, a Task-Agnostic Graph Navigation framework that wraps existing GUI-agent backbones with reusable destination memory. Offline exploration is distilled into compact records of app functions, target pages, and waypoint routes before any evaluated instruction is seen. When a future task's destination is covered, TAG-Nav retrieves the destination and provides route guidance without task-specific demonstrations. This separates mobile GUI destination navigation from full task completion: many tasks share the prefix of reaching the right page, while final actions may enter content, send data, make purchases, delete information, or change sensitive settings. To evaluate route finding without downstream execution confounds, we introduce DestBench, a navigation-only protocol that asks whether an agent reaches the correct pre-action destination screen. DestBench enables a controlled graph-memory study comparing no-memory agents with random, automated BFS, and clean graph memory. Graph-derived memory improves strict navigation success rate of UI-TARS-1.5-7B from 52.4% without graph memory to 58.3% with automated BFS memory, and improves Qwen-VL-Max success rate from 59.5% to 71.4% on the same conditions. Together, TAG-Nav and DestBench show that destination coverage and route cleanliness strongly shape graph-memory value, making graph quality a measured variable rather than an assumed benefit of exploration.

cs.AI↗

arXiv2026

Which Trajectory Is Safer? Combining Rule and Learning for Safety Evaluation

Collision rate remains a dominant metric for evaluating autonomous driving trajectory safety, but its binary nature obscures substantial safety differences among collision-free trajectories. Existing surrogate-safety metrics provide finer-grained assessments, yet often trade off physical interpretability against adaptivity to interaction context. We introduce FutureRisk, a continuous metric that quantifies the latent future risk of a trajectory by how closely and persistently its realized interactions approach unsafe states. At its core, our Context-Adaptive Risk Model (CARM) combines a context-dependent safety-time distribution with an independently supervised critical-event-time distribution. Its safety criterion is bounded around a closed-form physical prior, while a bounded residual provides additional adaptation without abandoning this physical anchor. Experiments on DeepAccident and Bench2Drive show that FutureRisk achieves stronger agreement with native collision labels than existing surrogate-safety measures, discriminates risk among collision-free trajectories, and transfers zero-shot across datasets. Comparisons across four planners further show that FutureRisk reveals safety differences hidden by binary collision outcomes. Together, these results establish FutureRisk as a fine-grained, interpretable, and context-adaptive metric for trajectory safety evaluation. Code is available at https://github.com/Xuewen2025/safety-evaluation.

cs.RO↗

arXiv2026

TStruct: Learning Shared Temporal Structures for Long-Term Time-Series Forecasting

Long-term time series forecasting requires temporal dependency modeling that remains stable and generalizable under temporal distribution shifts. However, real-world time series are often noisy and non-stationary, making deep learning models prone to fitting incidental temporal interactions in the training data. A key question is whether there exist temporal relationship structures that recur across different temporal positions and variables. We analyze temporal relationship matrices on commonly used time series datasets and find that samples from different temporal positions and different variables exhibit shared dominant temporal relationship structures. This indicates that temporal relationship structures contain patterns that recur stably across samples and variables. To learn such shared structures, we propose \textbf{T}emporal \textbf{Struct}ure Network (\textbf{TStruct}), a simple yet effective network for directly learning shared and generalizable temporal dependencies. Specifically, TStruct directly parameterizes a set of shared temporal dependency matrices and generates normalized combination weights using dynamic sample information and static variable priors. These weights represent a soft allocation of each sample-variable pair over shared temporal dependency structures, enabling the model to generate adaptive temporal dependency matrices under shared structural constraints, rather than fitting incidental interactions in a fully unconstrained manner. Notably, this simple structure-aware design consistently outperforms several strong baselines across all datasets, achieving superior accuracy while maintaining high computational efficiency.

cs.LG↗

arXiv2026

DTA-GT: Direction- and Topology-Aware Graph Transformer for Neural Network Representation Learning

Architecture attribute predictors reduce the cost of neural architecture search by estimating accuracy or latency from a small set of evaluated candidates. The key difficulty is that neural architectures are operation-labeled DAGs: edge direction defines computational flow, but existing predictors often treat direction as a local structural cue. We propose DTA-GT, a Direction- and Topology-Aware Graph Transformer based on encoder-wide directional consistency, which preserves directed computational semantics across node initialization, global spectral encoding, pairwise interaction, and feature update. DTA-GT realizes this principle with direction- and topology-aware node initialization, Magnetic Laplacian spectral encoding, direction-dual structural attention, and a Direction-Sensitive MoE-FFN. Across accuracy and latency prediction on NAS-Bench-101 and NAS-Bench-201, DTA-GT consistently outperforms representative sequence-, GNN-, Transformer-, and hybrid predictors under limited-budget settings. Ablations and Magnetic spectral controls show that the improvements come from coordinated direction-aware stages and edge-orientation-sensitive spectral information, while mechanistic diagnostics confirm that the modules exhibit the intended directional behavior. These results support encoder-wide directional consistency as an effective design principle for architecture DAG representation.

cs.LG↗

arXiv2026

Calibrated Typed Decisions for Entity, Relation, and Event Extraction

Information extraction models often perform well on the datasets they are trained on, but transferring to new domains and annotation guidelines remains challenging. Supporting user-specified types alone does not resolve this problem. We present CalOn, a unified 0.6B-parameter encoder for entity, relation, and event extraction across datasets and schemas. The model jointly encodes text and type definitions, then directly identifies entity spans, relations between spans, and event triggers and arguments. Its extraction decisions consider the requested types together, using their combined evidence to identify relevant candidates before assigning a type. This gives all three tasks a common probabilistic formulation while keeping every extracted mention anchored to the input text. We train on diverse sources, and supplement supervised learning with a reward that encourages correct extractions and penalises incorrect ones. Across 40 benchmarks, CalOn outperforms the evaluated baselines on most benchmarks. It leads all reported relation extraction benchmarks and achieves the highest average F1 for event triggers and arguments with gold triggers. The same model also performs strongly on aspect sentiment triplet extraction and pronoun resolution through its span-pair formulation. A correction for the probability bias introduced during training improves confidence estimates for entities, relations, and event arguments after calibration, with little change in F1.

cs.CL↗