arXiv subjects

Xing Wu

Publications and source records attributed to Xing Wu.

At least 19 recordsLinked to original sources

arXiv2026

Environment Evolution for Terminal Agents

Scaling interactive and verifiable environments is critical for training terminal agents. As frontier models become more capable, environments synthesized from scratch become less challenging and thus provide limited learning signals. Recent co-evolution methods iteratively synthesize environments near the model's learnable frontier based on weaknesses exposed during rollouts. However, their dependence on on-policy rollouts limits generalization and the continuous provision of learning signals as the model becomes stronger. In this paper, we propose environment evolution, which incrementally increases environment difficulty off-policy and schedules the evolved environments generation by generation during training to provide continuous learning signals. We derive three evolution directions that influence environment difficulty from the multi-turn learning objective and then implement evolution along these directions through a loop-engineered multi-agent harness. Quantitative rollout experiments with Hy4 preview, Claude Opus 5, and GPT-5.6 Sol show that environment evolution consistently produces more difficult environments. We validate its effectiveness on Qwen3.6-27B and Qwen3.6-35B-A3B through simple long-horizon RL training, improving their performance by 14.4 and 18.0 percentage points on Terminal-Bench 2.1, respectively.

cs.AI↗

arXiv2026

LongGuard: Mechanistic Analysis and Training-Free Mitigation of Long-Context Failure in Safety Guardrails

Safety guardrails serve as the last line of defense against harmful inputs and outputs of large language models (LLMs), yet they are trained and evaluated almost exclusively on short text. We present LongGuard, a framework that evaluates, mechanistically analyzes, and mitigates long-context guardrail failure. We formulate the task as Safety Needle-in-a-Haystack (SafetyNIAH) over a 0.25k-32k length grid; across 15 mainstream guardrails, unsafe recall drops monotonically by more than 50% on average, and a paired Benign-Fill vs. Needle-Repeat design attributes the failure to proportional dilution of the unsafe needle rather than to absolute length. A three-layer attention-logit-behavior analysis on six guardrails locates the mechanism: attention mass on the unsafe needle is diluted, the unsafe-over-safe logit margin is compressed in lockstep, and the detection decision collapses accordingly, with this attention->logit->behavior chain remaining consistent after partialling out length. We further isolate a sparse set of guard-specialized retrieval heads that exhibit partial specificity relative to their base models. Building on the analysis, we propose two training-free mitigations - Chunked Detection (CD) and Attention-Head Sharpening (AHS) - and a deployment protocol, Context-Aware Hyperparameter Routing (CAHR), that selects configurations by context length and audit side. Across five benchmarks spanning synthetic data, long-context attacks, and reasoning-model outputs, CAHR-CD and CAHR-AHS improve the six-guardrail average by 22% and 13%, respectively. Code and data are available online.

cs.AI↗

arXiv2026

HealMed: Multilingual Evaluation of Large Language Models in Medicine

We present HealMed, an expert-reviewed benchmark for multilingual evaluation of large language models in medicine. HealMed contains 1,000 examples in each of nine languages, drawn from nine datasets and covering three task formats: MCQA, NLI and open-ended QA. The benchmark was developed over two years by 23 physicians and medical experts based across nine countries and regions. Each translation was evaluated and revised by two experts fluent in English and the corresponding target language. On HealMed, performance declined most in low-resource languages, although the size of the gap varied markedly across languages and models. The strongest proprietary models were the most stable across languages, whereas many open-source and medically specialized models showed larger and less consistent gaps. Medical specialization alone did not ensure multilingual robustness. Furthermore, expert revision could either raise or lower measured performance, indicating that translation quality materially affects cross-language evaluation results.

cs.CL↗

arXiv2026

Deep Research Pretraining via Predictive Navigation

Deep research agents are often trained on expensive, environment-grounded tool-use trajectories that require repeated retrieval, document inspection, and report evaluation. We introduce Deep Research Pretraining (DRP), an offline framework that derives predictive navigation supervision from naturally occurring evidence structures. Given a citation-bearing or hyperlinked passage, DRP constructs a proxy research objective, recovers linked evidence and graph-related alternatives, and converts them into search-open-write trajectories. This teaches models what to search for, which documents to inspect, and how to synthesize evidence, without a live retrieval environment or executed policy rollout. We instantiate DRP on scholarly citation graphs (DRP-Paper) and Wikipedia hyperlinks (DRP-Web), continually pretrain separate Qwen3-14B-Base models on 1B tokens, and fine-tune them on controlled fractions of 13K agent trajectories. Across five independently sampled subsets at each low-data budget, both variants consistently outperform matched no-DRP models on DeepResearch Bench. With one quarter of the SFT data, DRP-Web even surpasses a fixed no-DRP full-data checkpoint, with gains transferring to ResearchQA, WebWalkerQA, and SimpleQA. Starting from matched low-data SFT checkpoints, the DRP-Web advantage also persists through subsequent agentic RL. Source-matched and evidence-mismatch controls indicate that these improvements arise from evidence-conditioned navigation rather than domain exposure or agent-format imitation. DRP thus provides a promising complementary approach to trajectory-based agent training.

cs.CL↗

arXiv2026

Demystifying DRAM Read Disturbance: Bridging the Gap Between Experimental Characterization and Device-Level Modeling of RowHammer and RowPress Phenomena

DRAM read disturbance, like RowHammer and RowPress, is a critical robustness issue where accessing DRAM can cause unintended bitflips in other unaccessed DRAM locations. DRAM read disturbance bitflips significantly impact the safe, secure, and reliable operation of DRAM-based computing systems. Many prior works experimentally characterize these bitflips and propose mitigations based on empirical results. Other device-level works study their underlying physical mechanisms, but these mechanisms do not fully explain all major empirical observations. Our goal is to bridge the gap between experimental characterization and device-level modeling and understanding of RowHammer and RowPress, providing a principled foundation for future work on understanding, characterizing, and mitigating DRAM read disturbance. We first identify and demonstrate gaps and inconsistencies between the physical mechanisms of RowHammer and RowPress described by existing device-level models and experimental characterization of their bitflips. We focus on three fundamental metrics that should map to first-order physical mechanisms: 1) bitflip directions, 2) bitflip counts, and 3) the minimum number of aggressor row activations that trigger the first bitflips (i.e., ACmin). Second, we present a comprehensive and rigorous set of TCAD simulations that match phenomena observed in experimental characterizations of RowHammer and RowPress bitflips. From our results, we 1) summarize updated device-level error mechanisms for understanding RowHammer and RowPress bitflips, and 2) identify key modeling and simulation parameters that significantly affect whether simulation results match real-chip characterization. We discuss implications for 1) rigorous, comprehensive, and efficient experimental characterization methodologies of DRAM read disturbance bitflips, and 2) the design of DRAM read disturbance mitigation techniques.

cs.AR↗

arXiv2026

Future directions in nuclear $\beta$ decay at FRIB and beyond

Motivated by the opportunities presented for studies relevant to nuclear structure, astrophysics, and fundamental symmetries with nuclear $\beta$ decay, the Facility for Rare Isotope Beams (FRIB) Theory Alliance topical program ``Future Directions in Nuclear $\beta$ Decays at FRIB'' was held in September of 2025. This white paper summarizes the main points of discussion over the two-week program, and it aims to provide a snapshot of the current status of the field while also highlighting important questions and opportunities for future work. We provide an overview of the experimental tools and techniques that enable modern $\beta$ decay studies, discuss the current state of nuclear many-body approaches used to study $\beta$ decays, and highlight the important science questions that can be addressed by weak decays.

nucl-th↗

arXiv2026

TC-MAF: Train-Calibrated Bounded Multi-Evidence Fusion for Multimodal Industrial Anomaly Detection

Multimodal anomaly detection benefits from complementary RGB and 3D evidence, yet auxiliary RGB reconstruction is not equally reliable across product categories and class-wise test-time policy selection is usually unavailable. We propose TC-MAF, a base-anchored multi-evidence fusion design that combines a multimodal detector, complementary Dinomaly evidence, and a small cross-modal consistency cue under one fixed pixel-level fusion formula. A lightweight training-dispersion confidence (TDC) term scales auxiliary participation using only normal training statistics. On MVTec-3D, TC-MAF reaches 0.979 image-level AUROC and 0.990 pixel-level AUPRO, achieving the best mean results on both detection and localization among the compared multimodal methods. Systematic ablations show that the fusion structure itself is the dominant factor, while TDC provides a smaller but reproducible calibration gain over no calibration or arbitrary calibration. Additional experiments show that the same design remains effective under a pooled-statistics variant, auxiliary-branch and backbone substitutions, few-shot settings, a missing-3D setting, and cross-dataset evaluation on Eyecandies. Code is available at https://anonymous.4open.science/r/TC_MAF-C3BB.

cs.CV↗

arXiv2026

A practical methodology for $\Lambda$ global polarization extraction in fixed-target experiments

Non-central heavy-ion collisions generate large orbital angular momentum in the created medium, which leads to polarization of final-state particles via spin-orbit coupling, known as global spin polarization. The observation of significant global polarization of $\Lambda$ hyperon in heavy-ion collisions indicates that the quark-gluon plasma is the most vortical fluid known in nature. Exploring $\Lambda$ global polarization at lower energies is important for understanding spin dynamics across different regions of the quantum chromodynamics (QCD) phase diagram. Low-energy nuclear experiments are typically conducted with asymmetric detector acceptance, as in fixed-target collisions at RHIC-STAR, and at facilities such as FAIR, NICA, HIAF and HIRFL-CSR. The asymmetric rapidity coverage in these experiments enhances the coupling between directed flow and detector inefficiencies, creating significant bias in $\Lambda$ global polarization measurements. In this paper, we propose a methodology to eliminate such bias arising from asymmetric detector acceptance. The method is validated using realistic detector simulations based on the STAR fixed-target configuration.

physics.data-an↗

arXiv2026

From Blind Guess to Informed Judgment: Teaching LLMs to Evaluate Materials by Building Knowledge-Augmented Preference Signals

As candidate generation and high-throughput experimentation advance, the primary bottleneck in materials discovery is shifting from property prediction to making reliable evaluations among massive candidate sets. We propose a Knowledge-Augmented Preference Signals Framework, MaterEval, that automatically produces, for the same candidate, two evaluations: an informed judgment that follows expert rules and provides supporting evidence, and a rule-removed blind guess. By pairing the two evaluations as preference data, we guide general-purpose large language models (LLMs), originally lacking materials-specific criteria, from intuitive judgment toward reliable evaluation supported by explicit evidence. To balance throughput, cost, and reliability, we further introduce a fast-slow reasoning scheme that decouples large-scale rapid screening from in-depth review on a small subset. Using high-entropy alloy (HEA) assessment as a case study, we show that, without external retrieval and relying solely on internalized capabilities, small open-source LLMs achieve substantial gains in accuracy, conclusion consistency, and evidence discrimination, approaching the performance of rule-based closed-source LLMs. These results demonstrate that expert rules can be systematically transformed into learnable preference signals, enabling a low-cost and deployable evaluation module for autonomous materials discovery loops.

cs.CL↗

arXiv2026

Optical cycling on thorium monoxide (ThO) for an improved test of fundamental symmetries

Optical cycling refers to repeated excitation and spontaneous emission on an electronic transition in an atom or molecule. Optical cycling in molecules can enable a wide range of quantum control and readout techniques, but unfortunately it has only been demonstrated on a small class of alkali-like or alkaline-earth-like molecules. Thorium monoxide (ThO), a molecule used in one of the most precise permanent electron electric dipole moment (eEDM) searches (ACME [1]), does not fall into this category. In this work, we demonstrate the first optical cycling on this non-conventional class over a range of experimental parameter space, including laser intensity, polarization switching rate, and interaction time. We show that both the $J = 1, 2$ rotational levels of ThO molecule are capable of cycling $11(2)$ photons on average with a single laser, at $1.9(6) \times 10^{6}~\mathrm{s}^{-1}$ and $2.3(7)\times 10^6~\mathrm{s}^{-1}$ scattering rate, respectively, before population is lost to other vibronic levels. We outline a scheme to apply this demonstrated optical cycling in an ACME-style eEDM measurement, improving the detection efficiency by over fourfold compared to non-cycling fluorescence detection. This would lead to over a twofold enhancement in the statistical sensitivity of the eEDM search. This optical cycling scheme can be further extended to scatter about 100 photons, which would enable a wider range of quantum control and sensing using ThO molecules. [1] V. Andreev, D. G. Ang, D. DeMille, J. M. Doyle, G. Gabrielse, J. Haefner, N. R. Hutzler, Z. Lasner, C. Meisenhelder, B. R. O`Leary, C. D. Panda, A. D. West, E. P. West, and X. Wu, Nature 562, 355 (2018).

physics.atom-ph↗

arXiv2026

Toward Scalable Terminal Task Synthesis via Skill Graphs

Terminal agents have demonstrated strong potential for autonomous command-line execution, yet their training remains constrained by the scarcity of high-quality and diverse execution trajectories. Existing approaches mitigate this bottleneck by synthesizing large-scale terminal task instances for trajectory sampling. However, they primarily focus on scaling the number of tasks while providing limited control over the diversity of execution trajectories that agents actually experience during training. In this paper, we present SkillSynth, an automated framework for terminal task synthesis built on a scenario-mediated skill graph. SkillSynth first constructs a large-scale skill graph, where scenarios serve as intermediate transition nodes that connect diverse command-line skills. It then samples paths from this graph as abstractions of real-world workflows, and uses a multi-agent harness to instantiate them into executable task instances. By grounding task synthesis in graph-sampled workflow paths, SkillSynth explicitly controls the diversity of minimal execution trajectories required to solve the synthesized tasks. Experiments on Terminal-Bench demonstrate the effectiveness of SkillSynth. Moreover, task instances synthesized by SkillSynth have been adopted to train Hy3 Preview, contributing to its enhanced agentic capabilities in terminal-based settings.

cs.AI↗

arXiv2026

PolicyLong: Towards On-Policy Context Extension

Extending LLM context windows is hindered by scarce high-quality long-context data. Recent methods synthesize data with genuine long-range dependencies via information-theoretic verification, selecting contexts that reduce a base model's predictive entropy. However, their single-pass offline construction with a fixed model creates a fundamental off-policy gap: the static screening landscape misaligns with the model's evolving capabilities, causing the training distribution to drift. We propose PolicyLong, shifting data construction towards a dynamic on-policy paradigm. By iteratively re-executing data screening (entropy computation, retrieval, and verification) using the current model, PolicyLong ensures the training distribution tracks evolving capabilities, yielding an emergent self-curriculum. Crucially, both positive and hard negative contexts derive from the current model's entropy landscape, co-evolving what the model learns to exploit and resist. Experiments on RULER, HELMET, and LongBench-v2 (Qwen2.5-3B) show PolicyLong consistently outperforms EntropyLong and NExtLong, with gains growing at longer contexts (e.g., +2.54 at 128K on RULER), confirming the value of on-policy data evolution.

cs.LG↗

arXiv2026

TOOLCAD: Exploring Tool-Using Large Language Models in Text-to-CAD Generation with Reinforcement Learning

Computer-Aided Design (CAD) is an expert-level task that relies on long-horizon reasoning and coherent modeling actions. Large Language Models (LLMs) have shown remarkable advancements in enabling language agents to tackle real-world tasks. Notably, there has been no investigation into how tool-using LLMs optimally interact with CAD engines, hindering the emergence of LLM-based agentic text-to-CAD modeling systems. We propose ToolCAD, a novel agentic CAD framework deploying LLMs as tool-using agents for text-to-CAD generation. Furthermore, we introduce an interactive CAD modeling gym to rollout reasoning and tool-augmented interaction trajectories with the CAD engine, incorporating hybrid feedback and human supervision. Meanwhile, an end-to-end post-training strategy is presented to enable the LLM agent to elicit refined CAD Modeling Chain of Thought (CAD-CoT) and evolve into proficient CAD tool-using agents via online curriculum reinforcement learning. Our findings demonstrate ToolCAD fills the gap in adopting and training open-source LLMs for CAD tool-using agents, enabling them to perform comparably to proprietary models, paving the way for more accessible and robust autonomous text-to-CAD modeling systems.

cs.CV↗

arXiv2026

WRAP++: Web discoveRy Amplified Pretraining

Synthetic data rephrasing has emerged as a powerful technique for enhancing knowledge acquisition during large language model (LLM) pretraining. However, existing approaches operate at the single-document level, rewriting individual web pages in isolation. This confines synthesized examples to intra-document knowledge, missing cross-document relationships and leaving facts with limited associative context. We propose WRAP++ (Web discoveRy Amplified Pretraining), which amplifies the associative context of factual knowledge by discovering cross-document relationships from web hyperlinks and synthesizing joint QA over each discovered document pair. Concretely, WRAP++ discovers high-confidence relational motifs including dual-links and co-mentions, and synthesizes QA that requires reasoning across both documents. This produces relational knowledge absent from either source document alone, creating diverse entry points to the same facts. Because the number of valid entity pairs grows combinatorially, this discovery-driven synthesis also amplifies data scale far beyond single-document rewriting. Instantiating WRAP++ on Wikipedia, we amplify ~8.4B tokens of raw text into 80B tokens of cross-document QA data. On SimpleQA, OLMo-based models at both 7B and 32B scales trained with WRAP++ substantially outperform single-document approaches and exhibit sustained scaling gains, underscoring the advantage of cross-document knowledge discovery and amplification.

cs.CL↗

arXiv2026

Tianwen-2 target asteroid (469219) Kamo'oalewa probably develops an Itokawa-compositional but ultra-highly space-weathered surface

China's Tianwen-2 mission plans to return samples from a small, rapidly spinning Earth quasi-satellite (469219) Kamo'oalewa. Previous studies linked Kamo'oalewa to lunar composition and origin. Here, we propose another scenario. We reanalyzed the reflectance spectrum of Kamo'oalewa and obtained an absorption band center at 1.001+-0.028 um (error is 1sigma), consistent with LL chondrites. We then conducted space weathering (SW) experiments on meteorites and found that highly space-weathered LL chondrite powder (but not slab) successfully reproduced the reflectance spectrum of Kamo'oalewa. We further traced the dynamical origin of Kamo'oalewa and found that it probably originated from the v6 secular resonance, and more specifically, the Flora family. Kamo'oalewa exhibits a similar composition to Itokawa and 7 objects in the Flora family, but with a higher degree of space weathering. We, therefore, proposed that Kamo'oalewa probably originated from the Flora family and developed an Itokawa-compositional, highly space-weathered, fine-regolith-dominated surface.

astro-ph.EP↗

arXiv2026

Shape, regolith size and thickness, SMFe^0 content, and spectral type of Tianwen-2 target asteroid (469219) Kamo'oalewa

China's Tianwen-2 spacecraft will return samples from the near-Earth asteroid (469219) Kamo'oalewa. We previously reported that Kamo'oalewa develops an LL-chondrite-compositional, highly space-weathered surface. This study aims to estimate Kamo'oalewa's shape, regolith grain size and thickness, sub-micrometer iron (SMFe0) content, and spectral type. Using the lightcurve data and the Cellinoid model, we modeled Kamo'oalewa's shape, rotation period, and pole orientation. We then estimated its global distribution of regolith critical size using the balance method of gravity, cohesive force, and centrifugal force. Furthermore, in the temperature range of 253.15 to 473.15 K, we measured the thermal parameters of laser-irradiated LL chondrite powder that best matches Kamo'oalewa's spectrum, estimating Kamo'oalewa's thermal inertia and skin depth (lower limit of regolith thickness). Using the radiative transfer mixing model, we also estimated the content of SMFe0 in Kamo'oalewa's regolith. Finally, using the MIT online spectral classification tool for the laser-irradiated LL chondrite powder, we obtained a virtual spectral type of Kamo'oalewa. Our model gives a size of 68 m x 46 m x 39 m, a rotation period of 27.66 minutes, and a pole orientation of 134.7 degrees longitude and -11.4 degrees latitude for Kamo'oalewa. Regolith grains with a size <2 cm can remain stable over 93.8% of the global surface area of Kamo'oalewa. Laser-irradiated LL chondrite powder shows a low thermal inertia (95.5 to 135.1 J m^-2 K^-1 s^-1/2), corresponding to a thermal skin depth of 3 to 3.5 mm on Kamo'oalewa. An SMFe0 content of 0.29 +- 0.05 wt.% is required to match Kamo'oalewa's spectrum. The virtual spectral type of Kamo'oalewa is given as "Sqw".

astro-ph.EP↗

arXiv2026

Model-independent test of the cosmic distance duality relation with recent observational data

We test the cosmic distance duality relation (CDDR) using two model-independent methods. Method I is based on the PAge parametrization, which characterizes the expansion history in terms of the cosmic age. Parametrizations of possible CDDR violations are constrained using observational data from Type Ia supernovae (SN), baryon acoustic oscillations (BAO), cosmic chronometers, and gamma-ray bursts (GRB), including the latest PantheonPlus and DES Dovekie SN samples and DESI DR2 BAO data. The results support the validity of the CDDR within $1\sigma$. Different combinations of data sets are further explored to assess the impact of various probes and calibration choices, demonstrating the robustness of this conclusion. Although GRB data extend to higher redshifts, their constraining power is significantly weaker than that of the other low-redshift probes. The PantheonPlus and DES Dovekie samples yield consistent results. Method II uses a non-parametric Gaussian process reconstruction of the luminosity distance from SN data, combined with BAO measurements to construct the observed CDDR violation and constrain its parametrizations. The results are consistent with those from Method I, and we find no evidence for a violation of the CDDR.

astro-ph.CO↗

arXiv2026

VideoTIR: Accurate Understanding for Long Videos with Efficient Tool-Integrated Reasoning

Existing Multimodal Large Language Models (MLLMs) often suffer from hallucinations in long video understanding (LVU), primarily due to the imbalance between textual and visual tokens. Observing that MLLMs handle short visual inputs well, recent LVU works alleviate hallucinations by automatically parsing the vast visual data into manageable segments that can be effectively processed by MLLMs. SFT-based tool-calling methods can serve this purpose, but they typically require vast amounts of fine-grained, high-quality data and suffer from constrained tool-calling trajectories. We propose a novel VideoTIR that leverages Reinforcement Learning (RL) to encourage proper usage of comprehensive multi-level toolkits for efficient long video understanding. VideoTIR explores both Zero-RL and SFT cold-starting to enable MLLMs to retrieve and focus on meaningful video segments/images/regions, enhancing long video understanding both accurately and efficiently. To reduce redundant tool-calling, we propose Toolkit Action Grouped Policy Optimization (TAGPO), which enhances the efficiency of the calling process through stepwise reward assignment and reuse of failed rollouts. Additionally, we develop a sandbox-based trajectory synthesis framework to generate high-quality trajectories data. Extensive experiments on three long-video QA benchmarks demonstrate the effectiveness and efficiency of our method.

cs.CV↗