arXiv subjects

Xing Wu

Publications and source records attributed to Xing Wu.

At least 37 records · Page 2Linked to original sources

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

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 $1σ$. 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

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

TSPO: Breaking the Double Homogenization Dilemma in Multi-turn Search Policy Optimization

Multi-turn tool-integrated reasoning enables Large Language Models (LLMs) to solve complex tasks through iterative information retrieval. However, current reinforcement learning (RL) frameworks for search-augmented reasoning predominantly rely on sparse outcome-level rewards, leading to a "Double Homogenization Dilemma." This manifests as (1) Process homogenization, where the thinking, reasoning, and tooling involved in generation are ignored. (2) Intra-group homogenization, coarse-grained outcome rewards often lead to inefficiencies in intra-group advantage estimation with methods like Group Relative Policy Optimization (GRPO) during sampling. To address this, we propose Turn-level Stage-aware Policy Optimization (TSPO). TSPO introduces the First-Occurrence Latent Reward (FOLR) mechanism, allocating partial rewards to the step where the ground-truth answer first appears, thereby preserving process-level signals and increasing reward variance within groups without requiring external reward models or any annotations. Extensive experiments demonstrate that TSPO significantly outperforms state-of-the-art baselines, achieving average performance gains of 24% and 13.6% on Qwen2.5-3B and 7B models, respectively. Code is available at https://github.com/Flipped-May/TSPO.

cs.AI↗

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

A cryogenic buffer gas beam source with in-situ ablation target replacement

The design and performance of a cryogenic buffer gas beam (CBGB) source with a load-lock system is presented. The ACME III electron electric dipole moment (eEDM) search experiment uses this source to produce a beam of cold, slow thorium monoxide (ThO) molecules. A key feature of the apparatus is its capability to replace ablation targets without interrupting vacuum or cryogenic conditions, increasing the average signal in the eEDM search. The source produces approximately $1.3 \times 10^{11}$ ground-state ThO molecules per pulse, with a rotational temperature of $4.8$ K, molecular beam solid angle of $0.31$ sr, and forward velocity of $200$ m/s. These parameters match the performance of traditional sources that require time-consuming thermal cycles for target replacement. A long-term yield improvement of about 40% is achieved when the load-lock system is used to replace targets biweekly.

physics.atom-ph↗

arXiv2026

MiLe Loss: a New Entropy-Weighed Loss for Mitigating the Bias of Learning Difficulties in Large Language Models

Generative language models are usually pretrained on large text corpus via predicting the next token (i.e., sub-word/word/phrase) given the previous ones. Recent works have demonstrated the impressive performance of large generative language models on downstream tasks. However, existing generative language models generally neglect an inherent challenge in text corpus during training, i.e., the imbalance between frequent tokens and infrequent ones. It can lead a language model to be dominated by common and easy-to-learn tokens, thereby overlooking the infrequent and difficult-to-learn ones. To alleviate that, we propose a MiLe Loss function for mitigating the bias of learning difficulties with tokens. During training, it can dynamically assess the learning difficulty of a to-be-learned token, according to the information entropy of the corresponding predicted probability distribution over the vocabulary. Then it scales the training loss adaptively, trying to lead the model to focus more on the difficult-to-learn tokens. On the Pile dataset, we train generative language models at different scales of 468M, 1.2B, and 6.7B parameters. Experiments reveal that models incorporating the proposed MiLe Loss can gain consistent performance improvement on downstream benchmarks.

cs.CL↗

arXiv2026

Put the Space of LoRA Initialization to the Extreme to Preserve Pre-trained Knowledge

Low-Rank Adaptation (LoRA) is the leading parameter-efficient fine-tuning method for Large Language Models (LLMs), but it still suffers from catastrophic forgetting. Recent work has shown that specialized LoRA initialization can alleviate catastrophic forgetting. There are currently two approaches to LoRA initialization aimed at preventing knowledge forgetting during fine-tuning: (1) making residual weights close to pre-trained weights, and (2) ensuring the space of LoRA initialization is orthogonal to pre-trained knowledge. The former is what current methods strive to achieve, while the importance of the latter is not sufficiently recognized. We find that the space of LoRA initialization is the key to preserving pre-trained knowledge rather than the residual weights. Existing methods like MiLoRA propose making the LoRA initialization space orthogonal to pre-trained weights. However, MiLoRA utilizes the null space of pre-trained weights. Compared to pre-trained weights, the input activations of pre-trained knowledge take into account the parameters of all previous layers as well as the input data, while pre-trained weights only contain information from the current layer. Moreover, we find that the effective ranks of input activations are much smaller than those of pre-trained weights. Thus, the null space of activations is more accurate and contains less pre-trained knowledge information compared to that of weights. Based on these, we introduce LoRA-Null, our proposed method that initializes LoRA in the null space of activations. Experimental results show that LoRA-Null effectively preserves the pre-trained world knowledge of LLMs while achieving good fine-tuning performance, as evidenced by extensive experiments. Code is available at {https://github.com/HungerPWAY/LoRA-Null}.

cs.CL↗

arXiv2026

LongBench Pro: A More Realistic and Comprehensive Bilingual Long-Context Evaluation Benchmark

The rapid expansion of context length in large language models (LLMs) has outpaced existing evaluation benchmarks. Current long-context benchmarks often trade off scalability and realism: synthetic tasks underrepresent real-world complexity, while fully manual annotation is costly to scale to extreme lengths and diverse scenarios. We present LongBench Pro, a more realistic and comprehensive bilingual benchmark of 1,500 naturally occurring long-context samples in English and Chinese spanning 11 primary tasks and 25 secondary tasks, with input lengths from 8k to 256k tokens. LongBench Pro supports fine-grained analysis with task-specific metrics and a multi-dimensional taxonomy of context requirement (full vs. partial dependency), length (six levels), and difficulty (four levels calibrated by model performance). To balance quality with scalability, we propose a Human-Model Collaborative Construction pipeline: frontier LLMs draft challenging questions and reference answers, along with design rationales and solution processes, to reduce the cost of expert verification. Experts then rigorously validate correctness and refine problematic cases. Evaluating 46 widely used long-context LLMs on LongBench Pro yields three findings: (1) long-context optimization contributes more to long-context comprehension than parameter scaling; (2) effective context length is typically shorter than the claimed context length, with pronounced cross-lingual misalignment; and (3) the "thinking" paradigm helps primarily models trained with native reasoning, while mixed-thinking designs offer a promising Pareto trade-off. In summary, LongBench Pro provides a robust testbed for advancing long-context understanding.

cs.CL↗

arXiv2025

EntropyLong: Effective Long-Context Training via Predictive Uncertainty

Training long-context language models to capture long-range dependencies requires specialized data construction. Current approaches, such as generic text concatenation or heuristic-based variants, frequently fail to guarantee genuine long-range dependencies. We propose EntropyLong, a novel data construction method that leverages predictive uncertainty to verify dependency quality. Our approach identifies high-entropy positions in documents, retrieves semantically relevant contexts from large corpora, and verifies their utility by assessing whether they reduce prediction entropy. This model-in-the-loop verification ensures each dependency represents measurable information gain rather than spurious correlation. We construct training samples with long-range dependencies by combining original documents with these verified contextual supplements. Using FineWebEdu and Cosmopedia, we generate a dataset of 128K-length sequences with verified dependencies. Models trained on this data demonstrate significant improvements on RULER benchmarks, particularly in tasks requiring distant information. Following instruction fine-tuning, our models also achieve substantial gains on LongBenchv2, demonstrating enhanced long-context understanding. Extensive ablation studies further validate the necessity and effectiveness of entropybased verification for long-context training.

cs.CL↗

arXiv2025

LiteLong: Resource-Efficient Long-Context Data Synthesis for LLMs

High-quality long-context data is essential for training large language models (LLMs) capable of processing extensive documents, yet existing synthesis approaches using relevance-based aggregation face challenges of computational efficiency. We present LiteLong, a resource-efficient method for synthesizing long-context data through structured topic organization and multi-agent debate. Our approach leverages the BISAC book classification system to provide a comprehensive hierarchical topic organization, and then employs a debate mechanism with multiple LLMs to generate diverse, high-quality topics within this structure. For each topic, we use lightweight BM25 retrieval to obtain relevant documents and concatenate them into 128K-token training samples. Experiments on HELMET and Ruler benchmarks demonstrate that LiteLong achieves competitive long-context performance and can seamlessly integrate with other long-dependency enhancement methods. LiteLong makes high-quality long-context data synthesis more accessible by reducing both computational and data engineering costs, facilitating further research in long-context language training.

cs.CL↗

arXiv2025

Libra: Large Chinese-based Safeguard for AI Content

Large language models (LLMs) excel in text understanding and generation but raise significant safety and ethical concerns in high-stakes applications. To mitigate these risks, we present Libra-Guard, a cutting-edge safeguard system designed to enhance the safety of Chinese-based LLMs. Leveraging a two-stage curriculum training pipeline, Libra-Guard enhances data efficiency by employing guard pretraining on synthetic samples, followed by fine-tuning on high-quality, real-world data, thereby significantly reducing reliance on manual annotations. To enable rigorous safety evaluations, we also introduce Libra-Test, the first benchmark specifically designed to evaluate the effectiveness of safeguard systems for Chinese content. It covers seven critical harm scenarios and includes over 5,700 samples annotated by domain experts. Experiments show that Libra-Guard achieves 86.79% accuracy, outperforming Qwen2.5-14B-Instruct (74.33%) and ShieldLM-Qwen-14B-Chat (65.69%), and nearing closed-source models like Claude-3.5-Sonnet and GPT-4o. These contributions establish a robust framework for advancing the safety governance of Chinese LLMs and represent a tentative step toward developing safer, more reliable Chinese AI systems.

cs.AI↗

arXiv2025

AIMatDesign: Knowledge-Augmented Reinforcement Learning for Inverse Materials Design under Data Scarcity

With the growing demand for novel materials, machine learning-driven inverse design methods face significant challenges in reconciling the high-dimensional materials composition space with limited experimental data. Existing approaches suffer from two major limitations: (I) machine learning models often lack reliability in high-dimensional spaces, leading to prediction biases during the design process; (II) these models fail to effectively incorporate domain expert knowledge, limiting their capacity to support knowledge-guided inverse design. To address these challenges, we introduce AIMatDesign, a reinforcement learning framework that addresses these limitations by augmenting experimental data using difference-based algorithms to build a trusted experience pool, accelerating model convergence. To enhance model reliability, an automated refinement strategy guided by large language models (LLMs) dynamically corrects prediction inconsistencies, reinforcing alignment between reward signals and state value functions. Additionally, a knowledge-based reward function leverages expert domain rules to improve stability and efficiency during training. Our experiments demonstrate that AIMatDesign significantly surpasses traditional machine learning and reinforcement learning methods in discovery efficiency, convergence speed, and success rates. Among the numerous candidates proposed by AIMatDesign, experimental synthesis of representative Zr-based alloys yielded a top-performing BMG with 1.7GPa yield strength and 10.2\% elongation, closely matching predictions. Moreover, the framework accurately captured the trend of yield strength variation with composition, demonstrating its reliability and potential for closed-loop materials discovery.

cs.LG↗

arXiv2025

dots.llm1 Technical Report

Mixture of Experts (MoE) models have emerged as a promising paradigm for scaling language models efficiently by activating only a subset of parameters for each input token. In this report, we present dots.llm1, a large-scale MoE model that activates 14B parameters out of a total of 142B parameters, delivering performance on par with state-of-the-art models while reducing training and inference costs. Leveraging our meticulously crafted and efficient data processing pipeline, dots.llm1 achieves performance comparable to Qwen2.5-72B after pretraining on 11.2T high-quality tokens and post-training to fully unlock its capabilities. Notably, no synthetic data is used during pretraining. To foster further research, we open-source intermediate training checkpoints at every one trillion tokens, providing valuable insights into the learning dynamics of large language models.

cs.CL↗

arXiv2025

LongMagpie: A Self-synthesis Method for Generating Large-scale Long-context Instructions

High-quality long-context instruction data is essential for aligning long-context large language models (LLMs). Despite the public release of models like Qwen and Llama, their long-context instruction data remains proprietary. Human annotation is costly and challenging, while template-based synthesis methods limit scale, diversity, and quality. We introduce LongMagpie, a self-synthesis framework that automatically generates large-scale long-context instruction data. Our key insight is that aligned long-context LLMs, when presented with a document followed by special tokens preceding a user turn, auto-regressively generate contextually relevant queries. By harvesting these document-query pairs and the model's responses, LongMagpie produces high-quality instructions without human effort. Experiments on HELMET, RULER, and Longbench v2 demonstrate that LongMagpie achieves leading performance on long-context tasks while maintaining competitive performance on short-context tasks, establishing it as a simple and effective approach for open, diverse, and scalable long-context instruction data synthesis.

cs.CL↗

arXiv2025

FMNet: Frequency-Assisted Mamba-Like Linear Attention Network for Camouflaged Object Detection

Camouflaged Object Detection (COD) is challenging due to the strong similarity between camouflaged objects and their surroundings, which complicates identification. Existing methods mainly rely on spatial local features, failing to capture global information, while Transformers increase computational costs. To address this, the Frequency-Assisted Mamba-Like Linear Attention Network (FMNet) is proposed, which leverages frequency-domain learning to efficiently capture global features and mitigate ambiguity between objects and the background. FMNet introduces the Multi-Scale Frequency-Assisted Mamba-Like Linear Attention (MFM) module, integrating frequency and spatial features through a multi-scale structure to handle scale variations while reducing computational complexity. Additionally, the Pyramidal Frequency Attention Extraction (PFAE) module and the Frequency Reverse Decoder (FRD) enhance semantics and reconstruct features. Experimental results demonstrate that FMNet outperforms existing methods on multiple COD datasets, showcasing its advantages in both performance and efficiency. Code available at https://github.com/Chranos/FMNet.

cs.CV↗

arXiv2025

CodePMP: Scalable Preference Model Pretraining for Large Language Model Reasoning

Large language models (LLMs) have made significant progress in natural language understanding and generation, driven by scalable pretraining and advanced finetuning. However, enhancing reasoning abilities in LLMs, particularly via reinforcement learning from human feedback (RLHF), remains challenging due to the scarcity of high-quality preference data, which is labor-intensive to annotate and crucial for reward model (RM) finetuning. To alleviate this issue, we introduce CodePMP, a scalable preference model pretraining (PMP) pipeline that utilizes a large corpus of synthesized code-preference pairs from publicly available high-quality source code. CodePMP improves RM finetuning efficiency by pretraining preference models on large-scale synthesized code-preference pairs. We evaluate CodePMP on mathematical reasoning tasks (GSM8K, MATH) and logical reasoning tasks (ReClor, LogiQA2.0), consistently showing significant improvements in reasoning performance of LLMs and highlighting the importance of scalable preference model pretraining for efficient reward modeling.

cs.AI↗