BZPEERReading, trust and collaboration for research
Menu

Discover

GraphFollow connections around a paper, topic or saved view.

Workspaces

HomePick up where your research flow left off.InboxHandle requests, replies and notifications from one place.CollectionsCurate reading lists, evidence sets and shareable dossiers.ResearchSee your private provenance graph, trail and imports.CollaborationMove from discovery into focused discussion rooms.PublishDraft, preview and publish full-text research articles.

Network

ExpertsFind specialists worth following or asking for help.CommunitiesJoin research circles organized around shared work.PartnersSee external organizations active around the same topics.

Opportunities

ListingsBrowse roles, calls and collaborations with clearer fit signals.

Account

Log inReturn to your reading and collaboration workspace.Create accountStart saving work, following people and joining teams.
Log inCreate account

Researcher profile

Xing Wu

Xing Wu contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate
Computation and LanguageArtificial Intelligencemath.APphysics.atom-phApplicationscond-mat.mes-halleess.ASGenomicsgr-qchep-thMachine Learningphysics.chem-phphysics.ins-detquant-phSound

Trust snapshot

Quick read

Trust 21 - EmergingVerification L1Unclaimed author
19works
0followers
15topics
4close collaborators

Actions

Decide how to stay connected

Follow researcher0

Identity and collaboration

How to connect with this researcher

Claiming links this public author record to a researcher profile and unlocks direct collaboration workflows.

Log in to claim

Direct collaboration

Open a focused conversation when the fit is right

Claim this author entity first to unlock direct invitations.

Research graph

See the researcher in context

Open full explorer

Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.

Building this graph slice

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

19 published item(s)

preprint2026arXiv

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.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

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.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

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}.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

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.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

ESimCSE: Enhanced Sample Building Method for Contrastive Learning of Unsupervised Sentence Embedding

Contrastive learning has been attracting much attention for learning unsupervised sentence embeddings. The current state-of-the-art unsupervised method is the unsupervised SimCSE (unsup-SimCSE). Unsup-SimCSE takes dropout as a minimal data augmentation method, and passes the same input sentence to a pre-trained Transformer encoder (with dropout turned on) twice to obtain the two corresponding embeddings to build a positive pair. As the length information of a sentence will generally be encoded into the sentence embeddings due to the usage of position embedding in Transformer, each positive pair in unsup-SimCSE actually contains the same length information. And thus unsup-SimCSE trained with these positive pairs is probably biased, which would tend to consider that sentences of the same or similar length are more similar in semantics. Through statistical observations, we find that unsup-SimCSE does have such a problem. To alleviate it, we apply a simple repetition operation to modify the input sentence, and then pass the input sentence and its modified counterpart to the pre-trained Transformer encoder, respectively, to get the positive pair. Additionally, we draw inspiration from the community of computer vision and introduce a momentum contrast, enlarging the number of negative pairs without additional calculations. The proposed two modifications are applied on positive and negative pairs separately, and build a new sentence embedding method, termed Enhanced Unsup-SimCSE (ESimCSE). We evaluate the proposed ESimCSE on several benchmark datasets w.r.t the semantic text similarity (STS) task. Experimental results show that ESimCSE outperforms the state-of-the-art unsup-SimCSE by an average Spearman correlation of 2.02% on BERT-base.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

High-sensitivity low-noise photodetector using large-area silicon photomultiplier

The application of silicon photomultiplier (SiPM) technology for weak-light detection at a single photon level has expanded thanks to its better photon detection efficiency in comparison to a conventional photomultiplier tube (PMT). SiPMs with large detection area have recently become commercially available, enabling applications where the photon flux is low both temporarily and spatially. On the other hand, several drawbacks exist in the usage of SiPMs such as a higher dark count rate, many readout channels, slow response time, and optical crosstalk; therefore, users need to carefully consider the trade-offs. This work presents a SiPM-embedded compact large-area photon detection module. Various techniques are adopted to overcome the disadvantages of SiPMs so that it can be generally utilized as an upgrade from a PMT. A simple cooling component and recently developed optical crosstalk suppression method are adopted to reduce the noise which is more serious for larger-area SiPMs. A dedicated readout circuit increases the response frequency and reduces the number of readout channels. We favorably compare this design with a conventional PMT and obtain both higher photon detection efficiency and larger-area acceptance.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Ill_posedness for a two_component Novikov system in Besov space

In this paper, we consider the Cauchy problem for a two-component Novikov system on the line. By specially constructed initial data $(ρ_0, u_0)$ in $B_{p, \infty}^{s-1}(\mathbb{R})\times B_{p, \infty}^s(\mathbb{R})$ with $s>\max\{2+\frac{1}{p}, \frac{5}{2}\}$ and $1\leq p \leq \infty$, we show that any energy bounded solution starting from $(ρ_0, u_0)$ does not converge back to $(ρ_0, u_0)$ in the metric of $B_{p, \infty}^{s-1}(\mathbb{R})\times B_{p, \infty}^s(\mathbb{R})$ as time goes to zero, thus results in discontinuity of the data-to-solution map and ill-posedness.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Measurement of the H$^3Δ_1$ Radiative Lifetime in ThO

The best limit on the electron electric dipole moment (eEDM) comes from the ACME II experiment [Nature \textbf{562} (2018), 355-360] which probes physics beyond the Standard Model at energy scales well above 1 TeV. ACME II measured the eEDM by monitoring electron spin precession in a cold beam of the metastable H$^3Δ_1$ state of thorium monoxide (ThO) molecules, with an observation time $τ\approx 1$ ms for each molecule. We report here a new measurement of the lifetime of the ThO (H$^3Δ_1$) state, $τ_H = 4.2\pm 0.5$ ms. Using an apparatus within which $τ\approx τ_H$ will enable a substantial reduction in uncertainty of an eEDM measurement.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Smoothed Contrastive Learning for Unsupervised Sentence Embedding

Contrastive learning has been gradually applied to learn high-quality unsupervised sentence embedding. Among the previous un-supervised methods, the latest state-of-the-art method, as far as we know, is unsupervised SimCSE (unsup-SimCSE). Unsup-SimCSE uses the InfoNCE1loss function in the training stage by pulling semantically similar sentences together and pushing apart dis-similar ones.Theoretically, we expect to use larger batches in unsup-SimCSE to get more adequate comparisons among samples and avoid overfitting. However, increasing the batch size does not always lead to improvements, but instead even lead to performance degradation when the batch size exceeds a threshold. Through statistical observation, we find that this is probably due to the introduction of low-confidence negative pairs after in-creasing the batch size. To alleviate this problem, we introduce a simple smoothing strategy upon the InfoNCE loss function, termedGaussian Smoothing InfoNCE (GS-InfoNCE).Specifically, we add random Gaussian noise vectors as negative samples, which act asa smoothing of the negative sample space.Though being simple, the proposed smooth-ing strategy brings substantial improvements to unsup-SimCSE. We evaluate GS-InfoNCEon the standard semantic text similarity (STS)task. GS-InfoNCE outperforms the state-of-the-art unsup-SimCSE by an average Spear-man correlation of 1.38%, 0.72%, 1.17% and0.28% on the base of BERT-base, BERT-large,RoBERTa-base and RoBERTa-large, respectively.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Stacked Autoencoder Based Multi-Omics Data Integration for Cancer Survival Prediction

Cancer survival prediction is important for developing personalized treatments and inducing disease-causing mechanisms. Multi-omics data integration is attracting widespread interest in cancer research for providing information for understanding cancer progression at multiple genetic levels. Many works, however, are limited because of the high dimensionality and heterogeneity of multi-omics data. In this paper, we propose a novel method to integrate multi-omics data for cancer survival prediction, called Stacked AutoEncoder-based Survival Prediction Neural Network (SAEsurv-net). In the cancer survival prediction for TCGA cases, SAEsurv-net addresses the curse of dimensionality with a two-stage dimensionality reduction strategy and handles multi-omics heterogeneity with a stacked autoencoder model. The two-stage dimensionality reduction strategy achieves a balance between computation complexity and information exploiting. The stacked autoencoder model removes most heterogeneities such as data's type and size in the first group of autoencoders, and integrates multiple omics data in the second autoencoder. The experiments show that SAEsurv-net outperforms models based on a single type of data as well as other state-of-the-art methods.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Text Smoothing: Enhance Various Data Augmentation Methods on Text Classification Tasks

Before entering the neural network, a token is generally converted to the corresponding one-hot representation, which is a discrete distribution of the vocabulary. Smoothed representation is the probability of candidate tokens obtained from a pre-trained masked language model, which can be seen as a more informative substitution to the one-hot representation. We propose an efficient data augmentation method, termed text smoothing, by converting a sentence from its one-hot representation to a controllable smoothed representation. We evaluate text smoothing on different benchmarks in a low-resource regime. Experimental results show that text smoothing outperforms various mainstream data augmentation methods by a substantial margin. Moreover, text smoothing can be combined with those data augmentation methods to achieve better performance.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Buffer-gas cooling of molecules in the low-density regime: Comparison between simulation and experiment

Cryogenic buffer gas cells have been a workhorse for the cooling of molecules in the last decades. The straightforward sympathetic cooling principle makes them applicable to a huge variety of different species. Notwithstanding this success, detailed simulations of buffer gas cells are rare, and have never been compared to experimental data in the regime of low to intermediate buffer gas densities. Here, we present a numerical approach based on a trajectory analysis, with molecules performing a random walk in the cell due to collisions with a homogeneous buffer gas. This method can reproduce experimental flux and velocity distributions of molecules emerging from the buffer gas cell for varying buffer gas densities. This includes the strong decrease in molecule output from the cell for increasing buffer gas density and the so-called boosting effect, when molecules are accelerated by buffer-gas atoms after leaving the cell. The simulations provide various insights which could substantially improve buffer-gas cell design.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Data Augmentation for Copy-Mechanism in Dialogue State Tracking

While several state-of-the-art approaches to dialogue state tracking (DST) have shown promising performances on several benchmarks, there is still a significant performance gap between seen slot values (i.e., values that occur in both training set and test set) and unseen ones (values that occur in training set but not in test set). Recently, the copy-mechanism has been widely used in DST models to handle unseen slot values, which copies slot values from user utterance directly. In this paper, we aim to find out the factors that influence the generalization ability of a common copy-mechanism model for DST. Our key observations include: 1) the copy-mechanism tends to memorize values rather than infer them from contexts, which is the primary reason for unsatisfactory generalization performance; 2) greater diversity of slot values in the training set increase the performance on unseen values but slightly decrease the performance on seen values. Moreover, we propose a simple but effective algorithm of data augmentation to train copy-mechanism models, which augments the input dataset by copying user utterances and replacing the real slot values with randomly generated strings. Users could use two hyper-parameters to realize a trade-off between the performances on seen values and unseen ones, as well as a trade-off between overall performance and computational cost. Experimental results on three widely used datasets (WoZ 2.0, DSTC2, and Multi-WoZ 2.0) show the effectiveness of our approach.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Distilling Knowledge from Pre-trained Language Models via Text Smoothing

This paper studies compressing pre-trained language models, like BERT (Devlin et al.,2019), via teacher-student knowledge distillation. Previous works usually force the student model to strictly mimic the smoothed labels predicted by the teacher BERT. As an alternative, we propose a new method for BERT distillation, i.e., asking the teacher to generate smoothed word ids, rather than labels, for teaching the student model in knowledge distillation. We call this kind of methodTextSmoothing. Practically, we use the softmax prediction of the Masked Language Model(MLM) in BERT to generate word distributions for given texts and smooth those input texts using that predicted soft word ids. We assume that both the smoothed labels and the smoothed texts can implicitly augment the input corpus, while text smoothing is intuitively more efficient since it can generate more instances in one neural network forward step.Experimental results on GLUE and SQuAD demonstrate that our solution can achieve competitive results compared with existing BERT distillation methods.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Non-uniform continuity of the generalized Camassa-Holm equation in Besov spaces

In this paper, we consider the Cauchy problem for the generalized Camassa-Holm equation proposed by Hakkaev and Kirchev (2005) \cite{Hakkaev 2005}. We prove that the solution map of the generalized Camassa-Holm equation is not uniformly continuous on the initial data in Besov spaces. Our result include the present work (2020) \cite{Li 2020,Li 2020-1} on Camassa-Holm equation with $Q=1$ and extends the previous non-uniform continuity in Sobolev spaces (2015) \cite{Mi 2015} to Besov spaces. In addition, the non-uniform continuity in critical space $B_{2, 1}^{\frac{3}{2}}(\mathbb{R})$ is the first to be considered in our paper.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

TransSent: Towards Generation of Structured Sentences with Discourse Marker

Structured sentences are important expressions in human writings and dialogues. Previous works on neural text generation fused semantic and structural information by encoding the entire sentence into a mixed hidden representation. However, when a generated sentence becomes complicated, the structure is difficult to be properly maintained. To alleviate this problem, we explicitly separate the modeling process of semantic and structural information. Intuitively, humans generate structured sentences by directly connecting discourses with discourse markers (such as and, but, etc.). Therefore, we propose a task that mimics this process, called discourse transfer. This task represents a structured sentence as (head discourse, discourse marker, tail discourse), and aims at tail discourse generation based on head discourse and discourse marker. We also propose a corresponding model called TransSent, which interprets the relationship between two discourses as a translation1 from the head discourse to the tail discourse in the embedding space. We experiment TransSent not only in discourse transfer task but also in free text generation and dialogue generation tasks. Automatic and human evaluation results show that TransSent can generate structured sentences with high quality, and has certain scalability in different tasks.

0 citations0 reviewsNo code linkedOpen access
preprint2019arXiv

Higher curvature corrections to pole-skipping

Recent developments have revealed a new phenomenon, i.e. the residues of the poles of the holographic retarded two point functions of generic operators vanish at certain complex values of the frequency and momentum. This so-called pole-skipping phenomenon can be determined holographically by the near horizon dynamics of the bulk equations of the corresponding fields. In particular, the pole-skipping point in the upper half plane of complex frequency has been shown to be closed related to many-body chaos, while those in the lower half plane also places universal and nontrivial constraints on the two point functions. In this paper, we study the effect of higher curvature corrections, i.e. the stringy correction and Gauss-Bonnet correction, to the (lower half plane) pole-skipping phenomenon for generic scalar, vector, and metric perturbations. We find that at the pole-skipping points, the frequencies $ω_n=-i2πnT$ are not explicitly influenced by both $R^2$ and $R^4$ corrections, while the momenta $k_n$ receive corresponding corrections.

0 citations0 reviewsNo code linkedOpen access
preprint2019arXiv

Room temperature 2D ferromagnetism in few-layered 1$T$-CrTe$_{2}$

Spin-related electronics using two dimensional (2D) van der Waals (vdW) materials as a platform are believed to hold great promise for revolutionizing the next generation spintronics. Although many emerging new phenomena have been unravelled in 2D electronic systems with spin long-range orderings, the scarcely reported room temperature magnetic vdW material has thus far hindered the related applications. Here, we show that intrinsic ferromagnetically aligned spin polarization can hold up to 316 K in a metallic phase of 1$T$-CrTe$_{2}$ in the few-layer limit. This room temperature 2D long range spin interaction may be beneficial from an itinerant enhancement. Spin transport measurements indicate an in-plane room temperature negative anisotropic magnetoresistance (AMR) in few-layered CrTe$_{2}$, but a sign change in the AMR at lower temperature, with -0.6$\%$ at 300 K and +5$\%$ at 10 K, respectively. This behavior may originate from the specific spin polarized band structure of CrTe$_{2}$. Our findings provide insights into magnetism in few-layered CrTe$_{2}$, suggesting potential for future room temperature spintronic applications of such 2D vdW magnets.

0 citations0 reviewsNo code linkedOpen access