arXiv subjects

Jia Liu

Publications and source records attributed to Jia Liu.

At least 145 records · Page 8Linked to original sources

arXiv2025

NTIRE 2025 XGC Quality Assessment Challenge: Methods and Results

This paper reports on the NTIRE 2025 XGC Quality Assessment Challenge, which will be held in conjunction with the New Trends in Image Restoration and Enhancement Workshop (NTIRE) at CVPR 2025. This challenge is to address a major challenge in the field of video and talking head processing. The challenge is divided into three tracks, including user generated video, AI generated video and talking head. The user-generated video track uses the FineVD-GC, which contains 6,284 user generated videos. The user-generated video track has a total of 125 registered participants. A total of 242 submissions are received in the development phase, and 136 submissions are received in the test phase. Finally, 5 participating teams submitted their models and fact sheets. The AI generated video track uses the Q-Eval-Video, which contains 34,029 AI-Generated Videos (AIGVs) generated by 11 popular Text-to-Video (T2V) models. A total of 133 participants have registered in this track. A total of 396 submissions are received in the development phase, and 226 submissions are received in the test phase. Finally, 6 participating teams submitted their models and fact sheets. The talking head track uses the THQA-NTIRE, which contains 12,247 2D and 3D talking heads. A total of 89 participants have registered in this track. A total of 225 submissions are received in the development phase, and 118 submissions are received in the test phase. Finally, 8 participating teams submitted their models and fact sheets. Each participating team in every track has proposed a method that outperforms the baseline, which has contributed to the development of fields in three tracks.

cs.CV↗

arXiv2025

Meta-Learning the Optimal Mixture of Strategies for Online Portfolio Selection

This paper presents an innovative online portfolio selection model, situated within a meta-learning framework, that leverages a mixture policies strategy. The core idea is to simulate a fund that employs multiple fund managers, each skilled in handling different market environments, and dynamically allocate our funding to these fund managers for investment. To address the non-stationary nature of financial markets, we divide the long-term process into multiple short-term processes to adapt to changing environments. We use a clustering method to identify a set of historically high-performing policies, characterized by low similarity, as candidate policies. Additionally, we employ a meta-learning method to search for initial parameters that can quickly adapt to upcoming target investment tasks, effectively providing a set of well-suited initial strategies. Subsequently, we update the initial parameters using the target tasks and determine the optimal mixture weights for these candidate policies. Empirical tests show that our algorithm excels in terms of training time and data requirements, making it particularly suitable for high-frequency algorithmic trading. To validate the effectiveness of our method, we conduct numerical tests on cross-training datasets, demonstrating its excellent transferability and robustness.

math.OC↗

arXiv2025

The HalfDome Multi-Survey Cosmological Simulations: N-body Simulations

Upcoming cosmological surveys have the potential to reach groundbreaking discoveries on multiple fronts, including the neutrino mass, dark energy, and inflation. Most of the key science goals require the joint analysis of datasets from multiple surveys to break parameter degeneracies and calibrate systematics. To realize such analyses, a large set of mock simulations that realistically model correlated observables is required. In this paper we present the N-body component of the HalfDome cosmological simulations, designed for the joint analysis of Stage-IV cosmological surveys, such as Rubin LSST, Euclid, SPHEREx, Roman, DESI, PFS, Simons Observatory, CMB-S4, and LiteBIRD. Our 300TB initial data release includes full-sky lightcones and halo catalogs between $z$=0--4 for 11 fixed cosmology realizations, as well as an additional run with local primordial non-Gaussianity ($f_{\rm NL}$=20). The simulations evolve $6144^3$ particles in a 3.75$\,h^{-1} {\rm Gpc}$ box, reaching a minimum halo mass of $\sim 6 \times 10^{12}\,h^{-1} M_\odot$ and maximum scale of $k \sim 1\,h{\rm Mpc}^{-1}$. Our data is publicly available: instructions to access the data and plans for future data releases can be found at https://halfdomesims.github.io.

astro-ph.CO↗

arXiv2025

Look Twice Before You Answer: Memory-Space Visual Retracing for Hallucination Mitigation in Multimodal Large Language Models

Despite their impressive capabilities, multimodal large language models (MLLMs) are prone to hallucinations, i.e., the generated content that is nonsensical or unfaithful to input sources. Unlike in LLMs, hallucinations in MLLMs often stem from the sensitivity of text decoder to visual tokens, leading to a phenomenon akin to "amnesia" about visual information. To address this issue, we propose MemVR, a novel decoding paradigm inspired by common cognition: when the memory of an image seen the moment before is forgotten, people will look at it again for factual answers. Following this principle, we treat visual tokens as supplementary evidence, re-injecting them into the MLLM through Feed Forward Network (FFN) as "key-value memory" at the middle trigger layer. This "look-twice" mechanism occurs when the model exhibits high uncertainty during inference, effectively enhancing factual alignment. Comprehensive experimental evaluations demonstrate that MemVR significantly mitigates hallucination across various MLLMs and excels in general benchmarks without incurring additional time overhead. The implementation is available from https://github.com/1zhou-Wang/MemVR

cs.CV↗

arXiv2025

Evaluating Temporal Plasticity in Foundation Time Series Models for Incremental Fine-tuning

Time series foundation models excel at diverse time series forecasting tasks, but their capacity for continuous improvement through incremental learning remains unexplored. We present the first comprehensive study investigating these models' temporal plasticity - their ability to progressively enhance performance through continual learning while maintaining existing capabilities. Through experiments on real-world datasets exhibiting distribution shifts, we evaluate both conventional deep learning models and foundation models using a novel continual learning framework. Our findings reveal that while traditional models struggle with performance deterioration during incremental fine-tuning, foundation models like Time-MoE and Chronos demonstrate sustained improvement in predictive accuracy. This suggests that optimizing foundation model fine-tuning strategies may be more valuable than developing domain-specific small models. Our research introduces new evaluation methodologies and insights for developing foundation time series models with robust continuous learning capabilities.

cs.LG↗

arXiv2025

Configuration interaction relativistic Hartree-Fock model

The configuration interaction relativistic Hartree-Fock (CI-RHF) model is developed in this work. Compared to the conventional configuration interaction shell model (CISM), the CI-RHF model can be applied to study the structural properties of a wide range of nuclei without readjusting any parameters, as the effective Hamiltonian for different model space can be deduced consistently from a universal density-dependent Lagrangian based on the Hartree-Fock single-particle basis. The convergence of intermediate-state excitations has been examined in evaluating the effective interactions, and the core-polarization effects are illustrated, by using $^{18}$O as an example. Employing the CI-RHF model, both the bulk properties and low-lying spectra of even-even nuclei $^{18\sim 28}$Ne have been well reproduced with the model space restricted to the $sd$ shell. Studies of the isotopic evolution concerning charge radii and low-lying spectra highlight the shell closure at $N=14$ for neon isotopes. Furthermore, the cross-shell calculations extending from the $sd$ to $pf$ shell successfully reproduced the low-lying spectra of $^{30}$Ne and $^{32}$Ne. Notably, remarkably low excitation energies $E(2^{+}_{1})$ of $^{30}$Ne suggest the disappearance of the conventional magicity $N=20$.

nucl-th↗

arXiv2025

Dark Matter Search with a Resonantly-Coupled Hybrid Spin System

Recent advances in tabletop quantum sensor technology have enabled searches for nongravitational interactions of dark matter (DM). Traditional axion DM experiments rely on sharp resonance, resulting in extensive scanning time to cover a wide mass range. In this work, we present a broadband approach in an alkali-${}^{21}$Ne spin system. We identify two distinct hybrid spin-coupled regimes: a self-compensation (SC) regime at low frequencies and a hybrid spin resonance (HSR) regime at higher frequencies. By utilizing these two distinct regimes, we significantly enhance the bandwidth of ${}^{21}$Ne nuclear spin compared to conventional nuclear magnetic resonance, while maintaining competitive sensitivity. We present a comprehensive broadband search for axion-like dark matter, covering 5 orders of magnitude of Compton frequencies range within $[10^{-2}, \, 10^3]$ Hz. We set new constraints on the axion dark matter interactions with neutrons and protons, accounting for the effects of DM stochasticity. For the axion-neutron coupling, our results reach a low value of $|g_{ann}|\le 3\times 10^{-10}$ in the frequency range $[2\times 10^{-2}, \, 4]$ Hz surpassing astrophysical limits and providing the strongest laboratory constraints in the $[10, \, 100]$ Hz range. For the axion-proton coupling, we offer the best terrestrial constraints for the frequency ranges $[2\times 10^{-2}, \, 5]$Hz and $[16, \, 7\times 10^{2}]$ Hz.

hep-ph↗

arXiv2025

Long-lived Sterile Neutrino Searches at Future Muon Colliders

We explore the potential of studying sterile neutrinos at a future high-energy muon collider, where these particles can generate small active neutrino masses via the seesaw mechanism and exhibit long-lived particle signatures. A Dirac sterile neutrino model with ${\rm U(1)}_{L_μ-L_τ}$ symmetry is introduced, where the heavy right-handed neutrino ($N_R$) produces tiny active neutrino masses, and the light left-handed neutrino ($N_L$) naturally behaves as a long-lived particle. The ${\rm U(1)}_{L_μ-L_τ}$ gauge symmetry also enhances sterile neutrino pair production at a future high-energy muon collider. Using the displaced vertex method, the muon collider can search for heavy sterile neutrino, especially for $m_L> m_W$. We find that a muon collider with $\sqrt{s} = 3~ (10)$ TeV and luminosity $L=1~(10)$ ab$^{-1}$ can probe $N_L$ masses of $m_L \in [100,~1500~(5000)]$ GeV and mixing angles $θ_{νL} \in [10^{-13},~10^{-6}]$.

hep-ph↗

arXiv2025

Hiding Functions within Functions: Steganography by Implicit Neural Representations

Deep steganography utilizes the powerful capabilities of deep neural networks to embed and extract messages, but its reliance on an additional message extractor limits its practical use due to the added suspicion it can raise from steganalyzers. To address this problem, we propose StegaINR, which utilizes Implicit Neural Representation (INR) to implement steganography. StegaINR embeds a secret function into a stego function, which serves as both the message extractor and the stego media for secure transmission on a public channel. Recipients need only use a shared key to recover the secret function from the stego function, allowing them to obtain the secret message. Our approach makes use of continuous functions, enabling it to handle various types of messages. To our knowledge, this is the first work to introduce INR into steganography. We performed evaluations on image and climate data to test our method in different deployment contexts.

cs.CR↗

arXiv2025

Every FLOP Counts: Scaling a 300B Mixture-of-Experts LING LLM without Premium GPUs

In this technical report, we tackle the challenges of training large-scale Mixture of Experts (MoE) models, focusing on overcoming cost inefficiency and resource limitations prevalent in such systems. To address these issues, we present two differently sized MoE large language models (LLMs), namely Ling-Lite and Ling-Plus (referred to as "Bailing" in Chinese, spelled Bǎilíng in Pinyin). Ling-Lite contains 16.8 billion parameters with 2.75 billion activated parameters, while Ling-Plus boasts 290 billion parameters with 28.8 billion activated parameters. Both models exhibit comparable performance to leading industry benchmarks. This report offers actionable insights to improve the efficiency and accessibility of AI development in resource-constrained settings, promoting more scalable and sustainable technologies. Specifically, to reduce training costs for large-scale MoE models, we propose innovative methods for (1) optimization of model architecture and training processes, (2) refinement of training anomaly handling, and (3) enhancement of model evaluation efficiency. Additionally, leveraging high-quality data generated from knowledge graphs, our models demonstrate superior capabilities in tool use compared to other models. Ultimately, our experimental findings demonstrate that a 300B MoE LLM can be effectively trained on lower-performance devices while achieving comparable performance to models of a similar scale, including dense and MoE models. Compared to high-performance devices, utilizing a lower-specification hardware system during the pre-training phase demonstrates significant cost savings, reducing computing costs by approximately 20%. The models can be accessed at https://huggingface.co/inclusionAI.

cs.LG↗

arXiv2025

Test Script Intention Generation for Mobile Application via GUI Image and Code Understanding

Testing is the most direct and effective technique to ensure software quality. Test scripts always play a more important role in mobile app testing than test cases for source code, due to the GUI-intensive and event-driven characteristics of mobile applications (app). Test scripts focus on user interactions and the corresponding response events, which is significant for testing the target app functionalities. Therefore, it is critical to understand the test scripts for better script maintenance and modification. There exist some mature code understanding (i.e., code comment generation) technologies that can be directly applied to functionality source code with business logic. However, such technologies will have difficulties when being applied to test scripts, because test scripts are loosely linked to apps under test (AUT) by widget selectors, and do not contain business logic themselves. In order to solve the test script understanding gap, this paper presents a novel approach, namely TestIntention, to infer the intention of GUI test scripts. Test intention refers to the user expectations of app behaviors for specific operations. TestIntention formalizes test scripts with an operation sequence model. For each operation within the sequence, TestIntention extracts the target widget selector and links the selector to the GUI layout information or the corresponding response events. For widgets identified by XPath, TestIntention utilizes the image understanding technologies to explore the detailed information of the widget images, the intention of which is understood with a deep learning model. For widgets identified by ID, TestIntention first maps the selectors to the response methods with business logic, and then adopts code understanding technologies to describe code in natural language form. Results of all operations are combined to generate test intention for test scripts.

cs.SE↗

arXiv2025

PCL: Prompt-based Continual Learning for User Modeling in Recommender Systems

User modeling in large e-commerce platforms aims to optimize user experiences by incorporating various customer activities. Traditional models targeting a single task often focus on specific business metrics, neglecting the comprehensive user behavior, and thus limiting their effectiveness. To develop more generalized user representations, some existing work adopts Multi-task Learning (MTL)approaches. But they all face the challenges of optimization imbalance and inefficiency in adapting to new tasks. Continual Learning (CL), which allows models to learn new tasks incrementally and independently, has emerged as a solution to MTL's limitations. However, CL faces the challenge of catastrophic forgetting, where previously learned knowledge is lost when the model is learning the new task. Inspired by the success of prompt tuning in Pretrained Language Models (PLMs), we propose PCL, a Prompt-based Continual Learning framework for user modeling, which utilizes position-wise prompts as external memory for each task, preserving knowledge and mitigating catastrophic forgetting. Additionally, we design contextual prompts to capture and leverage inter-task relationships during prompt tuning. We conduct extensive experiments on real-world datasets to demonstrate PCL's effectiveness.

cs.IR↗

arXiv2025

Byzantine-Robust Federated Learning over Ring-All-Reduce Distributed Computing

Federated learning (FL) has gained attention as a distributed learning paradigm for its data privacy benefits and accelerated convergence through parallel computation. Traditional FL relies on a server-client (SC) architecture, where a central server coordinates multiple clients to train a global model, but this approach faces scalability challenges due to server communication bottlenecks. To overcome this, the ring-all-reduce (RAR) architecture has been introduced, eliminating the central server and achieving bandwidth optimality. However, the tightly coupled nature of RAR's ring topology exposes it to unique Byzantine attack risks not present in SC-based FL. Despite its potential, designing Byzantine-robust RAR-based FL algorithms remains an open problem. To address this gap, we propose BRACE (Byzantine-robust ring-all-reduce), the first RAR-based FL algorithm to achieve both Byzantine robustness and communication efficiency. We provide theoretical guarantees for the convergence of BRACE under Byzantine attacks, demonstrate its bandwidth efficiency, and validate its practical effectiveness through experiments. Our work offers a foundational understanding of Byzantine-robust RAR-based FL design.

cs.CR↗

arXiv2025

Cosmological constraints using Minkowski functionals from the first year data of the Hyper Suprime-Cam

We use Minkowski functionals to analyse weak lensing convergence maps from the first-year data release of the Subaru Hyper Suprime-Cam (HSC-Y1) survey. Minkowski functionals provide a description of the morphological properties of a field, capturing the non-Gaussian features of the Universe matter-density distribution. Using simulated catalogs that reproduce survey conditions and encode cosmological information, we emulate Minkowski functionals predictions across a range of cosmological parameters to derive the best-fit from the data. By applying multiple scales cuts, we rigorously mitigate systematic effects, including baryonic feedback and intrinsic alignments. From the analysis, combining constraints of the angular power spectrum and Minkowski functionals, we obtain $S_8 \equiv σ_8\sqrt{Ω_{\rm m}/0.3} = {0.808}_{-0.046}^{+0.033}$ and $Ω_{\rm m} = {0.293}_{-0.043}^{+0.157}$. These results represent a $40\%$ improvement on the $S_8$ constraints compared to using power spectrum only. \newtext{Minkowski functionals results are consistent with other two-point, and higher order statistics constraints using the same data, being in agreement with CMB results from the Planck $S_8$ measurements. Our study demonstrates the power of Minkowski functionals beyond two-point statistics to constrain and break the degeneracy between $Ω_{\rm m}$ and $σ_8$.

astro-ph.CO↗

arXiv2025

GPT-4o as the Gold Standard: A Scalable and General Purpose Approach to Filter Language Model Pretraining Data

Large language models require vast amounts of high-quality training data, but effective filtering of web-scale datasets remains a significant challenge. This paper demonstrates that GPT-4o is remarkably effective at identifying high-quality training data, but its prohibitive cost makes it impractical at web-scale. We propose SIEVE, a lightweight alternative that matches GPT-4o accuracy at less than 1\% of the cost. SIEVE can perform up to 500 filtering operations for the cost of one GPT-4o filtering call. The key to SIEVE is a seamless integration of GPT-4o and lightweight text classification models, using active learning to fine-tune these models in the background with a small number of calls to GPT-4o. Once trained, it performs as well as GPT-4o at a tiny fraction of the cost. Through different filtering prompts, SIEVE can efficiently curate high quality data for general or specialized domains from web-scale corpora -- a valuable capability given the current scarcity of high-quality domain-specific datasets. Extensive experiments using automatic and human evaluation metrics show that SIEVE and GPT-4o achieve similar performance on five highly specific filtering prompts. In addition, when performing quality filtering on web crawl datasets, we demonstrate SIEVE can further improve over state-of-the-art quality filtering methods in the DataComp-LM challenge for selecting LLM pretraining data.

cs.CL↗

arXiv2025

Improved Paraphrase Generation via Controllable Latent Diffusion

Paraphrase generation strives to generate high-quality and diverse expressions of a given text, a domain where diffusion models excel. Though SOTA diffusion generation reconciles generation quality and diversity, textual diffusion suffers from a truncation issue that hinders efficiency and quality control. In this work, we propose \textit{L}atent \textit{D}iffusion \textit{P}araphraser~(LDP), a novel paraphrase generation by modeling a controllable diffusion process given a learned latent space. LDP achieves superior generation efficiency compared to its diffusion counterparts. It can facilitate only input segments to ensure paraphrase semantics, improving the results without external features. Experiments show that LDP better reconciles paraphrase generation quality and diversity than baselines. Further analysis shows that our method is also helpful to other similar text generations and domain adaptations

cs.CL↗

arXiv2025

Prediction-Assisted Online Distributed Deep Learning Workload Scheduling in GPU Clusters

The recent explosive growth of deep learning (DL) models has necessitated a compelling need for efficient job scheduling for distributed deep learning training with mixed parallelisms (DDLwMP) in GPU clusters. This paper proposes an adaptive shortest-remaining-processing-time-first (A-SRPT) scheduling algorithm, a novel prediction-assisted online scheduling approach designed to mitigate the challenges associated with DL cluster scheduling. By modeling each job as a graph corresponding to heterogeneous Deep Neural Network (DNN) models and their associated distributed training configurations, A-SRPT strategically assigns jobs to the available GPUs, thereby minimizing inter-server communication overhead. Observing that most DDLwMP jobs recur, A-SRPT incorporates a random forest regression model to predict training iterations. Crucially, A-SRPT maps the complex scheduling problem into a single-machine instance, which is addressed optimally by a preemptive "shortest-remaining-processing-time-first" strategy. This optimized solution serves as a guide for actual job scheduling within the GPU clusters, leading to a theoretically provable competitive scheduling efficiency. We conduct extensive real-world testbed and simulation experiments to verify our proposed algorithms.

cs.DC↗

arXiv2024

Natural Language Fine-Tuning

Large language model fine-tuning techniques typically depend on extensive labeled data, external guidance, and feedback, such as human alignment, scalar rewards, and demonstration. However, in practical application, the scarcity of specific knowledge poses unprecedented challenges to existing fine-tuning techniques. In this paper, focusing on fine-tuning tasks in specific domains with limited data, we introduce Natural Language Fine-Tuning (NLFT), which utilizes natural language for fine-tuning for the first time. By leveraging the strong language comprehension capability of the target LM, NLFT attaches the guidance of natural language to the token-level outputs. Then, saliency tokens are identified with calculated probabilities. Since linguistic information is effectively utilized in NLFT, our proposed method significantly reduces training costs. It markedly enhances training efficiency, comprehensively outperforming reinforcement fine-tuning algorithms in accuracy, time-saving, and resource conservation. Additionally, on the macro level, NLFT can be viewed as a token-level fine-grained optimization of SFT, thereby efficiently replacing the SFT process without the need for warm-up (as opposed to ReFT requiring multiple rounds of warm-up with SFT). Compared to SFT, NLFT does not increase the algorithmic complexity, maintaining O(n). Extensive experiments on the GSM8K dataset demonstrate that NLFT, with only 50 data instances, achieves an accuracy increase that exceeds SFT by 219%. Compared to ReFT, the time complexity and space complexity of NLFT are reduced by 78.27% and 92.24%, respectively. The superior technique of NLFT is paving the way for the deployment of various innovative LLM fine-tuning applications when resources are limited at network edges. Our code has been released at https://github.com/Julia-LiuJ/NLFT.

cs.CL↗