arXiv subjects

Yiming Liu

Publications and source records attributed to Yiming Liu.

At least 73 records · Page 4Linked to original sources

arXiv2025

Semantic Radio Access Networks: Architecture, State-of-the-Art, and Future Directions

Radio Access Network (RAN) is a bridge between user devices and the core network in mobile communication systems, responsible for the transmission and reception of wireless signals and air interface management. In recent years, Semantic Communication (SemCom) has represented a transformative communication paradigm that prioritizes meaning-level transmission over conventional bit-level delivery, thus providing improved spectrum efficiency, anti-interference ability in complex environments, flexible resource allocation, and enhanced user experience for RAN. However, there is still a lack of comprehensive reviews on the integration of SemCom into RAN. Motivated by this, we systematically explore recent advancements in Semantic RAN (SemRAN). We begin by introducing the fundamentals of RAN and SemCom, identifying the limitations of conventional RAN, and outlining the overall architecture of SemRAN. Subsequently, we review representative techniques of SemRAN across physical layer, data link layer, network layer, and security plane. Furthermore, we envision future services and applications enabled by SemRAN, alongside its current standardization progress. Finally, we conclude by identifying critical research challenges and outlining forward-looking directions to guide subsequent investigations in this burgeoning field.

eess.SP↗

arXiv2025

Probing top-quark electroweak couplings indirectly at the Electron-Ion Collider

Top quark electroweak interactions serve as a sensitive probe for beyond-Standard-Model physics, potentially exhibiting significant deviations from Standard Model predictions and offering crucial tests of fundamental symmetries in ultraviolet physics. This Letter investigates their measurement through deep inelastic scattering (DIS) processes involving top quark loops at the Electron-Ion Collider (EIC), within the Standard Model Effective Field Theory framework. We demonstrate that left-handed electron beam polarization can significantly enhance DIS cross sections for these interactions compared to right-handed scenario, highlighting the pivotal role of high polarization at the EIC in constraining these couplings. Compared to direct $pp \to t\bar{t}Z$ measurements at the Large Hadron Collider (LHC), DIS measurements could significantly improve these coupling constraints and effectively resolve parameter space degeneracies present in LHC data. Moreover, the expected precision from the EIC's high-luminosity phase matches that of LEP electroweak precision measurements, underscoring the EIC's significant potential to probe top quark properties.

hep-ph↗

arXiv2025

Perturbation-assisted Observation of the Lowest Vibrational Level of the $\mathrm{b}^{3}Π_{0}$ State of Ultracold LiK Molecules

The narrow transition from the lowest rovibrational level of the $\mathrm{X}^{1}Σ^{+}$ electronic ground state to the lowest vibrational level of the $\mathrm{b}^{3}Π_{0}$ potential provides opportunities for achieving magic-wavelength trapping of ultracold bialkali molecules for enhancing their rotational coherence times. Guided by existing spectroscopic data of several perturbed and deeply-bound rovibrational states of the $\mathrm{A}^{1}Σ^{+}$ potential [Grochola et al., Chem. Phys. Lett., 2012, 535, 17-20], we conducted a targeted spectroscopic search and report the first observation of the lowest vibrational level of the $\mathrm{b}^{3}Π_{0}$ state in $^{6}\mathrm{Li}^{40}\mathrm{K}$. The transition frequency from $|\mathrm{X}^{1}Σ^{+},\,v=0,\,J=0>$ to $|\mathrm{b}^{3}Π_{0},\,v'=0,\,J'=1>$ is determined to be 314,230.5(5)GHz. Assisted by microwave spectroscopy, we resolved the rotational structure of $|\mathrm{b}^{3}Π_{0},\,v'=0>$ and extracted a rotational constant of $h\times8.576(44)$ GHz for the $\mathrm{b}^{3}Π_{0}$ state. From this, we deducted an energy separation between $|\mathrm{b}^{3}Π_{0},v'=0,J'=0>$ and $|\mathrm{X}^{1}Σ^{+},v=0,J=0>$ of $hc\times$10,481.03(2) $\mathrm{cm}^{-1}$. Our work provides timely and precise information on the deeply-bound region of the $\mathrm{b}^{3}Π_{0}$ triplet excited potential of LiK, and benefits future applications of ultracold LiK isotopologues in quantum simulation and quantum computation that demand long coherence times.

physics.atom-ph↗

arXiv2025

SUPERChem: A Multimodal Reasoning Benchmark in Chemistry

Current benchmarks for evaluating the chemical reasoning capabilities of Large Language Models (LLMs) are limited by oversimplified tasks, lack of process-level evaluation, and misalignment with expert-level chemistry skills. To address these issues, we introduce SUPERChem, a benchmark of 500 expert-curated reasoning-intensive chemistry problems, covering diverse subfields and provided in both multimodal and text-only formats. Original content and an iterative curation pipeline eliminate flawed items and mitigate data contamination. Each problem is paired with an expert-authored solution path, enabling Reasoning Path Fidelity (RPF) scoring to evaluate reasoning quality beyond final-answer accuracy. Evaluations against a human baseline of 40.3% accuracy show that even the best-performing model, GPT-5 (High), reaches only 38.5%, followed closely by Gemini 2.5 Pro (37.9%) and DeepSeek-V3.1-Think (37.3%). SUPERChem elicits multi-step, multimodal reasoning, reveals model-dependent effects of visual information, and distinguishes high-fidelity reasoners from heuristic ones. By providing a challenging benchmark and a reliable evaluation framework, SUPERChem aims to facilitate the advancement of LLMs toward expert-level chemical intelligence. The dataset of the benchmark is available at https://huggingface.co/datasets/ZehuaZhao/SUPERChem.

cs.CL↗

arXiv2025

Data or Language Supervision: What Makes CLIP Better than DINO?

CLIP outperforms self-supervised models like DINO as vision encoders for vision-language models (VLMs), but it remains unclear whether this advantage stems from CLIP's language supervision or its much larger training data. To disentangle these factors, we pre-train CLIP and DINO under controlled settings -- using the same architecture, dataset, and training configuration -- achieving similar ImageNet accuracy. Embedding analysis shows that CLIP captures high-level semantics (e.g., object categories, text), while DINO is more responsive to low-level features like colors and styles. When integrated into VLMs and evaluated on 20 VQA benchmarks, CLIP excels at text-intensive tasks, while DINO slightly outperforms on vision-centric ones. Variants of language supervision (e.g., sigmoid loss, pre-trained language encoders) yield limited gains. Our findings provide scientific insights into vision encoder design and its impact on VLM performance.

cs.CV↗

arXiv2025

OpenRLHF: An Easy-to-use, Scalable and High-performance RLHF Framework

Large Language Models (LLMs) fine-tuned via Reinforcement Learning from Human Feedback (RLHF) and Reinforcement Learning with Verifiable Rewards (RLVR) significantly improve the alignment of human-AI values, further raising the upper bound of AI capabilities, particularly in reasoning-intensive, long-context Chain-of-Thought (CoT) tasks. However, existing frameworks commonly face challenges such as inference bottlenecks and complexity barriers, which restrict their accessibility to newcomers. To bridge this gap, we introduce \textbf{OpenRLHF}, a user-friendly, scalable, and easy-to-learn open-source RLHF framework built upon Ray, vLLM, DeepSpeed, and HuggingFace Transformers, featuring a simplified design, clear code structure, and comprehensive documentation to facilitate entry for researchers and practitioners. Experimental results show that OpenRLHF achieves superior training efficiency, with speedups ranging from 1.22x to 1.68x across different model sizes, compared to state-of-the-art frameworks. Additionally, it requires significantly fewer lines of code for implementation. OpenRLHF is publicly available at https://github.com/OpenRLHF/OpenRLHF, and has already been adopted by leading institutions to accelerate RLHF research and learning.

cs.AI↗

arXiv2025

Rethinking KL Regularization in RLHF: From Value Estimation to Gradient Optimization

Reinforcement Learning from Human Feedback (RLHF) leverages a Kullback-Leibler (KL) divergence loss to stabilize training and prevent overfitting. However, in methods such as GRPO, its implementation may be guided by principles from numerical value estimation-a practice that overlooks the term's functional role as an optimization loss. To analyze this issue, we establish a unified framework that connects two seemingly distinct implementation styles: using the mathematical term $k_n$ as a detached coefficient for the policy's score function ('$k_n$ in reward') or as a direct loss function through which gradients are propagated ('$k_n$ as loss'). We show that the latter can always be analyzed via an equivalent gradient coefficient in the former, unifying the two perspectives. Through this framework, we prove that the conventional '$k_1$ in reward' (like in PPO) is the principled loss for Reverse KL (RKL) regularization. We further establish a key finding: under on-policy conditions, the '$k_2$ as loss' formulation is, in fact, gradient-equivalent to '$k_1$ in reward'. This equivalence, first proven in our work, identifies both as the theoretically sound implementations of the RKL objective. In contrast, we show that the recently adopted '$k_3$ as loss' (like in GRPO) is merely a first-order, biased approximation of the principled loss. Furthermore, we argue that common off-policy implementations of '$k_n$ as loss' methods are biased due to neglected importance sampling, and we propose a principled correction. Our findings provide a comprehensive, gradient-based rationale for choosing and correctly implementing KL regularization, paving the way for more robust and effective RLHF systems.

cs.LG↗

arXiv2025

SSNet: Flexible and robust channel extrapolation for fluid antenna systems enabled by an self-supervised learning framework

Fluid antenna systems (FAS) signify a pivotal advancement in 6G communication by enhancing spectral efficiency and robustness. However, obtaining accurate channel state information (CSI) in FAS poses challenges due to its complex physical structure. Traditional methods, such as pilot-based interpolation and compressive sensing, are not only computationally intensive but also lack adaptability. Current extrapolation techniques relying on rigid parametric models do not accommodate the dynamic environment of FAS, while data-driven deep learning approaches demand extensive training and are vulnerable to noise and hardware imperfections. To address these challenges, this paper introduces a novel self-supervised learning network (SSNet) designed for efficient and adaptive channel extrapolation in FAS. We formulate the problem of channel extrapolation in FAS as an image reconstruction task. Here, a limited number of unmasked pixels (representing the known CSI of the selected ports) are used to extrapolate the masked pixels (the CSI of unselected ports). SSNet capitalizes on the intrinsic structure of FAS channels, learning generalized representations from raw CSI data, thus reducing dependency on large labelled datasets. For enhanced feature extraction and noise resilience, we propose a mix-of-expert (MoE) module. In this setup, multiple feedforward neural networks (FFNs) operate in parallel. The outputs of the MoE module are combined using a weighted sum, determined by a gating function that computes the weights of each FFN using a softmax function. Extensive simulations validate the superiority of the proposed model. Results indicate that SSNet significantly outperforms benchmark models, such as AGMAE and long short-term memory (LSTM) networks by using a much smaller labelled dataset.

eess.SP↗

arXiv2025

RIS-Assisted Joint Sensing and Communications via Fractionally Constrained Fractional Programming

This paper studies an uplink dual-functional sensing and communication system aided by a reconfigurable intelligent surface (RIS), whose reflection pattern is configured to trade-off sensing and communication functionalities. Specifically, the Bayesian Cramér-Rao lower bound (BCRLB) for estimating the azimuth angle of a sensing user is minimized while ensuring the signal-to-interference-plus-noise ratio constraints for communication users. We show that this problem can be formulated as a novel fractionally constrained fractional programming (FCFP) problem. To deal with this nontrivial optimization problem, we extend a quadratic transform technique, originally proposed to handle optimization problems containing fractional structures only in objectives, to the scenario where the constraints also include ratios. First, we consider the case where the fading coefficient is known. Using the quadratic transform, the FCFP problem can be turned into a sequence of subproblems that are convex except for the constant-modulus constraints which can be tackled using a penalty-based approach. To further reduce the computational complexity, we leverage the constant-modulus conditions and propose a novel linear transform. This new transform enables the FCFP problem to be turned into a sequence of linear programming (LP) subproblems, which can be solved efficiently. Then, we consider the case where the fading coefficient is unknown. A modified BCRLB is used to make the problem more tractable, and the proposed quadratic transform based algorithm is used to solve the problem. Numerical results unveil nontrivial and effective reflection patterns that can be synthesized by the RIS to facilitate both communication and sensing functionalities.

eess.SP↗

arXiv2025

Celler:A Genomic Language Model for Long-Tailed Single-Cell Annotation

Recent breakthroughs in single-cell technology have ushered in unparalleled opportunities to decode the molecular intricacy of intricate biological systems, especially those linked to diseases unique to humans. However, these progressions have also ushered in novel obstacles-specifically, the efficient annotation of extensive, long-tailed single-cell data pertaining to disease conditions. To effectively surmount this challenge, we introduce Celler, a state-of-the-art generative pre-training model crafted specifically for the annotation of single-cell data. Celler incorporates two groundbreaking elements: First, we introduced the Gaussian Inflation (GInf) Loss function. By dynamically adjusting sample weights, GInf Loss significantly enhances the model's ability to learn from rare categories while reducing the risk of overfitting for common categories. Secondly, we introduce an innovative Hard Data Mining (HDM) strategy into the training process, specifically targeting the challenging-to-learn minority data samples, which significantly improved the model's predictive accuracy. Additionally, to further advance research in this field, we have constructed a large-scale single-cell dataset: Celler-75, which encompasses 40 million cells distributed across 80 human tissues and 75 specific diseases. This dataset provides critical support for comprehensively exploring the potential of single-cell technology in disease research. Our code is available at https://github.com/AI4science-ym/HiCeller.

q-bio.GN↗

arXiv2025

Way to Build Native AI-driven 6G Air Interface: Principles, Roadmap, and Outlook

Artificial intelligence (AI) is expected to serve as a foundational capability across the entire lifecycle of 6G networks, spanning design, deployment, and operation. This article proposes a native AI-driven air interface architecture built around two core characteristics: compression and adaptation. On one hand, compression enables the system to understand and extract essential semantic information from the source data, focusing on task relevance rather than symbol-level accuracy. On the other hand, adaptation allows the air interface to dynamically transmit semantic information across diverse tasks, data types, and channel conditions, ensuring scalability and robustness. This article first introduces the native AI-driven air interface architecture, then discusses representative enabling methodologies, followed by a case study on semantic communication in 6G non-terrestrial networks. Finally, it presents a forward-looking discussion on the future of native AI in 6G, outlining key challenges and research opportunities.

cs.IT↗

arXiv2025

EEG-based Epileptic Prediction via a Two-stage Channel-aware Set Transformer Network

Epilepsy is a chronic, noncommunicable brain disorder, and sudden seizure onsets can significantly impact patients' quality of life and health. However, wearable seizure-predicting devices are still limited, partly due to the bulky size of EEG-collecting devices. To relieve the problem, we proposed a novel two-stage channel-aware Set Transformer Network that could perform seizure prediction with fewer EEG channel sensors. We also tested a seizure-independent division method which could prevent the adjacency of training and test data. Experiments were performed on the CHB-MIT dataset which includes 22 patients with 88 merged seizures. The mean sensitivity before channel selection was 76.4% with a false predicting rate (FPR) of 0.09/hour. After channel selection, dominant channels emerged in 20 out of 22 patients; the average number of channels was reduced to 2.8 from 18; and the mean sensitivity rose to 80.1% with an FPR of 0.11/hour. Furthermore, experimental results on the seizure-independent division supported our assertion that a more rigorous seizure-independent division should be used for patients with abundant EEG recordings.

eess.SP↗

arXiv2025

Multimodal Visual Image Based User Association and Beamforming Using Graph Neural Networks

This paper proposes an approach that leverages multimodal data by integrating visual images with radio frequency (RF) pilots to optimize user association and beamforming in a downlink wireless cellular network under a max-min fairness criterion. Traditional methods typically optimize wireless system parameters based on channel state information (CSI). However, obtaining accurate CSI requires extensive pilot transmissions, which lead to increased overhead and latency. Moreover, the optimization of user association and beamforming is a discrete and non-convex optimization problem, which is challenging to solve analytically. In this paper, we propose to incorporate visual camera data in addition to the RF pilots to perform the joint optimization of user association and beamforming. The visual image data helps enhance channel awareness, thereby reducing the dependency on extensive pilot transmissions for system optimization. We employ a learning-based approach based on using first a detection neural network that estimates user locations from images, and subsequently two graph neural networks (GNNs) that extract features for system optimization based on the location information and the received pilots, respectively. Then, a multimodal GNN is constructed to integrate the features for the joint optimization user association and beamforming. Simulation results demonstrate that the proposed method achieves superior performance, while having low computational complexity and being interpretable and generalizable, making it an effective solution as compared to traditional methods based only on RF pilots.

eess.SP↗

arXiv2025

NegVQA: Can Vision Language Models Understand Negation?

Negation is a fundamental linguistic phenomenon that can entirely reverse the meaning of a sentence. As vision language models (VLMs) continue to advance and are deployed in high-stakes applications, assessing their ability to comprehend negation becomes essential. To address this, we introduce NegVQA, a visual question answering (VQA) benchmark consisting of 7,379 two-choice questions covering diverse negation scenarios and image-question distributions. We construct NegVQA by leveraging large language models to generate negated versions of questions from existing VQA datasets. Evaluating 20 state-of-the-art VLMs across seven model families, we find that these models struggle significantly with negation, exhibiting a substantial performance drop compared to their responses to the original questions. Furthermore, we uncover a U-shaped scaling trend, where increasing model size initially degrades performance on NegVQA before leading to improvements. Our benchmark reveals critical gaps in VLMs' negation understanding and offers insights into future VLM development. Project page available at https://yuhui-zh15.github.io/NegVQA/.

cs.CL↗

arXiv2025

Asymptotic limits of spiked eigenvalues and eigenvectors of signal-plus-noise matrices with weak signals and heteroskedastic noise

This paper is to study a signal-plus-noise model in high dimensional settings when the dimension and the sample size are comparable. Specifically, we assume that the noise has a general covariance matrix that allows for heteroskedasticity, and that the deterministic signal has the same magnitude as the noise and can have a rank that tends to infinity. We develop the asymptotic limits of the left and right spiked singular vectors of the signal-plusnoise data matrix and the limits of the spiked eigenvalues of the corresponding Gram matrix. As an application, we propose a new criterion to estimate the number of clusters in clustering problems.

math.ST↗

arXiv2025

Learning a General Model: Folding Clothing with Topological Dynamics

The high degrees of freedom and complex structure of garments present significant challenges for clothing manipulation. In this paper, we propose a general topological dynamics model to fold complex clothing. By utilizing the visible folding structure as the topological skeleton, we design a novel topological graph to represent the clothing state. This topological graph is low-dimensional and applied for complex clothing in various folding states. It indicates the constraints of clothing and enables predictions regarding clothing movement. To extract graphs from self-occlusion, we apply semantic segmentation to analyze the occlusion relationships and decompose the clothing structure. The decomposed structure is then combined with keypoint detection to generate the topological graph. To analyze the behavior of the topological graph, we employ an improved Graph Neural Network (GNN) to learn the general dynamics. The GNN model can predict the deformation of clothing and is employed to calculate the deformation Jacobi matrix for control. Experiments using jackets validate the algorithm's effectiveness to recognize and fold complex clothing with self-occlusion.

cs.RO↗

arXiv2025

Automated Generation of Challenging Multiple-Choice Questions for Vision Language Model Evaluation

The rapid development of vision language models (VLMs) demands rigorous and reliable evaluation. However, current visual question answering (VQA) benchmarks often depend on open-ended questions, making accurate evaluation difficult due to the variability in natural language responses. To address this, we introduce AutoConverter, an agentic framework that automatically converts these open-ended questions into multiple-choice format, enabling objective evaluation while reducing the costly multiple-choice question creation process. Our experiments demonstrate that AutoConverter can generate correct and challenging multiple-choice questions, with VLMs demonstrating consistently similar or lower accuracy on these questions compared to human-created ones. Using AutoConverter, we construct VMCBench, a benchmark created by transforming 20 existing VQA datasets into a unified multiple-choice format, totaling 9,018 questions. We comprehensively evaluate 33 state-of-the-art VLMs on VMCBench, setting a new standard for scalable, consistent, and reproducible VLM evaluation.

cs.CV↗

arXiv2025

Towards Understanding the Robustness of Diffusion-Based Purification: A Stochastic Perspective

Diffusion-Based Purification (DBP) has emerged as an effective defense mechanism against adversarial attacks. The success of DBP is often attributed to the forward diffusion process, which reduces the distribution gap between clean and adversarial images by adding Gaussian noise. While this explanation is theoretically sound, the exact role of this mechanism in enhancing robustness remains unclear. In this paper, through empirical analysis, we propose that the intrinsic stochasticity in the DBP process is the primary factor driving robustness. To test this hypothesis, we introduce a novel Deterministic White-Box (DW-box) setting to assess robustness in the absence of stochasticity, and we analyze attack trajectories and loss landscapes. Our results suggest that DBP models primarily rely on stochasticity to avoid effective attack directions, while their ability to purify adversarial perturbations may be limited. To further enhance the robustness of DBP models, we propose Adversarial Denoising Diffusion Training (ADDT), which incorporates classifier-guided adversarial perturbations into the diffusion training process, thereby strengthening the models' ability to purify adversarial perturbations. Additionally, we propose Rank-Based Gaussian Mapping (RBGM) to improve the compatibility of perturbations with diffusion models. Experimental results validate the effectiveness of ADDT. In conclusion, our study suggests that future research on DBP can benefit from a clearer distinction between stochasticity-driven and purification-driven robustness.

cs.CV↗