arXiv subjects

Feng Liu

Publications and source records attributed to Feng Liu.

At least 181 records · Page 10Linked to original sources

arXiv2026

Quantum-geometry-enabled Landau-Zener tunneling in singular flat bands

Flat-band materials have attracted substantial interest for their intriguing quantum geometric effects. Here we investigate how singular flat bands (SFBs) respond to a static, uniform electric field and whether they can support single-particle dc transport. By constructing a minimal two-band lattice model, we show that away from the singular band crossing point (BCP), the Wannier-Stark (WS) spectrum of the flat band is well captured by an intraband Berry phase $Φ_{\mathrm{B}}$. The associated WS eigenstates are exponentially localized along the field direction, precluding dc transport. In contrast, near the BCP the interband Berry connection becomes prominent and drives Landau-Zener tunneling, which bends the flat-band WS ladder and delocalizes the SFB wavefunctions. Remarkably, this regime is governed solely by the maximal quantum distance $d$ through two geometric phases $(θ,φ)$: $θ$ characterizes the tunneling rate and $φ$ acts as a generalized Berry phase. These results highlight the essential role of quantum geometry in enabling nontrivial transport signatures in SFBs.

cond-mat.mes-hall↗

arXiv2026

DrawSim-PD: Simulating Student Science Drawings to Support NGSS-Aligned Teacher Diagnostic Reasoning

Developing expertise in diagnostic reasoning requires practice with diverse student artifacts, yet privacy regulations prohibit sharing authentic student work for teacher professional development (PD) at scale. We present DrawSim-PD, the first generative framework that simulates NGSS-aligned, student-like science drawings exhibiting controllable pedagogical imperfections to support teacher training. Central to our approach are apability profiles--structured cognitive states encoding what students at each performance level can and cannot yet demonstrate. These profiles ensure cross-modal coherence across generated outputs: (i) a student-like drawing, (ii) a first-person reasoning narrative, and (iii) a teacher-facing diagnostic concept map. Using 100 curated NGSS topics spanning K-12, we construct a corpus of 10,000 systematically structured artifacts. Through an expert-based feasibility evaluation, K--12 science educators verified the artifacts' alignment with NGSS expectations (>84% positive on core items) and utility for interpreting student thinking, while identifying refinement opportunities for grade-band extremes. We release this open infrastructure to overcome data scarcity barriers in visual assessment research.

cs.CY↗

arXiv2026

LocalScore: Local Density-Aware Similarity Scoring for Biometrics

Open-set biometrics faces challenges with probe subjects who may not be enrolled in the gallery, as traditional biometric systems struggle to detect these non-mated probes. Despite the growing prevalence of multi-sample galleries in real-world deployments, most existing methods collapse intra-subject variability into a single global representation, leading to suboptimal decision boundaries and poor open-set robustness. To address this issue, we propose LocalScore, a simple yet effective scoring algorithm that explicitly incorporates the local density of the gallery feature distribution using the k-th nearest neighbors. LocalScore is architecture-agnostic, loss-independent, and incurs negligible computational overhead, making it a plug-and-play solution for existing biometric systems. Extensive experiments across multiple modalities demonstrate that LocalScore consistently achieves substantial gains in open-set retrieval (FNIR@FPIR reduced from 53% to 40%) and verification (TAR@FAR improved from 51% to 74%). We further provide theoretical analysis and empirical validation explaining when and why the method achieves the most significant gains based on dataset characteristics.

cs.CV↗

arXiv2026

Visual-Guided Key-Token Regularization for Multimodal Large Language Model Unlearning

Unlearning in Multimodal Large Language Models (MLLMs) prevents the model from revealing private information when queried about target images. Existing MLLM unlearning methods largely adopt approaches developed for LLMs. They treat all answer tokens uniformly, disregarding their varying importance in the unlearning process. Moreover, these methods focus exclusively on the language modality, disregarding visual cues that indicate key tokens in answers. In this paper, after formulating the problem of unlearning in multimodal question answering for MLLMs, we propose Visual-Guided Key-Token Regularization (ViKeR). We leverage irrelevant visual inputs to predict ideal post-unlearning token-level distributions and use these distributions to regularize the unlearning process, thereby prioritizing key tokens. Further, we define key tokens in unlearning via information entropy and discuss ViKeR's effectiveness through token-level gradient reweighting, which amplifies updates on key tokens. Experiments on MLLMU and CLEAR benchmarks demonstrate that our method effectively performs unlearning while mitigating forgetting and maintaining response coherence.

cs.LG↗

arXiv2026

Per-parameter Task Arithmetic for Unlearning in Large Language Models

In large language model (LLM) unlearning, private information is required to be removed. Task arithmetic unlearns by subtracting a specific task vector (TV)--defined as the parameter difference between a privacy-information-tuned model and the original model. While efficient, it can cause over-forgetting by disrupting parameters essential for retaining other information. Motivated by the observation that each parameter exhibits different importance for forgetting versus retention, we propose a per-parameter task arithmetic (PerTA) mechanism to rescale the TV, allowing per-parameter adjustment. These weights quantify the relative importance of each parameter for forgetting versus retention, estimated via gradients (i.e., PerTA-grad) or the diagonal Fisher information approximation (i.e., PerTA-fisher). Moreover, we discuss the effectiveness of PerTA, extend it to a more general form, and provide further analysis. Extensive experiments demonstrate that PerTA consistently improves upon standard TV, and in many cases surpasses widely used training-based unlearning methods in both forgetting effectiveness and overall model utility. By retaining the efficiency of task arithmetic while mitigating over-forgetting, PerTA offers a principled and practical framework for LLM unlearning.

cs.LG↗

arXiv2026

NLPrompt: Noise-Label Prompt Learning for Vision-Language Models

The emergence of vision-language foundation models, such as CLIP, has revolutionized image-text representation, enabling a broad range of applications via prompt learning. Despite its promise, real-world datasets often contain noisy labels that can degrade prompt learning performance. In this paper, we demonstrate that using mean absolute error (MAE) loss in prompt learning, named PromptMAE, significantly enhances robustness against noisy labels while maintaining high accuracy. Though MAE is straightforward and recognized for its robustness, it is rarely used in noisy-label learning due to its slow convergence and poor performance outside prompt learning scenarios. To elucidate the robustness of PromptMAE, we leverage feature learning theory to show that MAE can suppress the influence of noisy samples, thereby improving the signal-to-noise ratio and enhancing overall robustness. Additionally, we introduce PromptOT, a prompt-based optimal transport data purification method to enhance the robustness further. PromptOT employs text features in vision-language models as prototypes to construct an optimal transportation matrix. This matrix effectively partitions datasets into clean and noisy subsets, allowing for the application of cross-entropy loss to the clean subset and MAE loss to the noisy subset. Our Noise-Label Prompt Learning method, named NLPrompt, offers a simple and efficient approach that leverages the expressive representations and precise alignment capabilities of vision-language models for robust prompt learning. We validate NLPrompt through extensive experiments across various noise settings, demonstrating significant performance improvements.

cs.CV↗

arXiv2026

Field Matters: A Lightweight LLM-enhanced Method for CTR Prediction

Click-through rate (CTR) prediction is a fundamental task in modern recommender systems. In recent years, the integration of large language models (LLMs) has been shown to effectively enhance the performance of traditional CTR methods. However, existing LLM-enhanced methods often require extensive processing of detailed textual descriptions for large-scale instances or user/item entities, leading to substantial computational overhead. To address this challenge, this work introduces LLaCTR, a novel and lightweight LLM-enhanced CTR method that employs a field-level enhancement paradigm. Specifically, LLaCTR first utilizes LLMs to distill crucial and lightweight semantic knowledge from small-scale feature fields through self-supervised field-feature fine-tuning. Subsequently, it leverages this field-level semantic knowledge to enhance both feature representation and feature interactions. In our experiments, we integrate LLaCTR with six representative CTR models across four datasets, demonstrating its superior performance in terms of both effectiveness and efficiency compared to existing LLM-enhanced methods. Our code is available at https://github.com/istarryn/LLaCTR.

cs.IR↗

arXiv2026

Structure-Aware Decoding Mechanisms for Complex Entity Extraction with Large-Scale Language Models

This paper proposes a structure-aware decoding method based on large language models to address the difficulty of traditional approaches in maintaining both semantic integrity and structural consistency in nested and overlapping entity extraction tasks. The method introduces a candidate span generation mechanism and structured attention modeling to achieve unified modeling of entity boundaries, hierarchical relationships, and cross-dependencies. The model first uses a pretrained language model to obtain context-aware semantic representations, then captures multi-granular entity span features through candidate representation combinations, and introduces hierarchical structural constraints during decoding to ensure consistency between semantics and structure. To enhance stability in complex scenarios, the model jointly optimizes classification loss and structural consistency loss, maintaining high recognition accuracy under multi-entity co-occurrence and long-sentence dependency conditions. Experiments conducted on the ACE 2005 dataset demonstrate significant improvements in Accuracy, Precision, Recall, and F1-Score, particularly in nested and overlapping entity recognition, where the model shows stronger boundary localization and structural modeling capability. This study verifies the effectiveness of structure-aware decoding in complex semantic extraction tasks, provides a new perspective for developing language models with hierarchical understanding, and establishes a methodological foundation for high-precision information extraction.

cs.CL↗

arXiv2026

Mechanical sensing of metamagnetic tricriticality in two-dimensional CrI3

Layered Ising metamagnets are antiferromagnetic (AF) materials consisting of monolayer Ising ferromagnets coupled to each other via interlayer AF interactions. They exhibit rich magnetic phase diagrams, featuring tricritical and critical end points, due to the competing magnetic interactions and the Ising anisotropy. While conventional thermodynamic probes can identify these critical points in bulk Ising metamagnets, achieving this in the two-dimensional (2D) limit, where enhanced fluctuation effects can substantially modify critical phenomena, remains to be realized. Here, we combine specific heat capacity (C_V) and magnetic circular dichroism measurements to identify these critical points, extract a tricritical exponent, and map out the complete magnetic phase diagram of 2D Ising metamagnetic CrI3. This is achieved in a nanomechanical device of 6-layer CrI3, in which a direct measurement of the temperature derivative of its mechanical resonance frequency gives C_V. The tricritical point is identified by the onset of an abrupt spin-flip transition on one side and, on the other side, by a vanishing specific heat λ-anomaly for a continuous AF phase transition. In contrast, only the spin-flip transition remains near the critical end point. Our results establish nanomechanical calorimetry as a general route to classify metamagnetic phase transitions and to study multicritical phenomena in 2D magnets.

cond-mat.mes-hall↗

arXiv2026

Let's Roll a BiFTA: Bi-refinement for Fine-grained Text-visual Alignment in Vision-Language Models

Recent research has shown that aligning fine-grained text descriptions with localized image patches can significantly improve the zero-shot performance of pre-trained vision-language models (e.g., CLIP). However, we find that both fine-grained text descriptions and localized image patches often contain redundant information, making text-visual alignment less effective. In this paper, we tackle this issue from two perspectives: \emph{View Refinement} and \emph{Description refinement}, termed as \textit{\textbf{Bi}-refinement for \textbf{F}ine-grained \textbf{T}ext-visual \textbf{A}lignment} (BiFTA). \emph{View refinement} removes redundant image patches with high \emph{Intersection over Union} (IoU) ratios, resulting in more distinctive visual samples. \emph{Description refinement} removes redundant text descriptions with high pairwise cosine similarity, ensuring greater diversity in the remaining descriptions. BiFTA achieves superior zero-shot performance on 6 benchmark datasets for both ViT-based and ResNet-based CLIP, justifying the necessity to remove redundant information in visual-text alignment.

cs.CV↗

arXiv2026

Does LLM Focus on the Right Words? Mitigating Context Bias in LLM-based Recommenders

Large language models (LLMs), owing to their extensive open-domain knowledge and semantic reasoning capabilities, have been increasingly integrated into recommender systems (RS). However, a substantial gap remains between the pre-training objectives of LLMs and the specific requirements of recommendation tasks. To address this gap, supervised fine-tuning (SFT) is commonly performed on specially curated recommendation datasets to further enhance their predictive ability. Despite its success, SFT exhibits a critical limitation: it induces Context Bias, whereby the model over-relies on auxiliary tokens, such as task descriptions and prefix-generated tokens, while underutilizing core user interaction tokens that encode user-specific preferences. This bias not only undermines recommendation accuracy but also raises unfairness concerns. To address this issue, we propose Group Distributionally Robust Optimization-based Tuning (GDRT), a novel fine-tuning paradigm that enforces consistent model performance across token groups with varying degrees of relevance to auxiliary tokens. By adaptively upweighting underperforming groups, typically those weakly correlated with auxiliary tokens, GDRT shifts the model's attention from superficial auxiliary cues to informative user interaction tokens, thereby mitigating context bias. Extensive experiments conducted on three public datasets demonstrate that GDRT effectively mitigates context bias, yielding substantial improvements in recommendation accuracy (with an average NDCG@10 gain of 24.29%) and significantly enhancing recommendation fairness. The code is available at https://github.com/WANGBohaO-jpg/GDRT.

cs.IR↗

arXiv2026

Superconductivity in non-centrosymmetric rhombohedral NbSe2

Crystal stacking offers a powerful yet underexplored route to engineer symmetry in layered superconductors. Here we report superconductivity in rhombohedral-stacked NbSe2 (3R-NbSe2), a non-centrosymmetric polytype in which global inversion symmetry is removed by stacking alone. Using comprehensive structural, transport, magnetic, and thermodynamic measurements, we establish superconductivity as a bulk property of the 3R phase and find that the in-plane upper critical field exceeds the Pauli paramagnetic limit, indicating the persistence of strong Ising-type spin-orbit coupling. Unlike the thickness-dependent superconductivity in centrosymmetric 2H-NbSe2, the superconducting transition temperature in 3R-NbSe2 shows little dependence on layer number but exhibits an unusually strong sensitivity to disorder. We further observe strongly enhanced nonlinear optical and electrical responses near the superconducting transition, consistent with stacking-induced inversion-symmetry breaking. Our results identify 3R-NbSe2 as a single-phase platform in which stacking engineering reshapes superconductivity and enables nonlinear transport phenomena in layered materials.

cond-mat.supr-con↗

arXiv2026

Advances in LLM Reasoning Enable Flexibility in Clinical Problem-Solving

Large Language Models (LLMs) have achieved high accuracy on medical question-answer (QA) benchmarks, yet their capacity for flexible clinical reasoning has been debated. Here, we asked whether advances in reasoning LLMs improve their cognitive flexibility in clinical reasoning. We assessed reasoning models from the OpenAI, Grok, Gemini, Claude, and DeepSeek families on the medicine abstraction and reasoning corpus (mARC), an adversarial medical QA benchmark which utilizes the Einstellung effect to induce inflexible overreliance on learned heuristic patterns in contexts where they become suboptimal. We found that strong reasoning models avoided Einstellung-based traps more often than weaker reasoning models, achieving human-level performance on mARC. On questions most commonly missed by physicians, the top 5 performing models answered 55% to 70% correctly with high confidence, indicating that these models may be less susceptible than humans to Einstellung effects. Our results indicate that strong reasoning models demonstrate improved flexibility in medical reasoning, achieving performance on par with humans on mARC.

cs.CL↗

arXiv2026

Matrix-Valued Passivity Indices: Foundations, Properties, and Stability Implications

The passivity index, a quantitative measure of a system's passivity deficiency or excess, has been widely used in stability analysis and control. Existing studies mostly rely on scalar forms of indices, which are restrictive for multi-input, multi-output (MIMO) systems. This paper extends the classical scalar indices to a systematic matrix-valued framework, referred to as passivity matrices. A broad range of classical results in passivity theory can be naturally generalized in this framework. We first show that, under the matrix representation, passivity indices essentially correspond to the curvature of the dissipativity functional under a second-variation interpretation. This result reveals that the intrinsic geometric structure of passivity consists of its directions and intensities, which a scalar index cannot fully capture. For linear time-invariant (LTI) systems, we examine the structural properties of passivity matrices with respect to the Loewner partial order and propose two principled criteria for selecting representative matrices. Compared with conventional scalar indices, the matrix-valued indices capture the passivity coupling among different input-output channels in MIMO systems and provide a more comprehensive description of system passivity. This richer information leads to lower passivation effort and less conservative stability assessment.

eess.SY↗

arXiv2026

DreamLoop: Controllable Cinemagraph Generation from a Single Photograph

Cinemagraphs, which combine static photographs with selective, looping motion, offer unique artistic appeal. Generating them from a single photograph in a controllable manner is particularly challenging. Existing image-animation techniques are restricted to simple, low-frequency motions and operate only in narrow domains with repetitive textures like water and smoke. In contrast, large-scale video diffusion models are not tailored for cinemagraph constraints and lack the specialized data required to generate seamless, controlled loops. We present DreamLoop, a controllable video synthesis framework dedicated to generating cinemagraphs from a single photo without requiring any cinemagraph training data. Our key idea is to adapt a general video diffusion model by training it on two objectives: temporal bridging and motion conditioning. This strategy enables flexible cinemagraph generation. During inference, by using the input image as both the first- and last- frame condition, we enforce a seamless loop. By conditioning on static tracks, we maintain a static background. Finally, by providing a user-specified motion path for a target object, our method provides intuitive control over the animation's trajectory and timing. To our knowledge, DreamLoop is the first method to enable cinemagraph generation for general scenes with flexible and intuitive controls. We demonstrate that our method produces high-quality, complex cinemagraphs that align with user intent, outperforming existing approaches.

cs.CV↗

arXiv2026

Band-Structure-Independent Topology from Nonsymmorphic Wannier Complexes

Nonsymmorphic symmetries can enforce band connectivity that obstructs a single-band Wannier description. We show that a fractional translation $\mathcal{L}$ connecting distinct high-symmetry Wyckoff positions generically renders the Wannier center of an individual band gauge ill-defined, requiring a symmetry-enforced multiband object -- a Wannier complex. We formulate a real-space topological classification of Wannier complexes and show that, when $\mathcal{L}$ is combined with certain point-group symmetries (notably $C_4$ and $C_3$), all symmetry-allowed Wannier-complex configurations carry a nontrivial quantized total electric polarization. This yields boundary phenomena that persist across symmetry-preserving deformations of the Hamiltonian, including parameter regimes with and without bulk gaps. We demonstrate the mechanism in minimal tight-binding models exhibiting M{ö}bius-twisted Wilson-loop structures and higher-order corner modes, and propose experimental signatures in a dielectric photonic crystal and a first-principles electronic platform octa-graphene, accompanied by a three-dimensional extension.

cond-mat.supr-con↗

arXiv2026

CTTA-T: Continual Test-Time Adaptation for Text Understanding via Teacher-Student with a Domain-aware and Generalized Teacher

Text understanding often suffers from domain shifts. To handle testing domains, domain adaptation (DA) is trained to adapt to a fixed and observed testing domain; a more challenging paradigm, test-time adaptation (TTA), cannot access the testing domain during training and online adapts to the testing samples during testing, where the samples are from a fixed domain. We aim to explore a more practical and underexplored scenario, continual test-time adaptation (CTTA) for text understanding, which involves a sequence of testing (unobserved) domains in testing. Current CTTA methods struggle in reducing error accumulation over domains and enhancing generalization to handle unobserved domains: 1) Noise-filtering reduces accumulated errors but discards useful information, and 2) accumulating historical domains enhances generalization, but it is hard to achieve adaptive accumulation. In this paper, we propose a CTTA-T (continual test-time adaptation for text understanding) framework adaptable to evolving target domains: it adopts a teacher-student framework, where the teacher is domain-aware and generalized for evolving domains. To improve teacher predictions, we propose a refine-then-filter based on dropout-driven consistency, which calibrates predictions and removes unreliable guidance. For the adaptation-generalization trade-off, we construct a domain-aware teacher by dynamically accumulating cross-domain semantics via incremental PCA, which continuously tracks domain shifts. Experiments show CTTA-T excels baselines.

cs.CL↗

arXiv2025

Generalising Traffic Forecasting to Regions without Traffic Observations

Traffic forecasting is essential for intelligent transportation systems. Accurate forecasting relies on continuous observations collected by traffic sensors. However, due to high deployment and maintenance costs, not all regions are equipped with such sensors. This paper aims to forecast for regions without traffic sensors, where the lack of historical traffic observations challenges the generalisability of existing models. We propose a model named GenCast, the core idea of which is to exploit external knowledge to compensate for the missing observations and to enhance generalisation. We integrate physics-informed neural networks into GenCast, enabling physical principles to regularise the learning process. We introduce an external signal learning module to explore correlations between traffic states and external signals such as weather conditions, further improving model generalisability. Additionally, we design a spatial grouping module to filter localised features that hinder model generalisability. Extensive experiments show that GenCast consistently reduces forecasting errors on multiple real-world datasets.

cs.LG↗