Matching
-
MAP-VLA: Memory-Augmented Prompting for Vision-Language-Action Model in Robotic Manipulation
2025-11-12 17:56:13Runhao Li, Wenkai Guo, Zhenyu Wu, Changyuan Wang, Haoyuan Deng, Zhenyu Weng, Yap-Peng Tan, Ziwei WangAbstract
Pre-trained Vision-Language-Action (VLA) models have achieved remarkable success in improving robustness and generalization for end-to-end robotic manipulation. However, these models struggle with long-horizon tasks due to their lack of memory and reliance solely on immediate sensory inputs. To address this limitation, we propose Memory-Augmented Prompting for Vision-Language-Action model (MAP-VLA), a novel framework that empowers pre-trained VLA models with demonstration-derived memory prompts to augment action generation for long-horizon robotic manipulation tasks. To achieve this, MAP-VLA first constructs a memory library from historical demonstrations, where each memory unit captures information about a specific stage of a task. These memory units are implemented as learnable soft prompts optimized through prompt tuning. Then, during real-time task execution, MAP-VLA retrieves relevant memory through trajectory similarity matching and dynamically integrates it into the VLA model for augmented action generation. Importantly, this prompt tuning and retrieval augmentation approach operates as a plug-and-play module for a frozen VLA model, offering a lightweight and flexible solution to improve task performance. Experimental results show that MAP-VLA delivers up to 7.0% absolute performance gains in the simulation benchmark and 25.0% on real robot evaluations for long-horizon tasks, surpassing the current state-of-the-art methods.
Abstract (translated)
URL
https://arxiv.org/abs/2511.09516
PDF
-
Learning by Neighbor-Aware Semantics, Deciding by Open-form Flows: Towards Robust Zero-Shot Skeleton Action Recognition
2025-11-12 14:54:53Yang Chen, Miaoge Li, Zhijie Rao, Deze Zeng, Song Guo, Jingcai GuoAbstract
Recognizing unseen skeleton action categories remains highly challenging due to the absence of corresponding skeletal priors. Existing approaches generally follow an "align-then-classify" paradigm but face two fundamental issues, i.e., (i) fragile point-to-point alignment arising from imperfect semantics, and (ii) rigid classifiers restricted by static decision boundaries and coarse-grained anchors. To address these issues, we propose a novel method for zero-shot skeleton action recognition, termed $\texttt{$\textbf{Flora}$}$, which builds upon $\textbf{F}$lexib$\textbf{L}$e neighb$\textbf{O}$r-aware semantic attunement and open-form dist$\textbf{R}$ibution-aware flow cl$\textbf{A}$ssifier. Specifically, we flexibly attune textual semantics by incorporating neighboring inter-class contextual cues to form direction-aware regional semantics, coupled with a cross-modal geometric consistency objective that ensures stable and robust point-to-region alignment. Furthermore, we employ noise-free flow matching to bridge the modality distribution gap between semantic and skeleton latent embeddings, while a condition-free contrastive regularization enhances discriminability, leading to a distribution-aware classifier with fine-grained decision boundaries achieved through token-level velocity predictions. Extensive experiments on three benchmark datasets validate the effectiveness of our method, showing particularly impressive performance even when trained with only 10\% of the seen data. Code is available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2511.09388
PDF
-
Spider4SSC & S2CLite: A text-to-multi-query-language dataset using lightweight ontology-agnostic SPARQL to Cypher parser
Abstract
We present Spider4SSC dataset and S2CLite parsing tool. S2CLite is a lightweight, ontology-agnostic parser that translates SPARQL queries into Cypher queries, enabling both in-situ and large-scale SPARQL to Cypher translation. Unlike existing solutions, S2CLite is purely rule-based (inspired by traditional programming language compilers) and operates without requiring an RDF graph or external tools. Experiments conducted on the BSBM42 and Spider4SPARQL datasets show that S2CLite significantly reduces query parsing errors, achieving a total parsing accuracy of 77.8% on Spider4SPARQL compared to 44.2% by the state-of-the-art S2CTrans. Furthermore, S2CLite achieved a 96.6\% execution accuracy on the intersecting subset of queries parsed by both parsers, outperforming S2CTrans by 7.3%. We further use S2CLite to parse Spider4SPARQL queries to Cypher and generate Spider4SSC, a unified Text-to-Query language (SQL, SPARQL, Cypher) dataset with 4525 unique questions and 3 equivalent sets of 2581 matching queries (SQL, SPARQL and Cypher). We open-source S2CLite for further development on GitHub (this http URL) and provide the clean Spider4SSC dataset for download.
Abstract (translated)
URL
https://arxiv.org/abs/2511.09354
PDF
-
mmJEE-Eval: A Bilingual Multimodal Benchmark for Evaluating Scientific Reasoning in Vision-Language Models
2025-11-12 13:52:37Arka Mukherjee, Shreya GhoshAbstract
Contemporary vision-language models (VLMs) perform well on existing multimodal reasoning benchmarks (78-85\% accuracy on MMMU, MathVista). Yet, these results fail to sufficiently distinguish true scientific reasoning articulation capabilities from pattern-matching. To address this gap, we introduce \textbf{mmJEE-Eval}, a multimodal bilingual (English and Hindi) benchmark comprising 1,460 questions from India's JEE Advanced examination (2019-2025) spanning pre-college Physics, Chemistry, and Mathematics domains. Our evaluation of 17 state-of-the-art models reveals that while frontier VLMs (GPT-5, Gemini 2.5 Pro/Flash) achieve 77-84\% accuracy on held-out 2025 questions, open-source models plateau at 37-45\% despite scaling to 400B parameters, a significant difference not observed on existing benchmarks. While closed frontiers from Google and OpenAI show high problem-solving accuracies (up to 100\% pass@3 scores), they fully collapse when the reasoning load is increased meta-cognitively (GPT-5 fixes just 5.2\% errors). Systematic ablations show mmJEE-Eval's difficulty stems from complexity and reasoning depth rather than memorization. Effectively, our benchmark segregates superior training and reasoning methodologies where alternatives fail. We publicly release our code and data: this https URL
Abstract (translated)
URL
https://arxiv.org/abs/2511.09339
PDF
-
Enriching Knowledge Distillation with Cross-Modal Teacher Fusion
2025-11-12 12:50:15Amir M. Mansourian, Amir Mohammad Babaei, Shohreh KasaeiAbstract
Multi-teacher knowledge distillation (KD), a more effective technique than traditional single-teacher methods, transfers knowledge from expert teachers to a compact student model using logit or feature matching. However, most existing approaches lack knowledge diversity, as they rely solely on unimodal visual information, overlooking the potential of cross-modal representations. In this work, we explore the use of CLIP's vision-language knowledge as a complementary source of supervision for KD, an area that remains largely underexplored. We propose a simple yet effective framework that fuses the logits and features of a conventional teacher with those from CLIP. By incorporating CLIP's multi-prompt textual guidance, the fused supervision captures both dataset-specific and semantically enriched visual cues. Beyond accuracy, analysis shows that the fused teacher yields more confident and reliable predictions, significantly increasing confident-correct cases while reducing confidently wrong ones. Moreover, fusion with CLIP refines the entire logit distribution, producing semantically meaningful probabilities for non-target classes, thereby improving inter-class consistency and distillation quality. Despite its simplicity, the proposed method, Enriching Knowledge Distillation (RichKD), consistently outperforms most existing baselines across multiple benchmarks and exhibits stronger robustness under distribution shifts and input corruptions.
Abstract (translated)
URL
https://arxiv.org/abs/2511.09286
PDF
-
4KDehazeFlow: Ultra-High-Definition Image Dehazing via Flow Matching
2025-11-12 07:16:52Xingchi Chen, Pu Wang, Xuerui Li, Chaopeng Li, Juxiang Zhou, Jianhou Gan, Dianjie Lu, Guijuan Zhang, Wenqi Ren, Zhuoran ZhengAbstract
Ultra-High-Definition (UHD) image dehazing faces challenges such as limited scene adaptability in prior-based methods and high computational complexity with color distortion in deep learning approaches. To address these issues, we propose 4KDehazeFlow, a novel method based on Flow Matching and the Haze-Aware vector field. This method models the dehazing process as a progressive optimization of continuous vector field flow, providing efficient data-driven adaptive nonlinear color transformation for high-quality dehazing. Specifically, our method has the following advantages: 1) 4KDehazeFlow is a general method compatible with various deep learning networks, without relying on any specific network architecture. 2) We propose a learnable 3D lookup table (LUT) that encodes haze transformation parameters into a compact 3D mapping matrix, enabling efficient inference through precomputed mappings. 3) We utilize a fourth-order Runge-Kutta (RK4) ordinary differential equation (ODE) solver to stably solve the dehazing flow field through an accurate step-by-step iterative method, effectively suppressing artifacts. Extensive experiments show that 4KDehazeFlow exceeds seven state-of-the-art methods. It delivers a 2dB PSNR increase and better performance in dense haze and color fidelity.
Abstract (translated)
URL
https://arxiv.org/abs/2511.09055
PDF
-
RadHARSimulator V2: Video to Doppler Generator
2025-11-12 06:15:53Weicheng GaoAbstract
Radar-based human activity recognition (HAR) still lacks a comprehensive simulation method. Existing software is developed based on models or motion-captured data, resulting in limited flexibility. To address this issue, a simulator that directly generates Doppler spectra from recorded video footage (RadHARSimulator V2) is presented in this paper. Both computer vision and radar modules are included in the simulator. In computer vision module, the real-time model for object detection with global nearest neighbor is first used to detect and track human targets in the video. Then, the high-resolution network is used to estimate two-dimensional poses of the detected human targets. Next, the three-dimensional poses of the detected human targets are obtained by nearest matching method. Finally, smooth temporal three-dimensional pose estimation is achieved through Kalman filtering. In radar module, pose interpolation and smoothing are first achieved through the Savitzky-Golay method. Second, the delay model and the mirror method are used to simulate echoes in both free-space and through-the-wall scenarios. Then, range-time map is generated using pulse compression, moving target indication, and DnCNN. Next, Doppler-time map (DTM) is generated using short-time Fourier transform and DnCNN again. Finally, the ridge features on the DTM are extracted using the maximum local energy method. In addition, a hybrid parallel-serial neural network architecture is proposed for radar-based HAR. Numerical experiments are conducted and analyzed to demonstrate the effectiveness of the designed simulator and the proposed network model. The open-source code of this work can be found in: this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2511.09022
PDF
-
Diffusion Policies with Value-Conditional Optimization for Offline Reinforcement Learning
2025-11-12 02:59:07Yunchang Ma, Tenglong Liu, Yixing Lan, Xin Yin, Changxin Zhang, Xinglong Zhang, Xin XuAbstract
In offline reinforcement learning, value overestimation caused by out-of-distribution (OOD) actions significantly limits policy performance. Recently, diffusion models have been leveraged for their strong distribution-matching capabilities, enforcing conservatism through behavior policy constraints. However, existing methods often apply indiscriminate regularization to redundant actions in low-quality datasets, resulting in excessive conservatism and an imbalance between the expressiveness and efficiency of diffusion modeling. To address these issues, we propose DIffusion policies with Value-conditional Optimization (DIVO), a novel approach that leverages diffusion models to generate high-quality, broadly covered in-distribution state-action samples while facilitating efficient policy improvement. Specifically, DIVO introduces a binary-weighted mechanism that utilizes the advantage values of actions in the offline dataset to guide diffusion model training. This enables a more precise alignment with the dataset's distribution while selectively expanding the boundaries of high-advantage actions. During policy improvement, DIVO dynamically filters high-return-potential actions from the diffusion model, effectively guiding the learned policy toward better performance. This approach achieves a critical balance between conservatism and explorability in offline RL. We evaluate DIVO on the D4RL benchmark and compare it against state-of-the-art baselines. Empirical results demonstrate that DIVO achieves superior performance, delivering significant improvements in average returns across locomotion tasks and outperforming existing methods in the challenging AntMaze domain, where sparse rewards pose a major difficulty.
Abstract (translated)
URL
https://arxiv.org/abs/2511.08922
PDF
-
LLM-Guided Probabilistic Fusion for Label-Efficient Document Layout Analysis
2025-11-12 02:25:58Ibne Farabi Shihab, Sanjeda Akter, Anuj SharmaAbstract
Document layout understanding remains data-intensive despite advances in semi-supervised learning. We present a framework that enhances semi-supervised detection by fusing visual predictions with structural priors from text-pretrained LLMs via principled probabilistic weighting. Given unlabeled documents, an OCR-LLM pipeline infers hierarchical regions which are combined with teacher detector outputs through inverse-variance fusion to generate refined this http URL method demonstrates consistent gains across model scales. With a lightweight SwiftFormer backbone (26M params), we achieve 88.2$\pm$0.3 AP using only 5\% labels on PubLayNet. When applied to document-pretrained LayoutLMv3 (133M params), our fusion framework reaches 89.7$\pm$0.4 AP, surpassing both LayoutLMv3 with standard semi-supervised learning (89.1$\pm$0.4 AP, p=0.02) and matching UDOP~\cite{udop} (89.8 AP) which requires 100M+ pages of multimodal pretraining. This demonstrates that LLM structural priors are complementary to both lightweight and pretrained architectures. Key findings include: (1) learned instance-adaptive gating improves over fixed weights by +0.9 AP with data-dependent PAC bounds correctly predicting convergence; (2) open-source LLMs enable privacy-preserving deployment with minimal loss (Llama-3-70B: 87.1 AP lightweight, 89.4 AP with LayoutLMv3); (3) LLMs provide targeted semantic disambiguation (18.7\% of cases, +3.8 AP gain) beyond simple text this http URL system cost includes \$12 for GPT-4o-mini API or 17 GPU-hours for local Llama-3-70B per 50K pages, amortized across training runs.
Abstract (translated)
URL
https://arxiv.org/abs/2511.08903
PDF
-
Asymmetric Cross-Modal Knowledge Distillation: Bridging Modalities with Weak Semantic Consistency
2025-11-12 02:24:24Riling Wei, Kelu Yao, Chuanguang Yang, Jin Wang, Zhuoyan Gao, Chao LiAbstract
Cross-modal Knowledge Distillation has demonstrated promising performance on paired modalities with strong semantic connections, referred to as Symmetric Cross-modal Knowledge Distillation (SCKD). However, implementing SCKD becomes exceedingly constrained in real-world scenarios due to the limited availability of paired modalities. To this end, we investigate a general and effective knowledge learning concept under weak semantic consistency, dubbed Asymmetric Cross-modal Knowledge Distillation (ACKD), aiming to bridge modalities with limited semantic overlap. Nevertheless, the shift from strong to weak semantic consistency improves flexibility but exacerbates challenges in knowledge transmission costs, which we rigorously verified based on optimal transport theory. To mitigate the issue, we further propose a framework, namely SemBridge, integrating a Student-Friendly Matching module and a Semantic-aware Knowledge Alignment module. The former leverages self-supervised learning to acquire semantic-based knowledge and provide personalized instruction for each student sample by dynamically selecting the relevant teacher samples. The latter seeks the optimal transport path by employing Lagrangian optimization. To facilitate the research, we curate a benchmark dataset derived from two modalities, namely Multi-Spectral (MS) and asymmetric RGB images, tailored for remote sensing scene classification. Comprehensive experiments exhibit that our framework achieves state-of-the-art performance compared with 7 existing approaches on 6 different model architectures across various datasets.
Abstract (translated)
URL
https://arxiv.org/abs/2511.08901
PDF
-
Compression then Matching: An Efficient Pre-training Paradigm for Multimodal Embedding
2025-11-11 17:23:02Da Li, Yuxiao Luo, Keping Bi, Jiafeng Guo, Wei Yuan, Biao Yang, Yan Wang, Fan Yang, Tingting Gao, Guorui ZhouAbstract
Vision-language models advance multimodal representation learning by acquiring transferable semantic embeddings, thereby substantially enhancing performance across a range of vision-language tasks, including cross-modal retrieval, clustering, and classification. An effective embedding is expected to comprehensively preserve the semantic content of the input while simultaneously emphasizing features that are discriminative for downstream tasks. Recent approaches demonstrate that VLMs can be adapted into competitive embedding models via large-scale contrastive learning, enabling the simultaneous optimization of two complementary objectives. We argue that the two aforementioned objectives can be decoupled: a comprehensive understanding of the input facilitates the embedding model in achieving superior performance in downstream tasks via contrastive learning. In this paper, we propose CoMa, a compressed pre-training phase, which serves as a warm-up stage for contrastive learning. Experiments demonstrate that with only a small amount of pre-training data, we can transform a VLM into a competitive embedding model. CoMa achieves new state-of-the-art results among VLMs of comparable size on the MMEB, realizing optimization in both efficiency and effectiveness.
Abstract (translated)
URL
https://arxiv.org/abs/2511.08480
PDF
-
DPRM: A Dual Implicit Process Reward Model in Multi-Hop Question Answering
2025-11-11 15:41:22Xinyi Wang, Yiping Song, Zhiliang Tian, Bo Liu, Tingjin Luo, Minlie HuangAbstract
In multi-hop question answering (MHQA) tasks, Chain of Thought (CoT) improves the quality of generation by guiding large language models (LLMs) through multi-step reasoning, and Knowledge Graphs (KGs) reduce hallucinations via semantic matching. Outcome Reward Models (ORMs) provide feedback after generating the final answers but fail to evaluate the process for multi-step reasoning. Traditional Process Reward Models (PRMs) evaluate the reasoning process but require costly human annotations or rollout generation. While implicit PRM is trained only with outcome signals and derives step rewards through reward parameterization without explicit annotations, it is more suitable for multi-step reasoning in MHQA tasks. However, existing implicit PRM has only been explored for plain text scenarios. When adapting to MHQA tasks, it cannot handle the graph structure constraints in KGs and capture the potential inconsistency between CoT and KG paths. To address these limitations, we propose the DPRM (Dual Implicit Process Reward Model). It trains two implicit PRMs for CoT and KG reasoning in MHQA tasks. Both PRMs, namely KG-PRM and CoT-PRM, derive step-level rewards from outcome signals via reward parameterization without additional explicit annotations. Among them, KG-PRM uses preference pairs to learn structural constraints from KGs. DPRM further introduces a consistency constraint between CoT and KG reasoning steps, making the two PRMs mutually verify and collaboratively optimize the reasoning paths. We also provide a theoretical demonstration of the derivation of process rewards. Experimental results show that our method outperforms 13 baselines on multiple datasets with up to 16.6% improvement on Hit@1.
Abstract (translated)
URL
https://arxiv.org/abs/2511.08364
PDF
-
Extreme Model Compression with Structured Sparsity at Low Precision
2025-11-11 15:37:55Dan Liu, Nikita Dvornik, Xue LiuAbstract
Deep neural networks (DNNs) are used in many applications, but their large size and high computational cost make them hard to run on devices with limited resources. Two widely used techniques to address this challenge are weight quantization, which lowers the precision of all weights, and structured sparsity, which removes unimportant weights while retaining the important ones at full precision. Although both are effective individually, they are typically studied in isolation due to their compounded negative impact on model accuracy when combined. In this work, we introduce SLOPE Structured Sparsity at Low Precision), a unified framework, to effectively combine structured sparsity and low-bit quantization in a principled way. We show that naively combining sparsity and quantization severely harms performance due to the compounded impact of both techniques. To address this, we propose a training-time regularization strategy that minimizes the discrepancy between full-precision weights and their sparse, quantized counterparts by promoting angular alignment rather than direct matching. On ResNet-18, SLOPE achieves $\sim20\times$ model size reduction while retaining $\sim$99% of the original accuracy. It consistently outperforms state-of-the-art quantization and structured sparsity methods across classification, detection, and segmentation tasks on models such as ResNet-18, ViT-Small, and Mask R-CNN.
Abstract (translated)
URL
https://arxiv.org/abs/2511.08360
PDF
-
Work-in-Progress: Function-as-Subtask API Replacing Publish/Subscribe for OS-Native DAG Scheduling
2025-11-11 14:31:22Takahiro Ishikawa-Aso, Atsushi Yano, Yutaro Kobayashi, Takumi Jin, Yuuki Takano, Shinpei KatoAbstract
The Directed Acyclic Graph (DAG) task model for real-time scheduling finds its primary practical target in Robot Operating System 2 (ROS 2). However, ROS 2's publish/subscribe API leaves DAG precedence constraints unenforced: a callback may publish mid-execution, and multi-input callbacks let developers choose topic-matching policies. Thus preserving DAG semantics relies on conventions; once violated, the model collapses. We propose the Function-as-Subtask (FasS) API, which expresses each subtask as a function whose arguments/return values are the subtask's incoming/outgoing edges. By minimizing description freedom, DAG semantics is guaranteed at the API rather than by programmer discipline. We implement a DAG-native scheduler using FasS on a Rust-based experimental kernel and evaluate its semantic fidelity, and we outline design guidelines for applying FasS to Linux Linux sched_ext.
Abstract (translated)
URL
https://arxiv.org/abs/2511.08297
PDF
-
ImagebindDC: Compressing Multi-modal Data with Imagebind-based Condensation
2025-11-11 13:55:46Yue Min, Shaobo Wang, Jiaze Li, Tianle Niu, Junxin Fan, Yongliang Miao, Lijin Yang, Linfeng ZhangAbstract
Data condensation techniques aim to synthesize a compact dataset from a larger one to enable efficient model training, yet while successful in unimodal settings, they often fail in multimodal scenarios where preserving intricate inter-modal dependencies is crucial. To address this, we introduce ImageBindDC, a novel data condensation framework operating within the unified feature space of ImageBind. Our approach moves beyond conventional distribution-matching by employing a powerful Characteristic Function (CF) loss, which operates in the Fourier domain to facilitate a more precise statistical alignment via exact infinite moment matching. We design our objective to enforce three critical levels of distributional consistency: (i) uni-modal alignment, which matches the statistical properties of synthetic and real data within each modality; (ii) cross-modal alignment, which preserves pairwise semantics by matching the distributions of hybrid real-synthetic data pairs; and (iii) joint-modal alignment, which captures the complete multivariate data structure by aligning the joint distribution of real data pairs with their synthetic counterparts. Extensive experiments highlight the effectiveness of ImageBindDC: on the NYU-v2 dataset, a model trained on just 5 condensed datapoints per class achieves lossless performance comparable to one trained on the full dataset, achieving a new state-of-the-art with an 8.2\% absolute improvement over the previous best method and more than 4$\times$ less condensation time.
Abstract (translated)
URL
https://arxiv.org/abs/2511.08263
PDF
-
LatentPrintFormer: A Hybrid CNN-Transformer with Spatial Attention for Latent Fingerprint identification
2025-11-11 11:20:49Arnab Maity, Manasa, Pavan Kumar C, Raghavendra RamachandraAbstract
Latent fingerprint identification remains a challenging task due to low image quality, background noise, and partial impressions. In this work, we propose a novel identification approach called LatentPrintFormer. The proposed model integrates a CNN backbone (EfficientNet-B0) and a Transformer backbone (Swin Tiny) to extract both local and global features from latent fingerprints. A spatial attention module is employed to emphasize high-quality ridge regions while suppressing background noise. The extracted features are fused and projected into a unified 512-dimensional embedding, and matching is performed using cosine similarity in a closed-set identification setting. Extensive experiments on two publicly available datasets demonstrate that LatentPrintFormer consistently outperforms three state-of-the-art latent fingerprint recognition techniques, achieving higher identification rates across Rank-10.
Abstract (translated)
URL
https://arxiv.org/abs/2511.08119
PDF
-
Gateways to Tractability for Satisfiability in Pearl's Causal Hierarchy
Abstract
Pearl's Causal Hierarchy (PCH) is a central framework for reasoning about probabilistic, interventional, and counterfactual statements, yet the satisfiability problem for PCH formulas is computationally intractable in almost all classical settings. We revisit this challenge through the lens of parameterized complexity and identify the first gateways to tractability. Our results include fixed-parameter and XP-algorithms for satisfiability in key probabilistic and counterfactual fragments, using parameters such as primal treewidth and the number of variables, together with matching hardness results that map the limits of tractability. Technically, we depart from the dynamic programming paradigm typically employed for treewidth-based algorithms and instead exploit structural characterizations of well-formed causal models, providing a new algorithmic toolkit for causal reasoning.
Abstract (translated)
URL
https://arxiv.org/abs/2511.08091
PDF
-
Taming Identity Consistency and Prompt Diversity in Diffusion Models via Latent Concatenation and Masked Conditional Flow Matching
2025-11-11 10:00:32Aditi Singhania, Arushi Jain, Krutik Malani, Riddhi Dhawan, Souymodip Chakraborty, Vineet Batra, Ankit PhogatAbstract
Subject-driven image generation aims to synthesize novel depictions of a specific subject across diverse contexts while preserving its core identity features. Achieving both strong identity consistency and high prompt diversity presents a fundamental trade-off. We propose a LoRA fine-tuned diffusion model employing a latent concatenation strategy, which jointly processes reference and target images, combined with a masked Conditional Flow Matching (CFM) objective. This approach enables robust identity preservation without architectural modifications. To facilitate large-scale training, we introduce a two-stage Distilled Data Curation Framework: the first stage leverages data restoration and VLM-based filtering to create a compact, high-quality seed dataset from diverse sources; the second stage utilizes these curated examples for parameter-efficient fine-tuning, thus scaling the generation capability across various subjects and contexts. Finally, for filtering and quality assessment, we present CHARIS, a fine-grained evaluation framework that performs attribute-level comparisons along five key axes: identity consistency, prompt adherence, region-wise color fidelity, visual quality, and transformation diversity.
Abstract (translated)
URL
https://arxiv.org/abs/2511.08061
PDF
-
Numerical Sensitivity and Robustness: Exploring the Flaws of Mathematical Reasoning in Large Language Models
2025-11-11 09:25:18Zhishen Sun, Guang Dai, Ivor Tsang, Haishan YeAbstract
LLMs have made significant progress in the field of mathematical reasoning, but whether they have true the mathematical understanding ability is still controversial. To explore this issue, we propose a new perturbation framework to evaluate LLMs' reasoning ability in complex environments by injecting additional semantically irrelevant perturbation sentences and gradually increasing the perturbation intensity. At the same time, we use an additional perturbation method: core questioning instruction missing, to further analyze the LLMs' problem-solving mechanism. The experimental results show that LLMs perform stably when facing perturbation sentences without numbers, but there is also a robustness boundary. As the perturbation intensity increases, the performance exhibits varying degrees of decline; when facing perturbation sentences with numbers, the performance decreases more significantly, most open source models with smaller parameters decrease by nearly or even more than 10%, and further increasing with the enhancement of perturbation intensity, with the maximum decrease reaching 51.55%. Even the most advanced commercial LLMs have seen a 3%-10% performance drop. By analyzing the reasoning process of LLMs in detail, We find that models are more sensitive to perturbations with numerical information and are more likely to give incorrect answers when disturbed by irrelevant numerical information. The higher the perturbation intensity, the more obvious these defects are. At the same time, in the absence of core questioning instruction, models can still maintain an accuracy of 20%-40%, indicating that LLMs may rely on memory templates or pattern matching to complete the task, rather than logical reasoning. In general, our work reveals the shortcomings and limitations of current LLMs in their reasoning capabilities, which is of great significance for the further development of LLMs.
Abstract (translated)
URL
https://arxiv.org/abs/2511.08022
PDF
-
State of the Art in Text Classification for South Slavic Languages: Fine-Tuning or Prompting?
2025-11-11 08:54:15Taja Kuzman Punger\v{s}ek, Peter Rupnik, Ivan Porupski, Vuk Dini\'c, Nikola Ljube\v{s}i\'cAbstract
Until recently, fine-tuned BERT-like models provided state-of-the-art performance on text classification tasks. With the rise of instruction-tuned decoder-only models, commonly known as large language models (LLMs), the field has increasingly moved toward zero-shot and few-shot prompting. However, the performance of LLMs on text classification, particularly on less-resourced languages, remains under-explored. In this paper, we evaluate the performance of current language models on text classification tasks across several South Slavic languages. We compare openly available fine-tuned BERT-like models with a selection of open-source and closed-source LLMs across three tasks in three domains: sentiment classification in parliamentary speeches, topic classification in news articles and parliamentary speeches, and genre identification in web texts. Our results show that LLMs demonstrate strong zero-shot performance, often matching or surpassing fine-tuned BERT-like models. Moreover, when used in a zero-shot setup, LLMs perform comparably in South Slavic languages and English. However, we also point out key drawbacks of LLMs, including less predictable outputs, significantly slower inference, and higher computational costs. Due to these limitations, fine-tuned BERT-like models remain a more practical choice for large-scale automatic text annotation.
Abstract (translated)
URL
https://arxiv.org/abs/2511.07989
PDF
