Recent advances in text-to-speech technology have enabled highly realistic voice generation, fueling audio-based deepfake attacks such as fraud and impersonation. While audio anti-spoofing systems are critical for detecting such threats, prior research has predominantly focused on acoustic-level perturbations, leaving **the impact of linguistic variation largely unexplored**. In this paper, we investigate the linguistic sensitivity of both open-source and commercial anti-spoofing detectors by introducing **TAPAS** (Transcript-to-Audio Perturbation Anti-Spoofing), a novel framework for transcript-level adversarial attacks. Our extensive evaluation shows that even minor linguistic perturbations can significantly degrade detection accuracy: attack success rates exceed **60%** on several open-source detector–voice pairs, and the accuracy of one commercial detector drops from **100%** on synthetic audio to just **32%**. Through a comprehensive feature attribution analysis, we find that linguistic complexity and model-level audio embedding similarity are key factors contributing to detector vulnerabilities. To illustrate the real-world risks, we replicate a recent Brad Pitt audio deepfake scam and demonstrate that TAPAS can bypass commercial detectors. These findings underscore the **need to move beyond purely acoustic defenses** and incorporate linguistic variation into the design of robust anti-spoofing systems. Our source code is available at https://github.com/nqbinh17/audio_linguistic_adversarial.
论文检索
输入标题、作者或关键词，从 7,591 篇学术成果中精准定位
To See a World in a Spark of Neuron: Disentangling Multi-Task Interference for Training-Free Model Merging
PDF ↗Fine-tuning pre-trained models on targeted datasets enhances task-specific performance but often comes at the expense of generalization. Model merging techniques, which integrate multiple fine-tuned models into a single multi-task model through task arithmetic, offer a promising solution. However, task interference remains a fundamental challenge, leading to performance degradation and suboptimal merged models. Existing approaches largely overlooked the fundamental roles of neurons, their connectivity, and activation, resulting in a merging process and a merged model that does not consider how neurons relay and process information. In this work, we present the first study that relies on neuronal mechanisms for model merging. Specifically, we decomposed task-specific representations into two complementary neuronal subspaces that regulate input sensitivity and task adaptability. Leveraging this decomposition, we introduced NeuroMerging, a novel merging framework developed to mitigate task interference within neuronal subspaces, enabling training-free model fusion across diverse tasks. Through extensive experiments, we demonstrated that NeuroMerging achieved superior performance compared to existing methods on multi-task benchmarks across both natural language and vision domains. Our findings highlighted the importance of aligning neuronal mechanisms in model merging, offering new insights into mitigating task interference and improving knowledge fusion. Our project is available at [this http URL](https://ZzzitaoFang.github.io/projects/NeuroMerging/).
Large language models (LLMs) demonstrate remarkable capabilities in understanding complex tasks and have achieved commendable performance in graph-related tasks, such as node classification, link prediction, and subgraph classification. These tasks primarily depend on the local reasoning capabilities of the graph structure. However, research has yet to address the graph partitioning task that requires global perception abilities. Our preliminary findings reveal that vanilla LLMs can only handle graph partitioning on extremely small-scale graphs. To overcome this limitation, we propose a three-phase pipeline to empower LLMs for large-scale graph partitioning: coarsening, reasoning, and refining. The coarsening phase reduces graph complexity. The reasoning phase captures both global and local patterns to generate a coarse partition. The refining phase ensures topological consistency by projecting the coarse-grained partitioning results back to the original graph structure. Extensive experiments demonstrate that our framework enables LLMs to perform graph partitioning across varying graph scales, validating both the effectiveness of LLMs for partitioning tasks and the practical utility of our proposed methodology.
Large language models (LLMs) are increasingly recognized as powerful tools for scientific discovery, particularly in molecular science. A fundamental requirement for these models is the ability to accurately understand molecular structures, commonly encoded in the SMILES representation. However, current LLMs struggle to interpret SMILES, even failing to carry out basic tasks such as counting molecular rings. To address this limitation, we introduce CLEANMOL, a novel framework that formulates SMILES parsing into a suite of clean and deterministic tasks explicitly designed to promote graph-level molecular comprehension. These tasks span from subgraph matching to global graph matching, providing structured supervision aligned with molecular structural properties. We construct a molecular pretraining dataset with adaptive difficulty scoring and pre-train open-source LLMs on these tasks. Our results show that CLEANMOL not only enhances structural comprehension but also achieves the best or competes with the baseline on the Mol-Instructions benchmark.
TreeReview: A Dynamic Tree of Questions Framework for Deep and Efficient LLM-based Scientific Peer Review
PDF ↗While Large Language Models (LLMs) have shown significant potential in assisting peer review, current methods often struggle to generate thorough and insightful reviews while maintaining efficiency. In this paper, we propose TreeReview, a novel framework that models paper review as a hierarchical and bidirectional question-answering process. TreeReview first constructs a tree of review questions by recursively decomposing high-level questions into fine-grained sub-questions and then resolves the question tree by iteratively aggregating answers from leaf to root to get the final review. Crucially, we incorporate a dynamic question expansion mechanism to enable deeper probing by generating follow-up questions when needed. We construct a benchmark derived from ICLR and NeurIPS venues to evaluate our method on full review generation and actionable feedback comments generation tasks. Experimental results of both LLM-based and human evaluation show that TreeReview outperforms strong baselines in providing comprehensive, in-depth, and expert-aligned review feedback, while reducing LLM token usage by up to 80% compared to computationally intensive approaches.
Large Language Models are commonly judged by their scores on standard benchmarks, yet such scores often overstate real capability since they mask the mix of skills a task actually demands. For example, ARC is assumed to test reasoning, while HellaSwag is designed to evaluate commonsense. However, we lack a systematic way to verify if these benchmarks actually measure these labels. We introduce **BENCHMARK PROFILING**, a diagnostic framework that decomposes benchmark performance into ten cognitively grounded abilities. The method combines gradient-based importance scoring with targeted parameter ablation to compute an Ability Impact Score (AIS) that quantifies how much each ability contributes to a model’s success on a given benchmark. Profiling three instruction-tuned models across ten widely used benchmarks yields four key findings: (i) most benchmarks draw on several abilities rather than one, (ii) datasets with similar labels rely on distinct ability mixtures, (iii) code-generation benchmarks reward broad, multi-skill improvement and thus show only modest gains from narrow domain-specific fine-tuning, and (iv) abilities irrelevant to the task could negatively affect performance. **BENCHMARK PROFILING** therefore explains why performance gains do not always translate into user-perceived competence and offer a transparent tool for benchmark audit and model interpretability.
Context-Aware Hierarchical Taxonomy Generation for Scientific Papers via LLM-Guided Multi-Aspect Clustering
PDF ↗The rapid growth of scientific literature demands efficient methods to organize and synthesize research findings. Existing taxonomy construction methods, leveraging unsupervised clustering or direct prompting of large language models (LLMs), often lack coherence and granularity. We propose a novel context-aware hierarchical taxonomy generation framework that integrates LLM-guided multi-aspect encoding with dynamic clustering. Our method leverages LLMs to identify key aspects of each paper (e.g., methodology, dataset, evaluation) and generates aspect-specific paper summaries, which are then encoded and clustered along each aspect to form a coherent hierarchy. In addition, we introduce a new evaluation benchmark of 156 expert-crafted taxonomies encompassing 11.6k papers, providing the first naturally annotated dataset for this task. Experimental results demonstrate that our method significantly outperforms prior approaches, achieving state-of-the-art performance in taxonomy coherence, granularity, and interpretability.
Methods for learning taxonomies from data have been widely studied. We study a specific version of this task, called commonality identification, where only the set of entities is given and we need to find meaningful ways to group those entities. While LLMs should intuitively excel at this task, it is difficult to directly use such models in large domains. In this paper, we instead use LLMs to describe the different properties that are satisfied by each of the entities individually. We then use pre-trained embeddings to cluster these properties, and finally group entities that have properties which belong to the same cluster. To achieve good results, it is paramount that the properties predicted by the LLM are sufficiently diverse. We find that this diversity can be improved by prompting the LLM to structure the predicted properties into different facets of knowledge.
Recent advances in LLM-based recommendation have shown promise, yet their cross-domain generalization is hindered by a fundamental mismatch between language-centric pretraining and the recommendation task. Existing methods, relying on language-level knowledge, fail to capture dynamic, item-level user interests across domains. To bridge this gap, we propose RecBase, a domain-agnostic foundational model pretrained with a recommendation-oriented objective. RecBase leverages a large-scale, heterogeneous, cross-domain corpus with unified textual representations and feature mappings to enhance cross-domain generalization. To further align item semantics across domains, we introduce a unified item tokenizer that encodes items into hierarchical concept identifiers, enabling structured representation and efficient vocabulary sharing. The model is trained using an autoregressive objective to capture complex item-level sequential patterns. On eight real-world datasets, our 1.5B-parameter model matches or surpasses the performance of LLM baselines up to 7B parameters in zero-shot and cross-domain recommendation tasks.
Conceptual operationalizations of empathy in NLP are varied, with some having specific behaviors and properties, while others are more abstract. How these variations relate to one another and capture properties of empathy observable in text remains unclear. To provide insight into this, we analyze the transfer performance of empathy models adapted to empathy tasks with different theoretical groundings. We study (1) the dimensionality of empathy definitions, (2) the correspondence between the defined dimensions and measured/observed properties, and (3) the conduciveness of the data to represent them, finding they have a significant impact to performance compared to other transfer setting features. Characterizing the theoretical grounding of empathy tasks as direct, abstract, or adjacent further indicates that tasks that directly predict specified empathy components have higher transferability. Our work provides empirical evidence for the need for precise and multidimensional empathy operationalizations.
OntologyRAG-Q: Resource Development and Benchmarking for Retrieval-Augmented Question Answering in Qur’anic Tafsir
PDF ↗This paper introduces essential resources for Qur’anic studies: an annotated Tafsir ontology, a dataset of approximately 4,200 question-answer pairs, and a collection of 15 structured Tafsir books available in two formats. We present a comprehensive framework for handling sensitive Qur’anic Tafsir data that spans the entire pipeline from dataset construction through evaluation and error analysis. Our work establishes new benchmarks for retrieval and question-answering tasks on Qur’anic content, comparing performance across state-of-the-art embedding models and large language models (LLMs).We introduce OntologyRAG-Q, a novel retrieval-augmented generation approach featuring our custom Ayat-Ontology chunking method that segments Tafsir content at the verse level using ontology-driven structure. Benchmarking reveals strong performance across various LLMs, with GPT-4 achieving the highest results, followed closely by ALLaM. Expert evaluations show our system achieves 69.52% accuracy and 74.36% correctness overall, though multi-hop and context-dependent questions remain challenging. Our analysis demonstrates that answer position within documents significantly impacts retrieval performance, and among the evaluation metrics tested, BERT-recall and BERT-F1 correlate most strongly with expert assessments. The resources developed in this study are publicly available at https://github.com/sazani/OntologyRAG-Q.git.
Automated Knowledge Graph Construction using Large Language Models and Sentence Complexity Modelling
PDF ↗We introduce CoDe-KG, an open-source, end-to-end pipeline for extracting sentence-level knowledge graphs by combining robust coreference resolution with syntactic sentence decomposition. Using our model, we contribute a dataset of over 150 000 knowledge triples, which is open source. We also contribute a training corpus of 7248 rows for sentence complexity, 200 rows of gold human annotations for coreference resolution using lung-cancer abstracts from PubMed, 900 rows of gold human annotations for sentence conversion policies from sentences in the abstract, and 398 triples of gold human annotations. We systematically select optimal prompt-model pairs across five complexity categories, showing that hybrid chain-of-thought and few-shot prompting yields up to 99.8% exact-match accuracy on sentence simplification. On relation extraction (RE), our pipeline achieves 65.8% macro-F1 on REBEL, an 8-point gain over the prior state of the art, and 75.7% micro-F1 on WebNLG2, while matching or exceeding performance on Wiki-NRE and CaRB. Ablation studies demonstrate that integrating coreference and decomposition increases recall on rare relations by over 20%
Multi-hop complex reasoning over incomplete knowledge graphs (KGs) has been extensively studied, but research on numerical knowledge graphs (NKGs) remains relatively limited. Recent approaches focus on separately encoding entities and numerical values, using neural networks to process query encodings for reasoning. However, in complex multi-hop reasoning tasks, numerical values are not merely symbols, and they carry specific semantics and logical relationships that must be accurately represented. The CNR-NST framework can perform binary operations on numerical attributes in NKGs, enabling it to infer new numerical attributes from existing knowledge. Our approach effectively handles up to 102 types of complex numerical reasoning queries. On three public datasets, CNR-NST demonstrates SOTA performance in complex numerical queries, achieving an average improvement of over 40% compared to existing methods. Notably, this work expands the query types for complex multi-hop numerical reasoning and introduces a new evaluation metric for numerical answers, which has been validated through comprehensive experiments.
Pierce the Mists, Greet the Sky: Decipher Knowledge Overshadowing via Knowledge Circuit Analysis
PDF ↗Large Language Models (LLMs), despite their remarkable capabilities, are hampered by hallucinations. A particularly challenging variant, knowledge overshadowing, occurs when one piece of activated knowledge inadvertently masks another relevant piece, leading to erroneous outputs even with high-quality training data. Current understanding of overshadowing is largely confined to inference-time observations, lacking deep insights into its origins and internal mechanisms during model training. Therefore, we introduce **PhantomCircuit, a novel framework designed to comprehensively analyze and detect knowledge overshadowing.** By innovatively employing knowledge circuit analysis, PhantomCircuit dissects the function of key components in the circuit and how the attention pattern dynamics contribute to the overshadowing phenomenon and its evolution throughout the training process. Extensive experiments demonstrate PhantomCircuit’s effectiveness in identifying such instances, offering novel insights into this elusive hallucination and providing the research community with a new methodological lens for its potential mitigation. Our code can be found in https://github.com/halfmorepiece/PhantomCircuit.
To confront the ever-evolving safety risks of LLMs, automated jailbreak attacks have proven effective for proactively identifying security vulnerabilities at scale. Existing approaches, including GCG and AutoDAN, modify adversarial prompts to induce LLMs to generate responses that strictly follow a fixed affirmative template. However, we observed that the reliance on the rigid output template is ineffective for certain malicious requests, leading to suboptimal jailbreak performance. In this work, we aim to develop a method that is universally effective across all hostile requests. To achieve this, we explore LLMs’ intrinsic safety mechanism: a refusal stance towards the adversarial prompt is formed in a confined region and ultimately leads to a rejective response. In light of this, we propose Stance Manipulation (SM), a novel automated jailbreak approach that generates jailbreak prompts to suppress the refusal stance and induce affirmative responses. Our experiments across four mainstream open-source LLMs demonstrate the superiority of SM’s performance. Under commenly used setting, SM achieves success rates over 77.1% across all models on Advbench. Specifically, for Llama-2-7b-chat, SM outperforms the best baseline by 25.4%. In further experiments with extended iterations in a speedup setup, SM achieves over 92.2% attack success rate across all models. Our code is publicly available at https://github.com/Zed630/Stance-Manipulation.
In recent years, the field of knowledge graph completion has focused on increasingly sophisticated models, which perform well on link prediction tasks, but are less scalable than earlier methods and are not suitable for learning entity embeddings. As a result, shallow models such as TransE and ComplEx remain the most popular choice in many settings. However, the strengths and limitations of such models remain poorly understood. In this paper, we present a unifying framework and systematically analyze a number of variants and extensions of existing shallow models, empirically showing that MuRE and its extension, ExpressivE, are highly competitive. Motivated by the strong empirical results of MuRE, we also theoretically analyze the expressivity of its associated scoring function, surprisingly finding that it can capture the same class of rule bases as state-of-the-art region-based embedding models.
As online news consumption grows, personalized recommendation systems have become integral to digital journalism. However, these systems risk reinforcing filter bubbles and political polarization by failing to incorporate diverse perspectives. Stance detection—identifying a text’s position on a target—can help mitigate this by enabling viewpoint-aware recommendations and data-driven analyses of media bias. Yet, existing stance detection research remains largely limited to short texts and high-resource languages. To address these gaps, we introduce K-News-Stance, the first Korean dataset for article-level stance detection, comprising 2,000 news articles with article-level and 21,650 segment-level stance annotations across 47 societal issues. We also propose JoA-ICL, a Journalism-guided Agentic In-Context Learning framework that employs a language model agent to predict the stances of key structural segments (e.g., leads, quotes), which are then aggregated to infer the overall article stance. Experiments showed that JoA-ICL outperforms existing stance detection methods, highlighting the benefits of segment-level agency in capturing the overall position of long-form news articles. Two case studies further demonstrate its broader utility in promoting viewpoint diversity in news recommendations and uncovering patterns of media bias.
Implicit assumptions and priors are often necessary in text-to-image generation tasks, especially when textual prompts lack sufficient context. However, these assumptions can sometimes reflect societal biases, low variance, or outdated concepts in the training data. We present Embedding-only Editing (EmbEdit), a method designed to efficiently edit implicit assumptions and priors in the text-to-image model without affecting unrelated objects or degrading overall performance. Given a “source” prompt (e.g., “nurse”) that elicits an assumption (e.g., a female nurse) and a “destination” prompt or distribution (e.g. equal gender chance), EmbEdit only fine-tunes the word token embedding (WTE) of the target object (i.e. token “nurse”’s WTE). Our method prevents unintended effects on other objects in the model’s knowledge base, as the WTEs for unrelated objects and the model weights remain unchanged. Further, our method can be applied to any text-to-image model with a text encoder. It is highly efficient, modifying only 768, 2048, and 4864 parameters for Stable Diffusion 1.4, Stable Diffusion XL, and FLUX, respectively, matching each model’s WTE dimension. Additionally, changes could be easily reversed by restoring the original WTE layers. The results show that EmbEdit outperforms previous methods in various models, tasks, and editing scenarios (both single and sequential multiple edits), achieving at least a 6.01% improvement (from 87.17% to 93.18%).
Open benchmarks are essential for evaluating and advancing large language models, offering reproducibility and transparency. However, their accessibility makes them likely targets of test set contamination. In this work, we introduce **DyePack**, a framework that leverages backdoor attacks to identify models that used benchmark test sets during training, **without requiring access to the loss, logits, or any internal details of the model.** Like how banks mix dye packs with their money to mark robbers, DyePack mixes backdoor samples with the test data to flag models that trained on it. We propose a principled design incorporating multiple backdoors with stochastic targets, **enabling exact false positive rate (FPR) computation when flagging every model.** This provably prevents false accusations while providing strong evidence for every detected case of contamination. We evaluate DyePack on five models across three datasets, covering both multiple-choice and open-ended generation tasks. For multiple-choice questions, it successfully detects all contaminated models with guaranteed FPRs as low as 0.000073% on MMLU-Pro and 0.000017% on Big-Bench-Hard using eight backdoors. For open-ended generation tasks, it generalizes well and identifies all contaminated models on Alpaca with a guaranteed false positive rate of just 0.127% using six backdoors.
Human mobility prediction is vital for urban services, but often fails to account for abrupt changes from external events. Existing spatiotemporal models struggle to leverage textual descriptions detailing these events. We propose SeMob, an LLM-powered semantic synthesis pipeline for dynamic mobility prediction. Specifically, SeMob employs a multi-agent framework where LLM-based agents automatically extract and reason about spatiotemporally related text from complex online texts. Fine-grained relevant contexts are then incorporated with spatiotemporal data through our proposed innovative progressive fusion architecture. The rich pre-trained event prior contributes enriched insights about event-driven prediction, and hence results in a more aligned forecasting model. Evaluated on a dataset constructed through our pipeline, SeMob achieves maximal reductions of 13.92% in MAE and 11.12% in RMSE compared to the spatiotemporal model. Notably, the framework exhibits pronounced superiority especially within spatiotemporal regions close to an event’s location and time of occurrence.