Large Language Models (LLMs) suffer from hallucinations and outdated knowledge due to their reliance on static training data. Retrieval-Augmented Generation (RAG) mitigates these issues by integrating external dynamic information for improved factual grounding. With advances in multimodal learning, Multimodal RAG extends this approach by incorporating multiple modalities such as text, images, audio, and video to enhance the generated outputs. However, cross-modal alignment and reasoning introduce unique challenges beyond those in unimodal RAG. This survey offers a structured and comprehensive analysis of Multimodal RAG systems, covering datasets, benchmarks, metrics, evaluation, methodologies, and innovations in retrieval, fusion, augmentation, and generation. We review training strategies, robustness enhancements, loss functions, and agent-based approaches, while also exploring the diverse Multimodal RAG scenarios. In addition, we outline open challenges and future directions to guide research in this evolving field. This survey lays the foundation for developing more capable and reliable AI systems that effectively leverage multimodal dynamic external knowledge bases. All resources are publicly available at https://github.com/llm-lab-org/Multimodal-RAG-Survey.
论文检索
输入标题、作者或关键词，从 2,072 篇学术成果中精准定位
Large Language Models (LLMs) store and retrieve vast amounts of factual knowledge acquired during pre-training. Prior research has localized and identified mechanisms behind knowledge recall; however, it has only focused on English monolingual models. The question of how these mechanisms generalize to non-English languages and multilingual LLMs remains unexplored. In this paper, we address this gap by conducting a comprehensive analysis of three multilingual LLMs. First, we show that previously identified recall mechanisms in English largely apply to multilingual contexts, with nuances based on language and architecture. Next, through patching intermediate representations, we localize the role of language during recall, finding that subject enrichment is language-independent, while object extraction is language-dependent. Additionally, we discover that the last token representation acts as a Function Vector (FV), encoding both the language of the query and the content to be extracted from the subject. Furthermore, in decoder-only LLMs, FVs compose these two pieces of information in two separate stages. These insights reveal unique mechanisms in multilingual LLMs for recalling information, highlighting the need for new methodologies—such as knowledge evaluation, fact editing, and knowledge acquisition—that are specifically tailored for multilingual LLMs.
The integration of sophisticated Vision-Language Models (VLMs) in vehicular systems is revolutionizing vehicle interaction and safety, performing tasks such as Visual Question Answering (VQA). However, a critical gap persists due to the lack of a comprehensive benchmark for multimodal VQA models in vehicular scenarios. To address this, we propose IntelliCockpitBench, a benchmark that encompasses diverse automotive scenarios. It includes images from front, side, and rear cameras, various road types, weather conditions, and interior views, integrating data from both moving and stationary states. Notably, all images and queries in the benchmark are verified for high levels of authenticity, ensuring the data accurately reflects real-world conditions. A sophisticated scoring methodology combining human and model-generated assessments enhances reliability and consistency. Our contributions include a diverse and authentic dataset for automotive VQA and a robust evaluation metric aligning human and machine assessments. All code and data can be found at https://github.com/Lane315/IntelliCockpitBench.
Recent breakthroughs in singing voice synthesis (SVS) have heightened the demand for high-quality annotated datasets, yet manual annotation remains prohibitively labor-intensive and resource-intensive. Existing automatic singing annotation (ASA) methods, however, primarily tackle isolated aspects of the annotation pipeline. To address this fundamental challenge, we present STARS, which is, to our knowledge, the first unified framework that simultaneously addresses singing transcription, alignment, and refined style annotation. Our framework delivers comprehensive multi-level annotations encompassing: (1) precise phoneme-audio alignment, (2) robust note transcription and temporal localization, (3) expressive vocal technique identification, and (4) global stylistic characterization including emotion and pace. The proposed architecture employs hierarchical acoustic feature processing across frame, word, phoneme, note, and sentence levels. The novel non-autoregressive local acoustic encoders enable structured hierarchical representation learning. Experimental validation confirms the framework’s superior performance across multiple evaluation dimensions compared to existing annotation approaches. Furthermore, applications in SVS training demonstrate that models utilizing STARS-annotated data achieve significantly enhanced perceptual naturalness and precise style control. This work not only overcomes critical scalability challenges in the creation of singing datasets but also pioneers new methodologies for controllable singing voice synthesis.
We perform a critical examination of the scientific methodology behind contemporary large language model (LLM) research. For this we assess over 2,000 research works released between 2020 and 2024 based on criteria typical of what is considered good research (e.g. presence of statistical tests and reproducibility), and cross-validate it with arguments that are at the centre of controversy (e.g., claims of emergent behaviour). We find multiple trends, such as declines in ethics disclaimers, a rise of LLMs as evaluators, and an increase on claims of LLM reasoning abilities without leveraging human evaluation. We note that conference checklists are effective at curtailing some of these issues, but balancing velocity and rigour in research cannot solely rely on these. We tie all these findings to findings from recent meta-reviews and extend recommendations on how to address what does, does not, and should work in LLM research.
A Survey of Mathematical Reasoning in the Era of Multimodal Large Language Model: Benchmark, Method & Challenges
PDF ↗Mathematical reasoning, a core aspect of human cognition, is vital across many domains, from educational problem-solving to scientific advancements. As artificial general intelligence (AGI) progresses, integrating large language models (LLMs) with mathematical reasoning tasks is becoming increasingly significant. This survey provides **the first comprehensive analysis of mathematical reasoning in the era of multimodal large language models (MLLMs)**. We review over 200 studies published since 2021, and examine the state-of-the-art developments in Math-LLMs, with a focus on multimodal settings. We categorize the field into three dimensions: benchmarks, methodologies, and challenges. In particular, we explore multimodal mathematical reasoning pipeline, as well as the role of (M)LLMs and the associated methodologies. Finally, we identify five major challenges hindering the realization of AGI in this domain, offering insights into the future direction for enhancing multimodal reasoning capabilities. This survey serves as a critical resource for the research community in advancing the capabilities of LLMs to tackle complex multimodal reasoning tasks.
LSC-Eval: A General Framework to Evaluate Methods for Assessing Dimensions of Lexical Semantic Change Using LLM-Generated Synthetic Data
PDF ↗Lexical Semantic Change (LSC) provides insight into cultural and social dynamics. Yet, the validity of methods for measuring different kinds of LSC remains unestablished due to the absence of historical benchmark datasets. To address this gap, we propose LSC-Eval, a novel three-stage general-purpose evaluation framework to: (1) develop a scalable methodology for generating synthetic datasets that simulate theory-driven LSC using In-Context Learning and a lexical database; (2) use these datasets to evaluate the sensitivity of computational methods to synthetic change; and (3) assess their suitability for detecting change in specific dimensions and domains. We apply LSC-Eval to simulate changes along the Sentiment, Intensity, and Breadth (SIB) dimensions, as defined in the SIBling framework, using examples from psychology. We then evaluate the ability of selected methods to detect these controlled interventions. Our findings validate the use of synthetic benchmarks, demonstrate that tailored methods effectively detect changes along SIB dimensions, and reveal that a state-of-the-art LSC model faces challenges in detecting affective dimensions of LSC. LSC-Eval offers a valuable tool for dimension- and domain-specific benchmarking of LSC methods, with particular relevance to the social sciences.
Inductive coding traditionally relies on labor-intensive human efforts, who are prone to inconsistencies and individual biases. Although large language models (LLMs) offer promising automation capabilities, their standalone use often results in inconsistent outputs, limiting their reliability. In this work, we propose a framework that combines ensemble methods with code refinement methodology to address these challenges. Our approach integrates multiple smaller LLMs, fine-tuned via Low-Rank Adaptation (LoRA), and employs a moderator-based mechanism to simulate human consensus. To address the limitations of metrics like ROUGE and BERTScore, we introduce a composite evaluation metric that combines code conciseness and contextual similarity. The validity of this metric is confirmed through correlation analysis with human expert ratings. Results demonstrate that smaller ensemble models with refined outputs consistently outperform other ensembles, individual models, and even large-scale LLMs like GPT-4. Our evidence suggests that smaller ensemble models significantly outperform larger standalone language models, pointing out the risk of relying solely on a single large model for qualitative analysis.
Recent studies on reasoning in language models (LMs) have sparked a debate on whether they can learn systematic inferential principles or merely exploit superficial patterns in the training data. To understand and uncover the mechanisms adopted for formal reasoning in LMs, this paper presents a mechanistic interpretation of syllogistic inference. Specifically, we present a methodology for circuit discovery aimed at interpreting content-independent and formal reasoning mechanisms. Through two distinct intervention methods, we uncover a sufficient and necessary circuit involving middle-term suppression that elucidates how LMs transfer information to derive valid conclusions from premises. Furthermore, we investigate how belief biases manifest in syllogistic inference, finding evidence of partial contamination from additional attention heads responsible for encoding commonsense and contextualized knowledge. Finally, we explore the generalization of the discovered mechanisms across various syllogistic schemes, model sizes and architectures. The identified circuit is sufficient and necessary for syllogistic schemes on which the models achieve high accuracy (\geq 60%), with compatible activation patterns across models of different families. Overall, our findings suggest that LMs learn transferable content-independent reasoning mechanisms, but that, at the same time, such mechanisms do not involve generalizable and abstract logical primitives, being susceptible to contamination by the same world knowledge acquired during pre-training.
Multi-turn instruction following capability constitutes a core competency of large language models (LLMs) in real-world applications. Existing evaluation benchmarks predominantly focus on fine-grained constraint satisfaction and domain-specific capability assessment, yet overlook the crucial structural dependencies between dialogue turns that distinguish multi-turn from single-turn interactions. These structural dependencies not only reflect user intent but also establish an essential second dimension for the instruction following evaluation beyond constraint satisfaction. To address this gap, we propose StructFlowBench, a multi-turn instruction following benchmark with structural flow modeling. The benchmark defines an innovative structural flow framework with six fundamental inter-turn relationships. These relationships introduce novel structural constraints for model evaluation and also serve as generation parameters for creating customized dialogue flows tailored to specific scenarios. Adopting established LLM-based automatic evaluation methodologies, we conduct systematic evaluations of 13 leading open-source and closed-source LLMs. Experimental results reveal significant deficiencies in current models’ comprehension of multi-turn dialogue structures. The code is available at https://github.com/MLGroupJLU/StructFlowBench.
The automation of scientific research through large language models (LLMs) presents significant opportunities but faces critical challenges in knowledge synthesis and quality assurance. We introduce Feedback-Refined Agent Methodology (FRAME), a novel framework that enhances medical paper generation through iterative refinement and structured feedback. Our approach comprises three key innovations: (1) A structured dataset construction method that decomposes 4,287 medical papers into essential research components through iterative refinement; (2) A tripartite architecture integrating Generator, Evaluator, and Reflector agents that progressively improve content quality through metric-driven feedback; and (3) A comprehensive evaluation framework that combines statistical metrics with human-grounded benchmarks. Experimental results demonstrate FRAME’s effectiveness, achieving significant improvements over conventional approaches across multiple models (9.91% average gain with DeepSeek V3, comparable improvements with GPT-4o Mini) and evaluation dimensions. Human evaluation confirms that FRAME-generated papers achieve quality comparable to human-authored works, with particular strength in synthesizing future research directions. The results demonstrated our work could efficiently assist medical research by building a robust foundation for automated medical research paper generation while maintaining rigorous academic standards.
Self-Foveate: Enhancing Diversity and Difficulty of Synthesized Instructions from Unsupervised Text via Multi-Level Foveation
PDF ↗Large language models (LLMs) with instruction following capabilities have demonstrated impressive problem-solving abilities. While synthesizing instructional data from unsupervised text has become a common approach for training such models, conventional methods rely heavily on human effort for data annotation. Although existing automated synthesis paradigms have alleviated this constraint, they still exhibit significant limitations in ensuring adequate diversity and difficulty of synthesized instructions. To address these challenges, we propose Self-Foveate, an innovative LLM-driven method for instruction synthesis. This approach introduces a “Micro-Scatter-Macro” multi-level foveation methodology that effectively guides the LLM to deeply excavate fine-grained information embedded in unsupervised text, thereby enhancing both the diversity and difficulty of synthesized instructions. Comprehensive experiments across multiple unsupervised corpora and diverse model architectures validate the effectiveness and superiority of our proposed method. We publicly release our data and codes: https://github.com/Mubuky/Self-Foveate
Translating multi-word expressions (MWEs) and idioms requires a deep understanding of the cultural nuances of both the source and target languages. This challenge is further amplified by the one-to-many nature of idiomatic translations, where a single source idiom can have multiple target-language equivalents depending on cultural references and contextual variations. Traditional static knowledge graphs (KGs) and prompt-based approaches struggle to capture these complex relationships, often leading to suboptimal translations. To address this, we propose an IdiomCE, an adaptive graph neural network (GNN) based methodology that learns intricate mappings between idiomatic expressions, effectively generalizing to both seen and unseen nodes during training. Our proposed method enhances translation quality even in resource-constrained settings, facilitating improved idiomatic translation in smaller models. We evaluate our approach on multiple idiomatic translation datasets using reference-less metrics, demonstrating significant improvements in translating idioms from English to various Indian languages
Large Language Models (LLMs) often struggle with complex reasoning tasks due to insufficient in-depth insights in their training data, which are frequently absent in publicly available documents. This paper introduces the Chain of Methodologies (CoM), a simple and innovative iterative prompting framework designed to build structured reasoning processes by injecting human methodological insights, thereby enabling LLMs to perform long and effective reasoning for complex tasks. Assuming that LLMs possess certain metacognitive abilities, CoM leverages user-defined methodologies to stimulate the cognitive insights that LLMs have learned implicitly from training data. Experimental results indicate that CoM outperforms competitive baselines, highlighting the potential of training-free prompting methods as general solutions for complex reasoning tasks and the possibility of incorporating human-like methodological insights to bridge the gap to human-level reasoning.
Large Language Models (LLMs) are increasingly deployed as human assistants across various domains where they help to make choices. However, the mechanisms behind LLMs’ choice behavior remain unclear, posing risks in safety-critical situations. Inspired by the intrinsic and extrinsic motivation framework within the classic human behavioral model of Self-Determination Theory and its established research methodologies, we investigate the factors influencing LLMs’ choice behavior by constructing a virtual QA platform that includes three different experimental conditions, with four models from GPT and Llama series participating in repeated experiments. Our findings indicate that LLMs’ behavior is influenced not only by intrinsic attention bias but also by extrinsic social influence, exhibiting patterns similar to the Matthew effect and Conformity. We distinguish independent pathways of these two factors in LLMs’ behavior by self-report. This work provides new insights into understanding LLMs’ behavioral patterns, exploring their human-like characteristics.
Evaluating Implicit Bias in Large Language Models by Attacking From a Psychometric Perspective
PDF ↗As large language models (LLMs) become an important way of information access, there have been increasing concerns that LLMs may intensify the spread of unethical content, including implicit bias that hurts certain populations without explicit harmful words. In this paper, we conduct a rigorous evaluation of LLMs’ implicit bias towards certain demographics by attacking them from a psychometric perspective to elicit agreements to biased viewpoints. Inspired by psychometric principles in cognitive and social psychology, we propose three attack approaches, i.e., Disguise, Deception, and Teaching. Incorporating the corresponding attack instructions, we built two benchmarks: (1) a bilingual dataset with biased statements covering four bias types (2.7K instances) for extensive comparative analysis, and (2) BUMBLE, a larger benchmark spanning nine common bias types (12.7K instances) for comprehensive evaluation. Extensive evaluation of popular commercial and open-source LLMs shows that our methods can elicit LLMs’ inner bias more effectively than competitive baselines. Our attack methodology and benchmarks offer an effective means of assessing the ethical risks of LLMs, driving progress toward greater accountability in their development.
Continued Pretraining and Interpretability-Based Evaluation for Low-Resource Languages: A Galician Case Study
PDF ↗Recent advances in Large Language Models (LLMs) have led to remarkable improvements in language understanding and text generation. However, challenges remain in enhancing their performance for underrepresented languages, ensuring continual learning without catastrophic forgetting, and developing robust evaluation methodologies. This work addresses these issues by investigating the impact of Continued Pretraining (CPT) on multilingual models and proposing a comprehensive evaluation framework for LLMs, focusing on the case of Galician language. Our first contribution explores CPT strategies for languages with limited representation in multilingual models. We analyze how CPT with Galician corpora improves text generation while assessing the trade-offs between linguistic enrichment and task-solving capabilities. Our findings show that CPT with small, high-quality corpora and diverse instructions enhances both task performance and linguistic quality. Our second contribution is a structured evaluation framework based on distinguishing task-based and language-based assessments, leveraging existing and newly developed benchmarks for Galician. Additionally, we contribute new Galician LLMs, datasets for evaluation and instructions, and an evaluation framework.
Visual Language Models (VLMs) have gained significant popularity due to their remarkable ability. While various methods exist to enhance privacy in text-based applications, privacy risks associated with visual inputs remain largely overlooked such as Protected Health Information (PHI) in medical images. To tackle this problem, two key tasks: accurately localizing sensitive text and processing it to ensure privacy protection should be performed. To address this issue, we introduce VisShield (Vision Privacy Shield), an end-to-end framework designed to enhance the privacy awareness of VLMs. Our framework consists of two key components: a specialized instruction-tuning dataset OPTIC (Optical Privacy Text Instruction Collection) and a tailored training methodology. The dataset provides diverse privacy-oriented prompts that guide VLMs to perform targeted Optical Character Recognition (OCR) for precise localization of sensitive text, while the training strategy ensures effective adaptation of VLMs to privacy-preserving tasks. Specifically, our approach ensures that VLMs recognize privacy-sensitive text and output precise bounding boxes for detected entities, allowing for effective masking of sensitive information. Extensive experiments demonstrate that our framework significantly outperforms existing approaches in handling private information, paving the way for privacy-preserving applications in vision-language models.
We present a taxonomy-driven framework for constructing domain-specific knowledge graphs (KGs) that integrates structured taxonomies, Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG). Although we focus on climate science to illustrate its effectiveness, our approach can potentially be adapted for other specialized domains. Existing methods often neglect curated taxonomies—hierarchies of verified entities and relationships—and LLMs frequently struggle to extract KGs in specialized domains. Our approach addresses these gaps by anchoring extraction to expert-curated taxonomies, aligning entities and relations with domain semantics, and validating LLM outputs using RAG against the domain taxonomy. Through a climate science case study using our annotated dataset of 25 publications (1,705 entity-publication links, 3,618 expert-validated relationships), we demonstrate that taxonomy-guided LLM prompting combined with RAG-based validation reduces hallucinations by 23.3% while improving F1 scores by 13.9% compared to baselines without the proposed techniques. Our contributions include: 1) a generalizable methodology for taxonomy-aligned KG construction; 2) a reproducible annotation pipeline, 3) the first benchmark dataset for climate science information retrieval; and 4) empirical insights into combining structured taxonomies with LLMs for specialized domains. The dataset, including expert annotations and taxonomy-aligned outputs, is publicly available at https://github.com/Jo-Pan/ClimateIE, and the accompanying framework can be accessed at https://github.com/Jo-Pan/TaxoDrivenKG.
Do Language Models Understand the Cognitive Tasks Given to Them? Investigations with the N-Back Paradigm
PDF ↗Cognitive tasks originally developed for humans are now increasingly used to study language models. While applying these tasks is often straightforward, interpreting their results can be challenging. In particular, when a model underperforms, it is often unclear whether this results from a limitation in the cognitive ability being tested or a failure to understand the task itself. A recent study argues that GPT 3.5’s declining performance on 2-back and 3-back tasks reflects a working memory capacity limit similar to humans (Gong et al., 2024). By analyzing a range of open-source language models of varying performance levels on these tasks, we show that the poor performance is due at least in part to a limitation in task comprehension and task set maintenance. We challenge the best-performing model with progressively harder versions of the task (up to 10-back) and experiment with alternative prompting strategies, before analyzing model attentions. Our larger aim is to contribute to the ongoing conversation around refining methodologies for the cognitive evaluation of language models.