论文检索

Supriti Sinhamahapatra, Jan Niehues

State-of-the-art (SOTA) Automatic Speech Recognition (ASR) systems primarily rely on acoustic information while disregarding additional multi-modal context. However, visual information are essential in disambiguation and adaptation. While most work focus on speaker images to handle noise conditions, this work also focuses on integrating presentation slides for the use cases of scientific presentation.In a first step, we create a benchmark for multi-modal presentation including an automatic analysis of transcribing domain-specific terminology. Next, we explore methods for augmenting speech models with multi-modal information. We mitigate the lack of datasets with accompanying slides by a suitable approach of data augmentation.Finally, we train a model using the augmented dataset, resulting in a relative reduction in word error rate of approximately 34%, across all words and 35%, for domain-specific terms compared to the baseline model.

Tianhao Niu, Yiming Cui, Baoxin Wang, Xiao Xu, Xin Yao, Qingfu Zhu, Dayong Wu, Shijin Wang, Wanxiang Che

Chart2code has recently received significant attention in the multimodal community due to its potential to reduce the burden of visualization and promote a more detailed understanding of charts. However, existing Chart2code-related training datasets suffer from at least one of the following issues: (1) limited scale, (2) limited type coverage, and (3) inadequate complexity. To address these challenges, we seek more diverse sources that better align with real-world user distributions and propose dual data synthesis pipelines: (1) synthesize based on online plotting code. (2) synthesize based on chart images in the academic paper. We create a large-scale Chart2code training dataset Chart2code53, including 53 chart types, 130K Chart-code pairs based on the pipeline. Experimental results demonstrate that even with few parameters, the model finetuned on Chart2code53 achieves state-of-the-art performance on multiple Chart2code benchmarks within open-source models.

Qihang Ma, Shengyu Li, Jie Tang, Dingkang Yang, Shaodong Chen, Yingyi Zhang, Chao Feng, Jiao Ran

Multi-modal keyphrase prediction (MMKP) aims to advance beyond text-only methods by incorporating multiple modalities of input information to produce a set of conclusive phrases. Traditional multi-modal approaches have been proven to have significant limitations in handling the challenging absence and unseen scenarios. Additionally, we identify shortcomings in existing benchmarks that overestimate model capability due to significant overlap in training tests. In this work, we propose leveraging vision-language models (VLMs) for the MMKP task. Firstly, we use two widely-used strategies, e.g., zero-shot and supervised fine-tuning (SFT) to assess the lower bound performance of VLMs. Next, to improve the complex reasoning capabilities of VLMs, we adopt Fine-tune-CoT, which leverages high-quality CoT reasoning data generated by a teacher model to finetune smaller models. Finally, to address the “overthinking” phenomenon, we propose a dynamic CoT strategy which adaptively injects CoT data during training, allowing the model to flexibly leverage its reasoning capabilities during the inference stage. We evaluate the proposed strategies on various datasets and the experimental results demonstrate the effectiveness of the proposed approaches. The code is available at https://github.com/bytedance/DynamicCoT.

Zitao Fang, Guodong Du, Shuyang Yu, Yifei Guo, Yiwei Zhang, Yiyao Cao, Jing Li, Ho-Kin Tang, Sim Kuan Goh

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/).

Yunhui Jang, Jaehyung Kim, Sungsoo Ahn

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.

Dongjun Kim, Gyuho Shim, Yongchan Chun, Minhyuk Kim, Chanjun Park, Heuiseok Lim

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.

Kun Zhu, Lizi Liao, Yuxuan Gu, Lei Huang, Xiaocheng Feng, Bing Qin

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.

Sadam Al-Azani, Maad Alowaifeer, Alhanoof Alhunief, Ahmed Abdelali

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.

Yize Cheng, Wenxiao Wang, Mazda Moayeri, Soheil Feizi

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.

Punit Kumar Singh, Nishant Kumar, Akash Ghosh, Kunal Pasad, Khushi Soni, Manisha Jaishwal, Sriparna Saha, Syukron Abu Ishaq Alfarozi, Asres Temam Abagissa, Kitsuchart Pasupa 等

Language Models (LMs) are primarily evaluated on globally popular sports, often overlooking regional and indigenous sporting traditions. To address this gap, we introduce CultSportQA, a benchmark designed to assess LMs’ understanding of traditional sports across 60 countries and 6 continents, encompassing four distinct cultural categories. The dataset features 33,000 multiple-choice questions (MCQs) across text and image modalities, categorized into primarily three key types: history-based, rule-based, and scenario-based. To evaluate model performance, we employ zero-shot, few-shot, and chain-of-thought (CoT) prompting across a diverse set of Large Language Models (LLMs), Small Language Models (SLMs), and Multimodal Large Language Models (MLMs). By providing a comprehensive multilingual and multicultural sports benchmark, CultSportQA establishes a new standard for assessing AI’s ability to understand and reason about traditional sports. The dataset will be publicly available, fostering research in culturally aware AI systems.

Xiaolong Wang, Zhaolu Kang, Wangyuxuan Zhai, Xinyue Lou, Yunghwei Lai, Ziyue Wang, Yawen Wang, Kaiyu Huang, Yile Wang, Peng Li 等

Multimodal Large Language Models (MLLMs) have demonstrated significant advances across numerous vision-language tasks. Due to their strong performance in image-text alignment, MLLMs can effectively understand image-text pairs with clear meanings. However, effectively resolving the inherent ambiguities in natural language and visual contexts remains challenging. Existing multimodal benchmarks typically overlook linguistic and visual ambiguities, relying mainly on unimodal context for disambiguation and thus failing to exploit the mutual clarification potential between modalities. To bridge this gap, we introduce MUCAR, a novel and challenging benchmark designed explicitly for evaluating multimodal ambiguity resolution across multilingual and cross-modal scenarios. MUCAR includes: (1) a multilingual dataset where ambiguous textual expressions are uniquely resolved by corresponding visual contexts, and (2) a dual-ambiguity dataset that systematically pairs ambiguous images with ambiguous textual contexts, with each combination carefully constructed to yield a single, clear interpretation through mutual disambiguation. Extensive evaluations involving 19 state-of-the-art multimodal models—encompassing both open-source and proprietary architectures—reveal substantial gaps compared to human-level performance, highlighting the need for future research into more sophisticated cross-modal ambiguity comprehension methods, further pushing the boundaries of multimodal reasoning.

Shiyu Li, Yang Tang, Ruijie Liu, Shi-Zhe Chen, Xi Chen

Large language models (LLMs) have recently demonstrated excellent performance in text embedding tasks. Previous work usually use LoRA to fine-tune existing LLMs, which are limited by the data and training gap between LLMs and embedding models. In this work, we introduce Conan-embedding-v2, a new 1.4B-parameter LLM trained from scratch and fine-tuned as a text embedder. First, we add news data and multilingual pairs for LLM pretraining to bridge the data gap. Based on this, we propose a cross-lingual retrieval dataset that enables the LLM to better integrate embeddings across different languages. Second, whereas LLMs use a causal mask with token-level loss, embedding models use a bidirectional mask with sentence-level loss. This training gap makes full fine-tuning less effective than LoRA. We introduce a soft-masking mechanism to gradually transition between these two types of masks, enabling the model to learn more comprehensive representations. Based on this, we propose a dynamic hard negative mining method that exposes the model to more difficult negative examples throughout the training process. Being intuitive and effective, with only approximately 1.4B parameters, Conan-embedding-v2 achieves SOTA performance on both the Massive Text Embedding Benchmark (MTEB) and Chinese MTEB (May 19, 2025).

Jingyao Li, Senqiao Yang, Sitong Wu, Han Shi, Chuanyang Zheng, Hong Xu, Jiaya Jia

In recent years, developing compact and efficient large language models (LLMs) has emerged as a thriving area of research. However, traditional Supervised Fine-Tuning (SFT), which relies on singular ground truth labels, often fails to capture token-level dependencies and linguistic diversity. To address these limitations, we propose a logits-based fine-tuning framework that integrates the strengths of supervised learning and knowledge distillation. Our approach constructs enriched training targets by combining teacher logits with ground truth labels, preserving both correctness and linguistic diversity. This ensures more reliable and effective training. To validate our approach, we constructed a large-scale 1.2M logits dataset and trained a series of science-focused models. Experimental results demonstrate that our method achieves significant improvements over current SOTA, with accuracy gains of 18% on Mawps and 22.7% on TabMWP. Across nine widely used mathematical benchmarks, our method consistently outperforms prior SFT models, achieving an average improvement of 7.28%. All code and datasets will be open-sourced.

Aida Ramezani, Yang Xu

Effective interactions between artificial intelligence (AI) and humans require an equitable and accurate representation of diverse cultures. It is known that current AI, particularly large language models (LLMs), possesses some degrees of cultural knowledge but not without limitations. We present a framework aimed at understanding the origin of these limitations. We hypothesize that there is a fundamental discordance between embedded ethics—how LLMs represent right versus wrong, and cultural inference—how LLMs infer cultural knowledge, specifically cultural norms. We demonstrate this by extracting low-dimensional subspaces that embed ethical principles of LLMs based on established benchmarks. We then show that how LLMs make errors in culturally distinctive scenarios significantly correlates with how they represent cultural norms with respect to these embedded ethics subspaces. Furthermore, we show that coercing cultural norms to be more aligned with the embedded ethics increases LLM performance in cultural inference. Our analyses of 12 language models, two large-scale cultural benchmarks spanning 75 countries and two ethical datasets indicate that 1) the ethics-culture discordance tends to be exacerbated in instruct-tuned models, and 2) how current LLMs represent ethics can impose limitations on their adaptation to diverse cultures particularly pertaining to non-Western and low-income regions.

Decheng Duan, Jitong Peng, Yingyi Zhang, Chengzhi Zhang

Structured information extraction from scientific literature is crucial for capturing core concepts and emerging trends in specialized fields. While existing datasets aid model development, most focus on specific publication sections due to domain complexity and the high cost of annotating scientific texts. To address this limitation, we introduce SciNLP—a specialized benchmark for full-text entity and relation extraction in the Natural Language Processing (NLP) domain. The dataset comprises 60 manually annotated full-text NLP publications, covering 7,072 entities and 1,826 relations. Compared to existing research, SciNLP is the first dataset providing full-text annotations of entities and their relationships in the NLP domain. To validate the effectiveness of SciNLP, we conducted comparative experiments with similar datasets and evaluated the performance of state-of-the-art supervised models on this dataset. Results reveal varying extraction capabilities of existing models across academic texts of different lengths. Cross-comparisons with existing datasets show that SciNLP achieves significant performance improvements on certain baseline models. Using models trained on SciNLP, we implemented automatic construction of a fine-grained knowledge graph for the NLP domain. Our KG has an average node degree of 3.2 per entity, indicating rich semantic topological information that enhances downstream applications. The dataset is publicly available at: https://github.com/AKADDC/SciNLP.

Xingyu Tan, Xiaoyang Wang, Qing Liu, Xiwei Xu, Xin Yuan, Liming Zhu, Wenjie Zhang

Retrieval-augmented generation (RAG) enhances large language models (LLMs) by incorporating external knowledge. Current hybrid RAG system retrieves evidence from both knowledge graphs (KGs) and text documents to support LLM reasoning. However, it faces challenges like handling multi-hop reasoning, multi-entity questions, multi-source verification, and effective graph utilization. To address these limitations, we present HydraRAG, a training-free framework that unifies graph topology, document semantics, and source reliability to support deep, faithful reasoning in LLMs. HydraRAG handles multi-hop and multi-entity problems through agent-driven exploration that combines structured and unstructured retrieval, increasing both diversity and precision of evidence. To tackle multi-source verification, HydraRAG uses a tri-factor cross-source verification (source trustworthiness assessment, cross-source corroboration, and entity-path alignment), to balance topic relevance with cross-modal agreement. By leveraging graph structure, HydraRAG fuses heterogeneous sources, guides efficient exploration, and prunes noise early. Comprehensive experiments on seven benchmark datasets show that HydraRAG achieves overall state-of-the-art results on all benchmarks with GPT-3.5-Turbo, outperforming the strong hybrid baseline ToG-2 by an average of 20.3% and up to 30.1%. Furthermore, HydraRAG enables smaller models (e.g., Llama-3.1-8B) to achieve reasoning performance comparable to that of GPT-4-Turbo. The source code is available on https://stevetantan.github.io/HydraRAG/.

Tony Woo, Sehun Lee, Kang-wook Kim, Gunhee Kim

Spoken dialogue systems increasingly employ large language models (LLMs) to leverage their advanced reasoning capabilities. However, direct application of LLMs in spoken communication often yield suboptimal results due to mismatches between optimal textual and verbal delivery. While existing approaches adapt LLMs to produce speech-friendly outputs, their impact on reasoning performance remains underexplored. In this work, we propose **Think-Verbalize-Speak**, a framework that decouples reasoning from spoken delivery to preserve the full reasoning capacity of LLMs. Central to our method is *verbalizing*, an intermediate step that translates thoughts into natural, speech-ready text. We also introduce **ReVerT**, a latency-efficient verbalizer based on incremental and asynchronous summarization. Experiments across multiple benchmarks show that our method enhances speech naturalness and conciseness with minimal impact on reasoning. The project page with the dataset and the source code is available at https://yhytoto12.github.io/TVS-ReVerT.

Sangyeon Cho, Mingi Kim, Jinkwon Hwang, Jaehoon Go, Minuk Ma, Sunjae Yoon, Junyeong Kim

Automated Audio Captioning (AAC) aims to generate natural language descriptions of audio content, enabling machines to interpret and communicate complex acoustic scenes. However, current AAC datasets often suffer from short and simplistic captions, limiting model expressiveness and semantic depth. To address this, we introduce **VggCaps**, a new multi-modal dataset that pairs audio with corresponding video and leverages large language models (LLMs) to generate rich, descriptive captions. VggCaps significantly outperforms existing benchmarks in caption length, lexical diversity, and human-rated quality. Furthermore, we propose **Multi2Cap**, a novel AAC framework that learns audio-visual representations through a AV-grounding module during pre-training and reconstructs visual semantics using audio alone at inference. This enables visually grounded captioning in audio-only scenarios. Experimental results on Clotho and AudioCaps demonstrate that Multi2Cap achieves state-of-the-art performance across multiple metrics, validating the effectiveness of cross-modal supervision and LLM-based generation in advancing AAC.

Xiaohan Yu, Pu Jian, Chong Chen

Retrieval-Augmented Generation (RAG) has demonstrated considerable effectiveness in open-domain question answering. However, when applied to heterogeneous documents, comprising both textual and tabular components, existing RAG approaches exhibit critical limitations. The prevailing practice of flattening tables and chunking strategies disrupts the intrinsic tabular structure, leads to information loss, and undermines the reasoning capabilities of LLMs in multi-hop, global queries. To address these challenges, we propose TableRAG, an SQL-based framework that unifies textual understanding and complex manipulations over tabular data. TableRAG iteratively operates in four steps: context-sensitive query decomposition, text retrieval, SQL programming and execution, and compositional intermediate answer generation. We also develop HeteQA, a novel benchmark designed to evaluate the multi-hop heterogeneous reasoning capabilities. Experimental results demonstrate that TableRAG consistently outperforms existing baselines on both public datasets and our HeteQA, establishing a new state-of-the-art for heterogeneous document question answering.

Shane Storks, Itamar Bar-Yossef, Yayuan Li, Zheyuan Zhang, Jason J Corso, Joyce Chai

Procedural mistake detection (PMD) is a challenging problem of classifying whether a human user (observed through egocentric video) has successfully executed a task (specified by a procedural text). Despite significant recent efforts, machine performance in the wild remains nonviable, and the reasoning processes underlying this performance are opaque. As such, we extend PMD to require generating visual self-dialog rationales to inform decisions. Given the impressive, mature image understanding capabilities observed in recent vision-and-language models (VLMs), we curate a suitable benchmark dataset for PMD based on individual frames. As our reformulation enables unprecedented transparency, we leverage a natural language inference (NLI) model to formulate two automated metrics for the coherence of generated rationales. We establish baselines for this reframed task, showing that VLMs struggle off-the-shelf, but with some trade-offs, their accuracy, coherence, and efficiency can be improved by incorporating these metrics into common inference and fine-tuning methods. Lastly, our multi-faceted metrics visualize common outcomes, highlighting areas for further improvement.