Automatic pronunciation assessment is typically performed by acoustic models trained on audio-score pairs. Although effective, these systems provide only numerical scores, without the information needed to help learners understand their errors. Meanwhile, large language models (LLMs) have proven effective in supporting language learning, but their potential for assessing pronunciation remains unexplored. In this work, we introduce TextPA, a zero-shot, Textual description-based Pronunciation Assessment approach. TextPA utilizes human-readable representations of speech signals, which are fed into an LLM to assess pronunciation accuracy and fluency, while also providing reasoning behind the assigned scores. Finally, a phoneme sequence match scoring method is used to refine the accuracy scores. Our work highlights a previously overlooked direction for pronunciation assessment. Instead of relying on supervised training with audio-score examples, we exploit the rich pronunciation knowledge embedded in written text. Experimental results show that our approach is both cost-efficient and competitive in performance. Furthermore, TextPA significantly improves the performance of conventional audio-score-trained models on out-of-domain data by offering a complementary perspective.
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
User Feedback in Human-LLM Dialogues: A Lens to Understand Users But Noisy as a Learning Signal
PDF ↗Once language models (LMs) are deployed, they can interact with users long-term, ideally evolving based on their feedback. Asking for direct user feedback can be disruptive; thus, we study harvesting implicit user feedback from user-LM interaction logs. We study two user-LM interaction datasets (WildChat and LMSYS). First, we analyze user feedback in the user-LLM conversation logs, providing insights into when and why such feedback occurs. Second, we study harvesting learning signals from such implicit user feedback. Specifically, we study whether incorporating the contents of user feedback (e.g., user wanted clarification), in addition to the polarity of the feedback, can improve the model performance. We observe mixed results, showing this helps in short human-designed questions (MTBench) but not on longer and more complex questions (WildBench). Together, we provide an in-depth study of implicit user feedback, showing its potential and limitations.
Reasoning language models (RLMs) excel at complex tasks by leveraging a chain-of-thought process to generate structured intermediate steps. However, language mixing, i.e., reasoning steps containing tokens from languages other than the prompt, has been observed in their outputs and shown to affect performance, though its impact remains debated. We present the first systematic study of language mixing in RLMs, examining its patterns, impact, and internal causes across 15 languages, 7 task difficulty levels, and 18 subject areas, and show how all three factors influence language mixing. Moreover, we demonstrate that the choice of reasoning language significantly affects performance: forcing models to reason in Latin or Han scripts via constrained decoding notably improves accuracy. Finally, we show that the script composition of reasoning traces closely aligns with that of the model’s internal representations, indicating that language mixing reflects latent processing preferences in RLMs. Our findings provide actionable insights for optimizing multilingual reasoning and open new directions for reasoning language control to build more interpretable and adaptable RLMs.
Modern LLMs face a major obstacle: each new pre-trained model version requires expensive and repetitive alignment. We propose a method that transfers fine-tuning updates across model versions. The key idea is to extract the *diff vector*, which is the difference in parameters induced by fine-tuning, from a *source* model version and apply it to the base of a different *target* version. We show that transferring diff vectors significantly improves the target base model, often achieving performance comparable to its fine-tuned counterpart. For example, applying the fine-tuning updates from Llama 3.0 8B to Llama 3.1 8B increases accuracy by 46.9% on IFEval and 15.7% on LiveCodeBench without further training, surpassing Llama 3.1 8B Instruct. In multilingual settings, we also observe accuracy gains relative to Llama 3.1 8B Instruct, including 4.7% for Malagasy and 15.5% for Turkish on Global MMLU. Our controlled experiments reveal that fine-tuning transfer works best when source and target models are linearly connected in parameter space. We also show that this transfer provides a stronger and more efficient starting point for subsequent fine-tuning. Finally, we propose an iterative *recycling-then-finetuning* approach for continuous model development, which improves both efficiency and effectiveness. Our findings suggest that fine-tuning transfer is a viable strategy to reduce training costs while maintaining model performance.
Code to Think, Think to Code: A Survey on Code-Enhanced Reasoning and Reasoning-Driven Code Intelligence in LLMs
PDF ↗Code and reasoning recently exhibit a mutually reinforcing relationship in large language models (LLMs): Code is abstract, modular, highly structured and has strong logic, guiding reasoning in training and inference. While reasoning translates high-level goals into small executable steps, enable more sophisticated code intellignece, solving real-world challenging software development problems. In this study, we examine how code serves as a structured medium for enhancing reasoning - providing verifiable execution paths, enforcing logical decomposition, and enabling runtime validation, and how advances in reasoning have transformed code intelligence from basic completion to sophisticated agent - enabling models to tackle complex software engineering tasks through deliberate planning and systematic debugging. Finally, we identify key challenges and propose future research directions may deepen the synergy, ultimately advancing LLM performance in both complex reasoning and code intelligence.
This study investigates the feasibility of automating clinical coding in Russian, a language with limited biomedical resources. We present a new dataset for ICD coding, which includes diagnosis fields from electronic health records (EHRs) annotated with over 10,000 entities and more than 1,500 unique ICD codes. This dataset serves as a benchmark for several state-of-the-art models, including BERT, LLaMA with LoRA, and RAG, with additional experiments examining transfer learning across domains (from PubMed abstracts to medical diagnosis) and terminologies (from UMLS concepts to ICD codes). We then apply the best-performing model to label an in-house EHR dataset containing patient histories from 2017 to 2021. Our experiments, conducted on a carefully curated test set, demonstrate that training with the automated predicted codes leads to a significant improvement in accuracy compared to manually annotated data from physicians. We believe our findings offer valuable insights into the potential for automating clinical coding in resource-limited languages like Russian, which could enhance clinical efficiency and data accuracy in these contexts. Our code and dataset are available at https://github.com/auto-icd-coding/ruccod.
Aspect-based sentiment analysis (ABSA) is a crucial task in information extraction and sentiment analysis, aiming to identify aspects with associated sentiment elements in text. However, existing ABSA datasets are predominantly English-centric, limiting the scope for multilingual evaluation and research. To bridge this gap, we present M-ABSA, a comprehensive dataset spanning 7 domains and 21 languages, making it the most extensive multilingual parallel dataset for ABSA to date. Our primary focus is on triplet extraction, which involves identifying aspect terms, aspect categories, and sentiment polarities. The dataset is constructed through an automatic translation process with human review to ensure quality. We perform extensive experiments using various baselines to assess performance and compatibility on M-ABSA. Our empirical findings highlight that the dataset enables diverse evaluation tasks, such as multilingual and multi-domain transfer learning, and large language model evaluation, underscoring its inclusivity and its potential to drive advancements in multilingual ABSA research.
Despite the impressive performance of LLMs on English-based tasks, little is known about their capabilities in specific languages such as Filipino. In this work, we address this gap by introducing FilBench, a Filipino-centric benchmark designed to evaluate LLMs across a diverse set of tasks and capabilities in Filipino, Tagalog, and Cebuano. We carefully curate the tasks in FilBench to reflect the priorities and trends of NLP research in the Philippines such as Cultural Knowledge, Classical NLP, Reading Comprehension, and Generation. By evaluating 27 state-of-the-art LLMs on FilBench, we find that several LLMs suffer from reading comprehension and translation capabilities. Our results indicate that FilBench is challenging, with the best model, GPT-4o, achieving only a score of 72.23%. Moreover, we also find that models trained specifically for Southeast Asian languages tend to underperform on FilBench, with the highest-performing model, SEA-LION v3 70B, achieving only a score of 61.07%. Our work demonstrates the value of curating language-specific LLM benchmarks to aid in driving progress on Filipino NLP and increasing the inclusion of Philippine languages in LLM development.
Multi-label classification is prevalent in real-world settings, but the behavior of Large Language Models (LLMs) in this setting is understudied. We investigate how autoregressive LLMs perform multi-label classification, focusing on subjective tasks, by analyzing the output distributions of the models at each label generation step. We find that the initial probability distribution for the first label often does not reflect the eventual final output, even in terms of relative order and find LLMs tend to suppress all but one label at each generation step. We further observe that as model scale increases, their token distributions exhibit lower entropy and higher single-label confidence, but the internal relative ranking of the labels improves. Finetuning methods such as supervised finetuning and reinforcement learning amplify this phenomenon. We introduce the task of distribution alignment for multi-label settings: aligning LLM-derived label distributions with empirical distributions estimated from annotator responses in subjective tasks. We propose both zero-shot and supervised methods which improve both alignment and predictive performance over existing approaches. We find one method – taking the max probability over all label generation distributions instead of just using the initial probability distribution – improves both distribution alignment and overall F1 classification without adding any additional computation.
We present REARANK, a large language model (LLM)-based listwise reasoning rerank- ing agent. REARANK explicitly reasons be- fore reranking, significantly improving both performance and interpretability. Leveraging reinforcement learning and data augmentation, REARANK achieves substantial improvements over baseline models across popular informa- tion retrieval benchmarks, notably requiring only 179 annotated samples. Built on top of Qwen2.5-7B, our REARANK-7B demonstrates performance comparable to GPT-4 on both in- domain and out-of-domain benchmarks and even surpasses GPT-4 on reasoning-intensive BRIGHT benchmarks. These results under- score the effectiveness of our approach and highlight how reinforcement learning can en- hance LLM reasoning capabilities in reranking.
Generative AI has established the opportunity to readily transform content from one medium to another. This capability is especially powerful for storytelling, where visual illustrations can illuminate a story originally expressed in text. In this paper, we focus on the task of narrative scene illustration, which involves automatically generating an image depicting a scene in a story. Motivated by recent progress on text-to-image models, we consider a pipeline that uses LLMs as an interface for prompting text-to-image models to generate scene illustrations given raw story text. We apply variations of this pipeline to a prominent story corpus in order to synthesize illustrations for scenes in these stories. We conduct a human annotation task to obtain pairwise quality judgments for these illustrations. The outcome of this process is the SceneIllustrations dataset, which we release as a new resource for future work on cross-modal narrative transformation. Through our analysis of this dataset and experiments modeling illustration quality, we demonstrate that LLMs can effectively verbalize scene knowledge implicitly evoked by story text. Moreover, this capability is impactful for generating and evaluating illustrations.
We present SWAN, a causal Transformer architecture in the decoder-only style that generalizes robustly to sequence lengths substantially longer than those seen during training. SWAN interleaves layers without positional encodings (NoPE) and sliding-window attention layers equipped with rotary positional encodings (SWA-RoPE), and applies a dynamic scaling mechanism for attention scores during inference. Experiments demonstrate that SWAN achieves strong length extrapolation without requiring additional long-context training. In addition, SWAN is more computationally efficient than the standard Transformer architecture, resulting in lower training cost and higher inference throughput. We further demonstrate that existing pre-trained decoder-only models can be adapted to the SWAN architecture with minimal continued training, enabling extended contexts. Overall, our work presents an effective approach for scaling language models to longer contexts in a robust and efficient manner.
As large language models (LLMs) are increasingly used to model and augment collective decision-making, it is critical to examine their alignment with human social reasoning. We present an empirical framework for assessing collective alignment, in contrast to prior work on the individual level. Using the Lost at Sea social psychology task, we conduct a large-scale online experiment (N=748), randomly assigning groups to leader elections with either visible demographic attributes (e.g. name, gender) or pseudonymous aliases. We then simulate matched LLM groups conditioned on the human data, benchmarking Gemini 2.5, GPT-4.1, Claude Haiku 3.5, and Gemma 3. LLM behaviors diverge: some mirror human biases; others mask these biases and attempt to compensate for them. We empirically demonstrate that human-AI alignment in collective reasoning depends on context, cues, and model-specific inductive biases. Understanding how LLMs align with collective human behavior is critical to advancing socially-aligned AI, and demands dynamic benchmarks that capture the complexities of collective reasoning.
Re-Align: Aligning Vision Language Models via Retrieval-Augmented Direct Preference Optimization
PDF ↗The emergence of large Vision Language Models (VLMs) has broadened the scope and capabilities of single-modal Large Language Models (LLMs) by integrating visual modalities, thereby unlocking transformative cross-modal applications in a variety of real-world scenarios. Despite their impressive performance, VLMs are prone to significant hallucinations, particularly in the form of cross-modal inconsistencies. Building on the success of Reinforcement Learning from Human Feedback (RLHF) in aligning LLMs, recent advancements have focused on applying direct preference optimization (DPO) on carefully curated datasets to mitigate these issues. Yet, such approaches typically introduce preference signals in a brute-force manner, neglecting the crucial role of visual information in the alignment process. In this paper, we introduce Re-Align, a novel alignment framework that leverages image retrieval to construct a dual-preference dataset, effectively incorporating both textual and visual preference signals. We further introduce rDPO, an extension of the standard direct preference optimization that incorporates an additional visual preference objective during fine-tuning. Our experimental results demonstrate that Re-Align not only mitigates hallucinations more effectively than previous methods but also yields significant performance gains in general visual question-answering (VQA) tasks. Moreover, we show that Re-Align maintains robustness and scalability across a wide range of VLM sizes and architectures. This work represents a significant step forward in aligning multimodal LLMs, paving the way for more reliable and effective cross-modal applications.
Large Language Models (LLMs) demonstrate exceptional capabilities in a multitude of NLP tasks. However, the efficacy of such models to languages other than English is often limited. Prior works have shown that encoder-only models such as BERT or XLM-RoBERTa show impressive cross lingual transfer of their capabilities from English to other languages. In this work, we propose a pretraining strategy that uses active forgetting to achieve similar cross lingual transfer in decoder-only LLMs. We show that LLMs pretrained with active forgetting are highly effective when adapting to new and unseen languages. Through extensive experimentation, we find that LLMs pretrained with active forgetting are able to learn better multilingual representations which translates to better performance in many downstream tasks.
Chinese idioms (成语, Chengyu) are concise four-character expressions steeped in history and culture, whose literal translations often fail to capture their full meaning. This complexity makes them challenging for language models to interpret and use correctly. Existing benchmarks focus on narrow tasks—multiple-choice cloze tests, isolated translation, or simple paraphrasing. We introduce CHENGYU-BENCH, a comprehensive benchmark featuring three tasks: (1) Evaluative Connotation, classifying idioms as positive or negative; (2) Appropriateness, detecting incorrect idiom usage in context; and (3) Open Cloze, filling blanks in longer passages without options. CHENGYU-BENCH comprises 2,937 human-verified examples covering 1,765 common idioms sourced from diverse corpora. We evaluate leading LLMs and find they achieve over 95% accuracy on Evaluative Connotation, but only ~85% on Appropriateness and ~40% top-1 accuracy in Open Cloze. Error analysis reveals that most mistakes arise from fundamental misunderstandings of idiom meanings. CHENGYU-BENCH demonstrates that while LLMs can reliably gauge idiom sentiment, they still struggle to grasp the cultural and contextual nuances essential for proper usage. The benchmark and code will be released upon paper acceptance.
Softmax self-attention often assigns disproportionate weight to semantically uninformative tokens such as punctuation and special tokens, a phenomenon known as attention noise. While recent methods like Cog Attention and the Differential Transformer have addressed this by introducing negative attention scores, they risk discarding useful information. In this paper, we propose the Integral Transformer, a novel self-attention mechanism that denoises attention by integrating signals sampled from the logit distribution. This approach mitigates noise while preserving the contributions of special tokens critical for model performance. Extensive experiments demonstrate that our model outperforms vanilla, Cog, and Differential attention variants on rigorous knowledge and reasoning benchmarks. Moreover, our analysis reveals that employing vanilla self-attention in the lower Transformer layers enhances performance and that the Integral Transformer more effectively balances attention distributions and reduces rank collapse in upper layers.
The Latin script is often used to informally write languages with non-Latin native scripts. In many cases (e.g., most languages in India), the lack of conventional spelling in the Latin script results in high spelling variability. Such romanization renders languages that are normally easily distinguished due to being written in different scripts – Hindi and Urdu, for example – highly confusable. In this work, we increase language identification (LID) accuracy for romanized text by improving the methods used to synthesize training sets. We find that training on synthetic samples which incorporate natural spelling variation yields higher LID system accuracy than including available naturally occurring examples in the training set, or even training higher capacity models. We demonstrate new state-of-the-art LID performance on romanized text from 20 Indic languages in the Bhasha-Abhijnaanam evaluation set (Madhani et al., 2023a), improving test F1 from the reported 74.7% (using a pretrained neural model) to 85.4% using a linear classifier trained solely on synthetic data and 88.2% when also training on available harvested text.
Recent advancements in large language model (LLM) agents have significantly accelerated scientific discovery automation, yet concurrently raised critical ethical and safety concerns. To systematically address these challenges, we introduce **SafeScientist**, an innovative AI scientist framework explicitly designed to enhance safety and ethical responsibility in AI-driven scientific exploration. SafeScientist proactively refuses ethically inappropriate or high-risk tasks and rigorously emphasizes safety throughout the research process. To achieve comprehensive safety oversight, we integrate multiple defensive mechanisms, including prompt monitoring, agent-collaboration monitoring, tool-use monitoring, and an ethical reviewer component. Complementing SafeScientist, we propose **SciSafetyBench** , a novel benchmark specifically designed to evaluate AI safety in scientific contexts, comprising 240 high-risk scientific tasks across 6 domains, alongside 30 specially designed scientific tools and 120 tool-related risk tasks. Extensive experiments demonstrate that SafeScientist significantly improves safety performance by 35% compared to traditional AI scientist frameworks, without compromising scientific output quality. Additionally, we rigorously validate the robustness of our safety pipeline against diverse adversarial attack methods, further confirming the effectiveness of our integrated approach. The code and data will be available at https://github.com/ulab-uiuc/SafeScientist.**Warning**: this paper contains example data that may be offensive or harmful.
Users often assume that large language models (LLMs) share their cognitive alignment of context and intent, leading them to omit critical information in question-answering (QA) and produce ambiguous queries. Responses based on misaligned assumptions may be perceived as hallucinations. Therefore, identifying possible implicit assumptions is crucial in QA. To address this fundamental challenge, we propose Conditional Ambiguous Question-Answering (CondAmbigQA), a benchmark comprising 2,000 ambiguous queries and condition-aware evaluation metrics. Our study pioneers “conditions” as explicit contextual constraints that resolve ambiguities in QA tasks through retrieval-based annotation, where retrieved Wikipedia fragments help identify possible interpretations for a given query and annotate answers accordingly. Experiments demonstrate that models considering conditions before answering improve answer accuracy by 11.75%, with an additional 7.15% gain when conditions are explicitly provided. These results highlight that apparent hallucinations may stem from inherent query ambiguity rather than model failure, and demonstrate the effectiveness of condition reasoning in QA, providing researchers with tools for rigorous evaluation.