This paper studies the problem of how to use large language models (LLMs) to identify the underlying partial differential equations (PDEs) out of very limited observations of a physical system. Previous methods usually utilize physical-informed neural networks (PINNs) to learn the PDE solver and coefficient of PDEs simultaneously, which could suffer from performance degradation under extreme data scarcity. Towards this end, this paper attempts to utilize LLMs to solve this problem without further fine-tuning by proposing a novel framework named LLM for PDE Discovery (LLM4PD). The core of our LLM4PD is to utilize a coarse-to-fine paradigm to automatically discover underlying PDEs. In the coarse phase, LLM4PD selects the crucial terms from a library with hierarchical prompts and incorporates a review agent to enhance the accuracy. In the fine phase, LLM4PD interacts with a PDE solver to optimize the coefficient of the selected terms with the optimization trajectory. We also provide an adaptive hybrid optimization strategy switching between fine-tuning and exploration to balance stability and efficiency. Extensive experiments on several systems validate the effectiveness of our proposed LLM4PD in different settings.
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
ASTPrompter: Preference-Aligned Automated Language Model Red-Teaming to Generate Low-Perplexity Unsafe Prompts
PDF ↗Existing LLM red-teaming approaches prioritize high attack success rate, often resulting in high-perplexity prompts. This focus overlooks low-perplexity attacks that are more difficult to filter, more likely to arise during benign usage, and more impactful as negative downstream training examples. In response, we introduce ASTPrompter, a single-step optimization method that uses contrastive preference learning to train an attacker to maintain low perplexity while achieving a high attack success rate (ASR). ASTPrompter achieves an attack success rate 5.1 times higher on Llama-8.1B while using inputs that are 2.1 times more likely to occur according to the frozen LLM. Furthermore, our attack transfers to Mistral-7B, Qwen-7B, and TinyLlama in both black- and white-box settings. Lastly, by tuning a single hyperparameter in our method, we discover successful attack prefixes along an efficient frontier between ASR and perplexity, highlighting perplexity as a previously under-considered factor in red-teaming.
Large language models (LLMs) have demonstrated remarkable capabilities in natural language processing tasks. However, they remain vulnerable to semantic inconsistency, where minor formatting variations result in divergent predictions for semantically equivalent inputs. Our comprehensive evaluation reveals that this brittleness persists even in state-of-the-art models such as GPT-4o, posing a serious challenge to their reliability. Through a mechanistic analysis, we find that semantic-equivalent input changes induce instability in internal representations, ultimately leading to divergent predictions. This reflects a deeper structural issue, where form and meaning are intertwined in the embedding space. We further demonstrate that existing mitigation strategies, including direct fine-tuning on format variations, do not fully address semantic inconsistency, underscoring the difficulty of the problem. Our findings highlight the need for deeper mechanistic understanding to develop targeted methods that improve robustness.
Improving Prompt Generalization for Cross-prompt Essay Trait Scoring from the Scoring-invariance Perspective
PDF ↗Cross-prompt trait scoring task aims to learn generalizable scoring capabilities from source- prompt data, enabling automatic scoring across multiple dimensions on unseen essays. Existing research on cross-prompt trait essay scoring primarily focuses on improving model generalization by obtaining prompt-invariant representations. In this paper, we approach the research problem from a different perspective on invariance learning and propose a scoring-invariant learning objective. This objective encourages the model to focus on intrinsic information within the essay that reflects its quality during training, thereby learning generic scoring features. To further enhance the model’s ability to score across multiple dimensions, we introduce a trait feature extraction network based on routing gates into the scoring architecture and propose a trait consistency scoring objective to encourage the model to balance the diversity of trait-specific features with scoring consistency across traits when learning trait-specific essay features. Extensive experiments demonstrate the effectiveness of our approach, showing advantages in multi-trait scoring performance and achieving significant improvements with low-resource prompts.
GRV-KBQA: A Three-Stage Framework for Knowledge Base Question Answering with Decoupled Logical Structure, Semantic Grounding and Structure-Aware Validation
PDF ↗Knowledge Base Question Answering (KBQA) is a fundamental task that enables natural language interaction with structured knowledge bases (KBs).Given a natural language question, KBQA aims to retrieve the answers from the KB. However, existing approaches, including retrieval-based, semantic parsing-based methods and large-language model-based methods often suffer from generating non-executable queries and inefficiencies in query execution. To address these challenges, we propose GRV-KBQA, a three-stage framework that decouples logical structure generation from semantic grounding and incorporates structure-aware validation to enhance accuracy. Unlike previous methods, GRV-KBQA explicitly enforces KB constraints to improve alignment between generated logical forms and KB structures. Experimental results on WebQSP and CWQ show that GRV-KBQA significantly improves performance over existing approaches. The ablation study conducted confirms the effectiveness of the decoupled logical form generation and validation mechanism of our framework.
Seeing is Believing: Emotion-Aware Audio-Visual Language Modeling for Expressive Speech Generation
PDF ↗We present an Audio-Visual Language Model (AVLM) for expressive speech generation by integrating full-face visual cues into a pre-trained expressive speech model. We explore multiple visual encoders and multimodal fusion strategies during pre-training to identify the most effective integration approach. Subsequent fine-tuning on emotion recognition and expressive dialogue tasks yields substantial gains over speech-only baselines (e.g., +5 F1 in emotion recognition). AVLM highlights the value of expressive visual information in guiding speech generation and offers a foundation for end-to-end multimodal conversational systems.
Event causality identification (ECI) is a challenging task that involves predicting causal relationships between events in text. Existing prompt-learning-based methods typically concatenate in-context examples only at the input layer, this shallow integration limits the model’s ability to capture the abstract semantic cues necessary for identifying complex causal relationships. To address this limitation, we propose a novel model called Deep In-Context Prompt (DICP), which injects in-context examples into the deeper layer of a pre-trained language model (PLM). This strategy enables the model to leverage the hierarchical semantic representations formed in deeper layers, thereby enhancing its capacity to learn high-level causal abstractions. Moreover, DICP introduces a multi-layer prompt injection mechanism, distributing diverse in-context examples across multiple transformer layers. This design allows the model to recognize a broader range of causal patterns and improves its generalization across different contexts. We evaluate the DICP model through extensive experiments on two widely used datasets, demonstrating its significant improvement in ECI performance compared to existing approaches. Furthermore, we explore the impact of varying the number of deep layers on performance, providing valuable insights into the optimal layer configuration for ECI tasks.
Currently, large language models (LLMs) based Open domain Natural language planning (LONG) has considerable room for improvement. E.g., non-reusable plans with incomplete intermediate states and missing steps hinder real-world applications. To remedy these flaws, this paper establishes a dataset with a baseline for LONG. The GOLD dataset provides the largest dataset for textual procedures, along with corresponding reusable formal planning domain definitions, to date. The baseline, DIGGER, leverages entity-attribute-level action models, which reveal relevant implicit physical properties (aka attributes) of salient entities in actions. DIGGER first extracts action models and builds typed entity lists from textual procedures. Then, it builds goal states for new tasks and instantiates grounded actions using domain prediction. At last, plans are generalized and translated into textual procedures by LLM. Reference-based metrics, LLM-as-a-Judge, and human evaluation are employed to comprehensively evaluate LONG. Experiments on GOLD validate that DIGGER is stronger and more generalizable than recently proposed approaches and LLMs. I.e., DIGGER is the best in seen domains and applicable to unseen domains without adaptation. Specifically, the BLEU-1 score increased from 0.385 to 0.408 on seen domains and rose to 0.310 on unseen domains.
Large Language Models (LLMs) have shown strong capabilities in zero-shot reasoning and generalization to new tasks. However, the zero-shot performance of general LLMs on complex tasks, such as multi-hop reasoning, remains suboptimal, while reasoning LLMs suffer from hallucinations and unfaithfulness. In this paper, to handle these limitations, we introduce a novel structure analysis method that helps LLMs better understand the question structure and guide the problem-solving process. We demonstrate that existing reasoning strategies, such as Chain-of-Thought and ReAct, significantly benefit from the LLM’s inherent understanding of semantic structure. We further ground our method in the theory of probabilistic graphical models to support its effectiveness. To enhance the reasoning process, we augment the structure analysis with refinement and retrieval capabilities, forming a multi-agent reasoning system called Structure-oriented Autonomous Reasoning Agents (SARA). Extensive experiments show that SARA significantly improves zero-shot performance on knowledge-intensive and mathematical tasks. Remarkably, our approach makes a general LLM competitive with dedicated reasoning models in several benchmarks and demonstrates strong robustness against corrupted reasoning paths.
Recent research has attempted to associate preference optimization (PO) performance with the underlying preference datasets. In this work, our observation is that the differences between the preferred response y^+ and dispreferred response y^- influence what LLMs can learn, which may not match the desirable differences to learn. Therefore, we use distance and reward margin to quantify these differences, and combine them to get Distance Calibrated Reward Margin (DCRM), a metric that measures the quality of a response pair for PO. Intuitively, DCRM encourages minimal noisy differences and maximal desired differences. With this, we study three types of commonly used preference datasets, classified along two axes: the source of the responses and the preference labeling function. We establish a general correlation between higher DCRM of the training set and better learning outcome. Inspired by this, we propose a best-of-N^2 pairing method that selects response pairs with the highest DCRM. Empirically, in various settings, our method produces training datasets that can further improve models’ performance on AlpacaEval, MT-Bench, and Arena-Hard over the existing training sets.
Scientific claim verification against tables typically requires predicting whether a claim is supported or refuted given a table. However, we argue that predicting the final label alone is insufficient: it reveals little about the model’s reasoning and offers limited interpretability. To address this, we reframe table–text alignment as an explanation task, requiring models to identify the table cells essential for claim verification. We build a new dataset by extending the SciTab benchmark with human-annotated cell-level rationales. Annotators verify the claim label and highlight the minimal set of cells needed to support their decision. After the annotation process, we utilize the collected information and propose a taxonomy for handling ambiguous cases. Our experiments show that (i) incorporating table alignment information improves claim verification performance, and (ii) most LLMs, while often predicting correct labels, fail to recover human-aligned rationales, suggesting that their predictions do not stem from faithful reasoning.
Toward Inclusive Language Models: Sparsity-Driven Calibration for Systematic and Interpretable Mitigation of Social Biases in LLMs
PDF ↗Large Language Models (LLMs) such as GPT and LLaMA excel in natural language tasks, e.g., text generation and machine translation. However, inherent biases from training on vast Internet datasets potentially amplify harmful stereotypes—widely held, oversimplified, and often inaccurate generalizations about groups of people. Our contribution introduces a novel, systematic, and architecture-aware method to identify and mitigate stereotypical bias in decoder-only transformer models. This interpretable approach operates without gradient access or retraining from scratch. We first evaluate bias and then apply a bias localization mechanism that correlates internal activations with a newly defined Context Influence (CI) Score. Our method pinpoints specific attention heads that consistently align with biased shifts in model predictions. To mitigate this, we introduce a soft pruning strategy that scales attention head parameters based on their correlation strength, followed by lightweight fine-tuning to maintain fluent text generation. Experiments across five models demonstrate our approach reduces bias by up to 37% on BBQ, 32% on StereoSet, and 33% on CrowS-Pairs while simultaneously improving reasoning performance on MMLU by up to 10%.
Anomaly detection (AD) is an important machine learning task with applications in fraud detection, content moderation, and user behavior analysis. However, AD is relatively understudied in a natural language processing (NLP) context, limiting its effectiveness in detecting harmful content, phishing attempts, and spam reviews. We introduce NLP-ADBench, the most comprehensive NLP anomaly detection (NLP-AD) benchmark to date, which includes eight curated datasets and 19 state-of-the-art algorithms. These span 3 end-to-end methods and 16 two-step approaches that adapt classical, non-AD methods to language embeddings from BERT and OpenAI. Our empirical results show that no single model dominates across all datasets, indicating a need for automated model selection. Moreover, two-step methods with transformer-based embeddings consistently outperform specialized end-to-end approaches, with OpenAI embeddings outperforming those of BERT. We release NLP-ADBench at https://github.com/USC-FORTIS/NLP-ADBench, providing a unified framework for NLP-AD and supporting future investigations.
We propose a novel automatic evaluation metric for open-ended text generation, which is a substantial improvement of the recently developed method, Fourier analysis of cross-entropy (FACE), hence, FACE-2. FACE-2 is a psycholinguistically inspired metric that extracts the dynamic patterns (spectrum) of text surprisal. Examined with open-ended text generation tasks, FACE-2 significantly outperforms a broad set of baseline metrics in revealing the model scaling effect, which scales up to models of 70B parameters, while many other existing metrics fail to capture this effect. We have also confirmed the advantage of FACE-2 in producing stronger agreement with human preferences from a large human-annotated dataset. We advocate for including metrics that mine the dynamics of likelihood in evaluating open-ended text generation, which covers broader aspects of human language than only using static likelihood-based or semantic-based metrics. Code repository: https://github.com/CLCS-SUSTech/FACEScore.
Chat-oriented dialogue systems that deliver tangible benefits, such as sharing news or frailty prevention for seniors, require proactive acquisition of specific user information via chats on user-favored topics. This study proposes the Proactive Information Acquisition (PIA) task to support the development of these systems. In this task, a system needs to acquire a user’s answers to predefined questions without making the user feel abrupt while engaging in a chat on a predefined topic. We created and analyzed a dataset of 650 PIA chats, identifying key challenges and effective strategies for recent LLMs. Our system, designed from these insights, surpassed the performance of LLMs prompted solely with task instructions. Finally, we demonstrate that automatic evaluation of this task is reasonably accurate, suggesting its potential as a framework to efficiently develop techniques for systems dealing with complex dialogue goals, extending beyond the scope of PIA alone. Our dataset is available at: https://github.com/CyberAgentAILab/PIA
ManuSearch: Democratizing Deep Search in Large Language Models with a Transparent and Open Multi-Agent Framework
PDF ↗Recent advances in web-augmented large language models (LLMs) have exhibited strong performance in complex reasoning tasks, yet these capabilities are mostly locked in proprietary systems with opaque architectures. In this work, we propose \textbf{ManuSearch}, a transparent and modular multi-agent framework designed to democratize deep search for LLMs. ManuSearch decomposes the search and reasoning process into three collaborative agents: (1) a solution planning agent that iteratively formulates sub-queries, (2) an Internet search agent that retrieves relevant documents via real-time web search, and (3) a structured webpage reading agent that extracts key evidence from raw web content. To rigorously evaluate deep reasoning abilities, we introduce \textbf{ORION}, a challenging benchmark focused on open-web reasoning over long-tail entities, covering both English and Chinese. Experimental results show that ManuSearch substantially outperforms prior open-source baselines and even surpasses leading closed-source systems. Our work paves the way for reproducible, extensible research in open deep search systems. We release the data and code in [https://github.com/RUCAIBox/ManuSearch](https://github.com/RUCAIBox/ManuSearch).
As machine translation (MT) rapidly advances in bridging global communication gaps, there is growing interest in variety-targeted translation for fine-grained language variants and specific translation styles. This translation variant aims to generate target outputs that are not only contextually accurate but also culturally sensitive. However, the lack of comprehensive evaluation benchmarks has hindered progress in this field. To bridge this gap, this work focuses on the translation across styles and language variants, aiming to establish a robust foundation for the automatic evaluation of fine-grained cultural and stylistic nuances, thereby fostering innovation in culturally sensitive translations. Specifically, we evaluate translations across four key dimensions: semantic preservation, cultural and regional specificity, expression style, and fluency at both the word and sentence levels. Through detailed human evaluations, we validate the high reliability of the proposed evaluation framework. On this basis, we thoroughly assess translations of state-of-the-art large language models (LLMs) for this task, highlighting their strengths and identifying areas for future improvement.
This paper presents a unified framework for extracting n-ary property information from materials science literature, addressing the critical challenge of capturing complex relationships that often span multiple sentences. We introduce three complementary approaches: RE-Composition, which transforms binary relations into n-ary structures; Direct EAE, which models polymer properties as events with multiple arguments; and LLM-Guided Assembly, which leverages high-confidence entity and relation outputs to guide structured extraction. Our framework is built upon two novel resources: MatSciNERE, a comprehensive corpus for materials science entities and relations, and PolyEE, a specialized corpus for polymer property events. Through strategic synthetic data generation for both NER and EAE tasks, we achieve significant performance improvements (up to 5.34 F1 points). Experiments demonstrate that our combined approaches outperform any single method, with the LLM-guided approach achieving the highest F1 score (71.53%). The framework enables more comprehensive knowledge extraction from scientific literature, supporting materials discovery and database curation applications. We plan to release our resources and trained models to the research community.
Self-correction has demonstrated potential in code generation by allowing language models to revise and improve their outputs through successive refinement. Recent studies have explored prompting-based strategies that incorporate verification or feedback loops using proprietary models, as well as training-based methods that leverage their strong reasoning capabilities. However, whether smaller models possess the capacity to effectively guide their outputs through self-reflection remains unexplored. Our findings reveal that smaller models struggle to exhibit reflective revision behavior across both self-correction paradigms. In response, we introduce CoCoS, an approach designed to enhance the ability of small language models for multi-turn code correction. Specifically, we propose an online reinforcement learning objective that trains the model to confidently maintain correct outputs while progressively correcting incorrect outputs as turns proceed. Our approach features an accumulated reward function that aggregates rewards across the entire trajectory and a fine-grained reward better suited to multi-turn correction scenarios. This facilitates the model in enhancing initial response quality while achieving substantial improvements through self-correction. With 1B-scale models, CoCoS achieves improvements of 35.8% on the MBPP and 27.7% on HumanEval compared to the baselines.
Oversensitivity occurs when language models defensively reject prompts that are actually benign. This behavior not only disrupts user interactions but also obscures the boundary between harmful and harmless content. Existing benchmarks rely on static datasets that degrade over time as models evolve, leading to data contamination and diminished evaluative power. To address this, we develop a framework that dynamically generates model-specific challenging datasets, capturing emerging defensive patterns and aligning with each model’s unique behavior. Building on this approach, we construct OverBench, a benchmark that aggregates these datasets across diverse LLM families, encompassing 450,000 samples from 25 models. OverBench provides a dynamic and evolving perspective on oversensitivity, allowing for continuous monitoring of defensive triggers as models advance, highlighting vulnerabilities that static datasets overlook.