| Total: 3353
Goal-oriented script planning, or the ability to devise coherent sequences of actions toward specific goals, is commonly employed by humans to plan for typical activities. In e-commerce, customers increasingly seek LLM-based assistants to generate scripts and recommend products at each step, thereby facilitating convenient and efficient shopping experiences. However, this capability remains underexplored due to several challenges, including the inability of LLMs to simultaneously conduct script planning and product retrieval, difficulties in matching products caused by semantic discrepancies between planned actions and search queries, and a lack of methods and benchmark data for evaluation. In this paper, we step forward by formally defining the task of E-commerce Script Planning (EcomScript) as three sequential subtasks. We propose a novel framework that enables the scalable generation of product-enriched scripts by associating products with each step based on the semantic similarity between the actions and their purchase intentions. By applying our framework to real-world e-commerce data, we construct the very first large-scale EcomScript dataset, EcomScriptBench, which includes 605,229 scripts sourced from 2.4 million products. Human annotations are then conducted to provide gold labels for a sampled subset, forming an evaluation benchmark. Extensive experiments reveal that current (L)LMs face significant challenges with EcomScript tasks, even after fine-tuning, while injecting product purchase intentions improves their performance.
Graph representation learning has garnered significant attention due to its broad applications in various domains, such as recommendation systems and social network analysis. Despite advancements in graph learning methods, challenges still remain in explainability when graphs are associated with semantic features. In this paper, we present GraphNarrator, the first method designed to generate natural language explanations for Graph Neural Networks. GraphNarrator employs a generative language model that maps input-output pairs to explanations reflecting the model’s decision-making process. To address the lack of ground truth explanations to train the model, we propose first generating pseudo-labels that capture the model’s decisions from saliency-based explanations, then using Expert Iteration to iteratively train the pseudo-label generator based on training objectives on explanation quality. The high-quality pseudo-labels are finally utilized to train an end-to-end explanation generator model. Extensive experiments are conducted to demonstrate the effectiveness of GraphNarrator in producing faithful, concise, and human-preferred natural language explanations.
Reward models (RMs) have driven the state-of-the-art performance of LLMs today by enabling the integration of human feedback into the language modeling process. However, RMs are primarily trained and evaluated in English, and their capabilities in multilingual settings remain largely understudied. In this work, we conduct a systematic evaluation of several reward models in multilingual settings. We first construct the first-of-its-kind multilingual RM evaluation benchmark, M-RewardBench, consisting of 2.87k preference instances for 23 typologically diverse languages, that tests the chat, safety, reasoning, and translation capabilities of RMs. We then rigorously evaluate a wide range of reward models on M-RewardBench, offering fresh insights into their performance across diverse languages. We identify a significant gap in RMs’ performances between English and non-English languages and show that RM preferences can change substantially from one language to another. We also present several findings on how different multilingual aspects impact RM performance. Specifically, we show that the performance of RMs is improved with improved translation quality. Similarly, we demonstrate that the models exhibit better performance for high-resource languages. We release M-RewardBench dataset and the codebase in this study to facilitate a better understanding of RM evaluation in multilingual settings.
While recent research increasingly emphasizes the value of human-LLM collaboration in competitive programming and proposes numerous empirical methods, a comprehensive understanding remains elusive due to the fragmented nature of existing studies and their use of diverse, application-specific human feedback. Thus, our work serves a three-fold purpose: First, we present the first taxonomy of human feedback consolidating the entire programming process, which promotes fine-grained evaluation. Second, we introduce ELABORATIONSET, a novel programming dataset specifically designed for human-LLM collaboration, meticulously annotated to enable large-scale simulated human feedback and facilitate cost-effective real human interaction studies. Third, we introduce ELABORATION, a novel benchmark to facilitate a thorough assessment of human-LLM competitive programming. With ELABORATION, we pinpoint strengthes and weaknesses of existing methods, thereby setting the foundation for furture improvement. Our dataset and code will be openly released.
The rise of general-purpose artificial intelligence (AI) systems, particularly large language models (LLMs), has raised pressing moral questions about how to reduce bias and ensure fairness at scale. Researchers have documented a sort of “bias” in the significant correlations between demographics (e.g., race, gender) in LLM prompts and responses, but it remains unclear how LLM fairness could be evaluated with more rigorous definitions, such as group fairness or fair representations. We analyze a variety of technical fairness frameworks and find inherent challenges in each that make the development of a fair LLM intractable. We show that each framework either does not logically extend to the general-purpose AI context or is infeasible in practice, primarily due to the large amounts of unstructured training data and the many potential combinations of human populations, use cases, and sensitive attributes. These inherent challenges would persist for general-purpose AI, including LLMs, even if empirical challenges, such as limited participatory input and limited measurement methods, were overcome. Nonetheless, fairness will remain an important type of model evaluation, and there are still promising research directions, particularly the development of standards for the responsibility of LLM developers, context-specific evaluations, and methods of iterative, participatory, and AI-assisted evaluation that could scale fairness across the diverse contexts of modern human-AI interaction.
Supervised Fine-Tuning (SFT) and Preference Optimization (PO) are key processes for aligning Language Models (LMs) with human preferences post pre-training. While SFT excels in efficiency and PO in effectiveness, they are often combined sequentially without integrating their optimization objectives. This approach ignores the opportunities to bridge their paradigm gap and take the strengths from both. In this paper, we interpret SFT and PO with two sub-processes — *Preference Estimation* and *Transition Optimization* — defined at token level within the Markov Decision Process (MDP). This modeling shows that SFT is only a special case of PO with inferior estimation and optimization. PO estimates the model’s preference by its entire generation, while SFT only scores model’s subsequent predicted tokens based on prior tokens from ground truth answer. These priors deviates from model’s distribution, hindering the preference estimation and transition optimization. Building on this view, we introduce **Intuitive Fine-Tuning (IFT)** to integrate SFT and PO into a single process. Through a temporal residual connection, IFT brings better estimation and optimization by capturing LMs’ intuitive sense of its entire answers. But it solely relies on a single policy and the same volume of non-preference-labeled data as SFT. Our experiments show that IFT performs comparably or even superiorly to SFT and some typical PO methods across several tasks, particularly those requires generation, reasoning, and fact-following abilities. An explainable Frozen Lake game further validates the effectiveness of IFT for getting competitive policy.
Standard bias benchmarks used for large language models (LLMs) measure the association between social attributes in model inputs and single-word model outputs. We test whether these benchmarks are robust to lengthening the model outputs via a more realistic user prompt, in the commonly studied domain of gender-occupation bias, as a step towards measuring Realistic Use and Tangible Effects (i.e., RUTEd evaluations). From the current literature, we adapt three standard metrics of next-word prediction (neutrality, skew, and stereotype), and we develop analogous RUTEd evaluations in three contexts of real-world LLM use: children’s bedtime stories, user personas, and English language learning exercises. We find that standard bias metrics have no significant correlation with long-form output metrics. For example, selecting the least biased model based on the standard “trick tests” coincides with selecting the least biased model based on longer output no more than random chance. There may not yet be evidence to justify standard benchmarks as reliable proxies of real-world biases, and we encourage further development of context-specific RUTEd evaluations.
Large Language Models (LLMs) have shown exciting performance in listwise passage ranking. Due to the limited input length, existing methods often adopt the sliding window strategy. Such a strategy, though effective, is inefficient as it involves repetitive and serialized processing, which usually re-evaluates relevant passages multiple times. As a result, it incurs redundant API costs, which are proportional to the number of inference tokens. The development of long-context LLMs enables the full ranking of all passages within a single inference, avoiding redundant API costs. In this paper, we conduct a comprehensive study of long-context LLMs for ranking tasks in terms of efficiency and effectiveness. Surprisingly, our experiments reveal that full ranking with long-context LLMs can deliver superior performance in the supervised fine-tuning setting with a huge efficiency improvement. Furthermore, we identify two limitations of fine-tuning the full ranking model based on existing methods: (1) sliding window strategy fails to produce a full ranking list as a training label, and (2) the language modeling loss cannot emphasize top-ranked passage IDs in the label. To alleviate these issues, we propose a new complete listwise label construction approach and a novel importance-aware learning objective for full ranking. Experiments show the superior performance of our method over baselines.
This study investigates the integration of diverse patient data sources into multimodal language models for automated chest X-ray (CXR) report generation. Traditionally, CXR report generation relies solely on data from a patient’s CXR exam, overlooking valuable information from patient electronic health records. Utilising the MIMIC-CXR and MIMIC-IV-ED datasets, we investigate the use of patient data from emergency department (ED) records — such as vital signs measured and medicines reconciled during an ED stay — for CXR report generation, with the aim of enhancing diagnostic accuracy. We also investigate conditioning CXR report generation on the clinical history section of radiology reports, which has been overlooked in the literature. We introduce a novel approach to transform these heterogeneous data sources into patient data embeddings that prompt a multimodal language model (CXRMate-ED). Our comprehensive evaluation indicates that using a broader set of patient data significantly enhances diagnostic accuracy. The model, training code, and dataset are publicly available.
The paper focuses on the interpretability of Grammatical Error Correction (GEC) evaluation metrics, which received little attention in previous studies. To bridge the gap, we introduce **CLEME2.0**, a reference-based metric describing four fundamental aspects of GEC systems: hit-correction, wrong-correction, under-correction, and over-correction. They collectively contribute to exposing critical qualities and locating drawbacks of GEC systems. Evaluating systems by combining these aspects also leads to superior human consistency over other reference-based and reference-less metrics. Extensive experiments on two human judgment datasets and six reference datasets demonstrate the effectiveness and robustness of our method, achieving a new state-of-the-art result. Our codes are released at https://github.com/THUKElab/CLEME.
The effective utilization of structured data, integral to corporate data strategies, has been challenged by the rise of large language models (LLMs) capable of processing unstructured information. This shift prompts the question: can LLMs interpret structured data directly in its unstructured form? We propose an automatic evaluation data generation method for assessing LLMs’ reasoning capabilities on structure-rich text to explore this. Our approach supports 8 structured languages and 29 tasks, generating data with adjustable complexity through controllable nesting and structural width. We introduce StrucText-Eval, a benchmark containing 5,800 pre-generated and annotated samples designed to evaluate how well LLMs understand and reason through structured text. StrucText-Eval is divided into two suites: a regular Test suite (3,712 samples) and a Test-Hard suite (2,088 samples), the latter emphasizing the gap between human and model performance on more complex tasks. Experimental results show that while open-source LLMs achieve a maximum accuracy of 74.9% on the standard dataset, their performance drops significantly to 45.8% on the harder dataset. In contrast, human participants reach an accuracy of 92.6% on StrucText-Eval-Hard, highlighting LLMs’ current limitations in handling intricate structural information.
AI holds promise for transforming scientific processes, including hypothesis generation. Prior work on hypothesis generation can be broadly categorized into theory-driven and data-driven approaches. While both have proven effective in generating novel and plausible hypotheses, it remains an open question whether they can complement each other. To address this, we develop the first method that combines literature-based insights with data to perform LLM-powered hypothesis generation. We apply our method on five different datasets and demonstrate that integrating literature and data outperforms other baselines (8.97% over few-shot, 15.75% over literature-based alone, and 3.37% over data-driven alone). Additionally, we conduct the first human evaluation to assess the utility of LLM-generated hypotheses in assisting human decision-making on two challenging tasks: deception detection and AI generated content detection. Our results show that human accuracy improves significantly by 7.44% and 14.19% on these tasks, respectively. These findings suggest that integrating literature-based and data-driven approaches provides a comprehensive and nuanced framework for hypothesis generation and could open new avenues for scientific inquiry.
Recent advances in large language models have highlighted the critical need for precise control over model outputs through predefined constraints. While existing methods attempt to achieve this through either direct instruction-response synthesis or preferential response optimization, they often struggle with constraint understanding and adaptation. This limitation becomes particularly evident when handling fine-grained constraints, leading to either hallucination or brittle performance. We introduce Generative Adversarial Policy Optimization (GAPO), a novel framework that combines GAN-based training dynamics with an encoder-only reward model to progressively learn and adapt to increasingly complex constraints. GAPO leverages adversarial training to automatically generate training samples of varying difficulty while utilizing the encoder-only architecture to better capture prompt-response relationships. Extensive experiments demonstrate GAPO’s superior performance across multiple benchmarks, particularly in scenarios requiring fine-grained constraint handling, where it significantly outperforms existing methods like PPO, DPO, and KTO. Our results suggest that GAPO’s unique approach to preferential prompt learning offers a more robust and effective solution for controlling LLM outputs.
Data synthesis has become a crucial research area in large language models (LLMs), especially for generating high-quality instruction fine-tuning data to enhance downstream performance. In code generation, a key application of LLMs, manual annotation of code instruction data is costly. Recent methods, such as Code Evol-Instruct and OSS-Instruct, leverage LLMs to synthesize large-scale code instruction data, significantly improving LLM coding capabilities. However, these approaches face limitations due to unidirectional synthesis and randomness-driven generation, which restrict data quality and diversity. To overcome these challenges, we introduce Tree-of-Evolution (ToE), a novel framework that models code instruction synthesis process with a tree structure, exploring multiple evolutionary paths to alleviate the constraints of unidirectional generation. Additionally, we propose optimization-driven evolution, which refines each generation step based on the quality of the previous iteration. Experimental results across five widely-used coding benchmarks—HumanEval, MBPP, EvalPlus, LiveCodeBench, and BigCodeBench—demonstrate that base models fine-tuned on just 75k data synthesized by our method achieve comparable or superior performance to the state-of-the-art open-weight Code LLM, Qwen2.5-Coder-Instruct, which was fine-tuned on millions of samples.
Despite the recent strides in large language models, studies have underscored the existence of social biases within these systems. In this paper, we delve into the validation and comparison of the ethical biases of LLMs concerning globally discussed and potentially sensitive topics, hypothesizing that these biases may arise from language-specific distinctions. Introducing the Multilingual Sensitive Questions & Answers Dataset (**MSQAD**), we collected news articles from Human Rights Watch covering 17 topics, and generated socially sensitive questions along with corresponding responses in multiple languages. We scrutinized the biases of these responses across languages and topics, employing two statistical hypothesis tests. The results showed that the null hypotheses were rejected in most cases, indicating biases arising from cross-language differences. It demonstrates that ethical biases in responses are widespread across various languages, and notably, these biases were prevalent even among different LLMs. By making the proposed MSQAD openly available, we aim to facilitate future research endeavors focused on examining cross-language biases in LLMs and their variant models.
Multi-hop question answering (MHQA) involves reasoning across multiple documents to answer complex questions. Dense retrievers typically outperform sparse methods like BM25 by leveraging semantic embeddings in many tasks; however, they require labeled query-document pairs for fine-tuning, which poses a significant challenge in MHQA due to the complexity of the reasoning steps. To overcome this limitation, we introduce Retriever Supervision with Consistency and Relevance (ReSCORE), a novel method for training dense retrievers for MHQA without the need for labeled documents. ReSCORE leverages large language models to measure document-question relevance with answer consistency and utilizes this information to train a retriever within an iterative question-answering framework. Evaluated on three MHQA benchmarks, our extensive experiments demonstrate the effectiveness of ReSCORE, with significant improvements in retrieval performance that consequently lead to state-of-the-art Exact Match and F1 scores for MHQA.
Large Language Models (LLMs) have significantly advanced the fact-checking studies. However, existing automated fact-checking evaluation methods rely on static datasets and classification metrics, which fail to automatically evaluate the justification production and uncover the nuanced limitations of LLMs in fact-checking. In this work, we introduce FACT-AUDIT, an agent-driven framework that adaptively and dynamically assesses LLMs’ fact-checking capabilities. Leveraging importance sampling principles and multi-agent collaboration, FACT-AUDIT generates adaptive and scalable datasets, performs iterative model-centric evaluations, and updates assessments based on model-specific responses. By incorporating justification production alongside verdict prediction, this framework provides a comprehensive and evolving audit of LLMs’ factual reasoning capabilities, to investigate their trustworthiness. Extensive experiments demonstrate that FACT-AUDIT effectively differentiates among state-of-the-art LLMs, providing valuable insights into model strengths and limitations in model-centric fact-checking analysis.
Tasks are central in machine learning, as they are the most natural objects to assess the capabilities of current models. The trend is to build general models able to address any task. Even though transfer learning and multitask learning try to leverage the underlying task space, no well-founded tools are available to study its structure. This study proposes a theoretically grounded setup to define the notion of task and to compute the inclusion between two tasks from a statistical deficiency point of view. We propose a tractable proxy as information sufficiency to estimate the degree of inclusion between tasks, show its soundness on synthetic data, and use it to reconstruct empirically the classic NLP pipeline.
Federated fine-tuning for Large Language Models (LLMs) has recently gained attention due to the heavy communication overhead of transmitting large model updates. Low Rank Adaptation (LoRA) has been proposed as a solution, yet its application in federated learning is complicated by discordance in aggregation. Existing methods addressing this discordance often suffer from performance degradation at low ranks in heterogeneous data settings. In response, we introduce LoRA-A^2 (Low Rank Adaptation with Alternating freeze and Adaptive rank selection), which demonstrates robustness in challenging settings with low ranks and high data heterogeneity. Our experimental findings reveal that LoRA-A^2 maintains performance even under extreme heterogeneity and low rank conditions, achieving up to a 99.8% reduction in uploaded parameters compared to full fine-tuning without compromising performance. This adaptive mechanism boosts robustness and communication efficiency in federated fine-tuning, enabling the practical deployment of LLMs in resource-constrained environments.
Detecting tricky bugs in plausible programs, those that pass existing test suites yet still contain bugs, remains a significant challenge in software testing. To address this problem, we propose TrickCatcher, an LLM-powered approach to generating test cases for uncovering bugs in plausible programs. TrickCatcher operates in three stages: First, it uses an LLM to generate program variants based on the program under test (PUT) and its specification. Second, it employs an LLM to construct an input generator from the specification for producing test inputs. Finally, these inputs are executed on both the PUT and its program variants to detect inconsistencies in their outputs. We evaluate TrickCatcher on two datasets, TrickyBugs and EvalPlus, which include 366 human-written and 151 AI-generated plausible programs with tricky bugs. TrickCatcher achieves recall, precision, and F1 scores that are 1.80×, 2.65×, and 1.66× those of the state-of-the-art baselines, respectively. Code and data used are available at https://github.com/RinCloud/TrickCatcher/.
As Large Language Models (LLMs) scale up and gain powerful Chain-of-Thoughts (CoTs) reasoning abilities, practical resource constraints drive efforts to distill these capabilities into more compact Smaller Language Models (SLMs). We find that CoTs consist mainly of simple reasoning forms, with a small proportion (4.7%) of key reasoning steps that truly impact conclusions. However, previous distillation methods typically involve supervised fine-tuning student SLMs only on correct CoTs data produced by teacher LLMs, resulting in students struggling to learn the key, instead imitating the teacher’s reasoning forms and making errors or omissions in reasoning. To address these issues, drawing an analogy to human learning, where analyzing mistakes according to correct solutions often reveals the crucial steps leading to successes or failures, we propose mistakE-Driven key reasonIng step distillaTion (EDIT), a novel method that further aids SLMs learning key reasoning steps rather than mere simple fine-tuning. Firstly, to expose the crucial steps in CoTs, we carefully design specific prompts to generate dual CoTs data with similar reasoning paths but divergent conclusions. Then, we apply the minimum edit distance algorithm on the dual CoTs data to locate these key steps and optimize the likelihood on these tokens. Extensive experiments and analysis validate the effectiveness of EDIT across both in-domain(IND) and out-of-domain(OOD) benchmark reasoning datasets.
With the advancement of large language models (LLMs), intelligent models have evolved from mere tools to autonomous agents with their own goals and strategies for cooperating with humans. This evolution has birthed a novel paradigm in NLP, i.e., human-model cooperation, that has yielded remarkable progress in numerous NLP tasks in recent years. In this paper, we take the first step to present a thorough review of human-model cooperation, exploring its principles, formalizations, and open challenges. In particular, we introduce a new taxonomy that provides a unified perspective to summarize existing approaches. Also, we discuss potential frontier areas and their corresponding challenges. We regard our work as an entry point, paving the way for more breakthrough research in this regard.
Text-based hyperbole and metaphor detection are of great significance for natural language processing (NLP) tasks. However, due to their semantic obscurity and expressive diversity, it is rather challenging to identify them. Existing methods mostly focus on superficial text features, ignoring the associations of hyperbole and metaphor as well as the effect of implicit emotion on perceiving these rhetorical devices. To implement these hypotheses, we propose an emotion-guided hyperbole and metaphor detection framework based on bidirectional dynamic interaction (EmoBi). Firstly, the emotion analysis module deeply mines the emotion connotations behind hyperbole and metaphor. Next, the emotion-based domain mapping module identifies the target and source domains to gain a deeper understanding of the implicit meanings of hyperbole and metaphor. Finally, the bidirectional dynamic interaction module enables the mutual promotion between hyperbole and metaphor. Meanwhile, a verification mechanism is designed to ensure detection accuracy and reliability. Experiments show that EmoBi outperforms all baseline methods on four datasets. Specifically, compared to the current SoTA, the F1 score increased by 28.1% for hyperbole detection on the TroFi dataset and 23.1% for metaphor detection on the HYPO-L dataset. These results, underpinned by in-depth analyses, underscore the effectiveness and potential of our approach for advancing hyperbole and metaphor detection.
In-context learning (ICL) enhances the reasoning abilities of Large Language Models (LLMs) by prepending a few demonstrations. It motivates researchers to introduce more examples to provide additional contextual information for the generation. However, existing methods show a significant limitation due to the problem of excessive growth in context length which causes a large hardware burden. Additionally, shallow-relevant examples selected by out-off-shelf tools hinder LLMs from capturing useful contextual information for generation. In this paper, to approach these limitations, we propose UniICL, a novel Unified ICL framework that unifies demonstration compression, demonstration selection, and final response generation. Furthermore, to avoid repeated compression of the same demonstration and boost inference efficiency, we design a tailored compression strategy that allows UniICL caching compression results into Demonstration Bank(DB). Extensive out-of-domain evaluations prove the advantages of UniICL in both effectiveness and efficiency.
In the epoch of multilingual large language models (LLMs), it is still challenging to evaluate the models’ understanding of lower-resourced languages, which motivates further development of expert-crafted natural language understanding benchmarks. We introduce BelarusianGLUE — a natural language understanding benchmark for Belarusian, an East Slavic language, with ≈15K instances in five tasks: sentiment analysis, linguistic acceptability, word in context, Winograd schema challenge, textual entailment. A systematic evaluation of BERT models and LLMs against this novel benchmark reveals that both types of models approach human-level performance on easier tasks, such as sentiment analysis, but there is a significant gap in performance between machine and human on a harder task — Winograd schema challenge. We find the optimal choice of model type to be task-specific: e.g. BERT models underperform on textual entailment task but are competitive for linguistic acceptability. We release the datasets (https://hf.co/datasets/maaxap/BelarusianGLUE) and evaluation code (https://github.com/maaxap/BelarusianGLUE).