Question answering (QA) requires accurately aligning user questions with structured queries, a process often limited by the scarcity of high-quality query-natural language (Q-NL) pairs. To overcome this, we present Q-NL Verifier, an approach to generating high-quality synthetic pairs of queries and NL translations. Our approach relies on large language models (LLMs) to generate semantically precise natural language paraphrases of structured queries. Building on these synthetic Q-NL pairs, we introduce a learned verifier component that automatically determines whether a generated paraphrase is semantically equivalent to the original query. Our experiments with the well-known LC-QuAD 2.0 benchmark show that Q-NL Verifier generalizes well to paraphrases from other models and even human-authored translations. Our approach strongly aligns with human judgments across varying query complexities and outperforms existing NLP metrics in assessing semantic correctness. We also integrate the verifier into QA pipelines, showing that verifier-filtered synthetic data has significantly higher quality in terms of translation correctness and enhances NL to Q translation accuracy. Lastly, we release an updated version of the LC-QuAD 2.0 benchmark containing our synthetic Q-NL pairs and verifier scores, offering a new resource for robust and scalable QA.
Test-time inference has emerged as a powerful paradigm for enabling language models to ``think'' longer and more carefully about complex challenges, much like skilled human experts. While reinforcement learning (RL) can drive self-improvement in language models on verifiable tasks, some models exhibit substantial gains while others quickly plateau. For instance, we find that Qwen-2.5-3B far exceeds Llama-3.2-3B under identical RL training for the game of Countdown. This discrepancy raises a critical question: what intrinsic properties enable effective self-improvement? We introduce a framework to investigate this question by analyzing four key cognitive behaviors -- verification, backtracking, subgoal setting, and backward chaining -- that both expert human problem solvers and successful language models employ. Our study reveals that Qwen naturally exhibits these reasoning behaviors, whereas Llama initially lacks them. In systematic experimentation with controlled behavioral datasets, we find that priming Llama with examples containing these reasoning behaviors enables substantial improvements during RL, matching or exceeding Qwen's performance. Importantly, the presence of reasoning behaviors, rather than correctness of answers, proves to be the critical factor -- models primed with incorrect solutions containing proper reasoning patterns achieve comparable performance to those trained on correct solutions. Finally, leveraging continued pretraining with OpenWebMath data, filtered to amplify reasoning behaviors, enables the Llama model to match Qwen's self-improvement trajectory. Our findings establish a fundamental relationship between initial reasoning behaviors and the capacity for improvement, explaining why some language models effectively utilize additional computation while others plateau.
Large language models (LLMs) have demonstrated significant progress in multilingual language understanding and generation. However, due to the imbalance in training data, their capabilities in non-English languages are limited. Recent studies revealed the English-pivot multilingual mechanism of LLMs, where LLMs implicitly convert non-English queries into English ones at the bottom layers and adopt English for thinking at the middle layers. However, due to the absence of explicit supervision for cross-lingual alignment in the intermediate layers of LLMs, the internal representations during these stages may become inaccurate. In this work, we introduce a deep supervision fine-tuning method (DFT) that incorporates additional supervision in the internal layers of the model to guide its workflow. Specifically, we introduce two training objectives on different layers of LLMs: one at the bottom layers to constrain the conversion of the target language into English, and another at the middle layers to constrain reasoning in English. To effectively achieve the guiding purpose, we designed two types of supervision signals: logits and feature, which represent a stricter constraint and a relatively more relaxed guidance. Our method guides the model to not only consider the final generated result when processing non-English inputs but also ensure the accuracy of internal representations. We conducted extensive experiments on typical English-centric large models, LLaMA-2 and Gemma-2, and the results on multiple multilingual datasets show that our method significantly outperforms traditional fine-tuning methods.
Incorrect boundary division, complex semantic representation, and differences in pronunciation and meaning often lead to errors in Chinese Named Entity Recognition(CNER). To address these issues, this paper proposes HREB-CRF framework: Hierarchical Reduced-bias EMA with CRF. The proposed method amplifies word boundaries and pools long text gradients through exponentially fixed-bias weighted average of local and global hierarchical attention. Experimental results on the MSRA, Resume, and Weibo datasets show excellent in F1, outperforming the baseline model by 1.1\%, 1.6\%, and 9.8\%. The significant improvement in F1 shows evidences of strong effectiveness and robustness of approach in CNER tasks.
Large Language Models (LLMs) have shown significant promise across various natural language processing tasks. However, their application in the field of pathology, particularly for extracting meaningful insights from unstructured medical texts such as pathology reports, remains underexplored and not well quantified. In this project, we leverage state-of-the-art language models, including the GPT family, Mistral models, and the open-source Llama models, to evaluate their performance in comprehensively analyzing pathology reports. Specifically, we assess their performance in cancer type identification, AJCC stage determination, and prognosis assessment, encompassing both information extraction and higher-order reasoning tasks. Based on a detailed analysis of their performance metrics in a zero-shot setting, we developed two instruction-tuned models: Path-llama3.1-8B and Path-GPT-4o-mini-FT. These models demonstrated superior performance in zero-shot cancer type identification, staging, and prognosis assessment compared to the other models evaluated.
Evolution, the engine behind the survival and growth of life on Earth, operates through the population-based process of reproduction. Inspired by this principle, this paper formally defines a newly emerging problem -- the population-based evolution of large language models (LLMs) -- and introduces a novel framework. Starting with a population of parent LLMs, our framework enables the population to evolve through four key operations: (i) crossover, merging the weights of different parents to create offspring LLMs, (ii) mutation, introducing small, random changes to model weights to foster diversity, (iii) selection, prioritizing high-performing models, and (iv) succession, transferring the learned experience from parent to offspring LLMs. With only 200 samples per new task, the LLM population evolves rapidly to adapt to the task at hand, without any gradients. Experiments on 12 datasets show that our framework consistently outperforms existing multi-LLM merging and adaptation methods, achieving accuracy gains of up to 54.8% over the best LLM in the initial population. Moreover, our framework allows for the evolution of LLMs across multiple new tasks simultaneously, scaling effectively with populations of up to 40 LLMs, and even zero-shot generalization to unseen held-out tasks. We have open-sourced the code on GitHub and released the weights of 10 parent LLMs, fine-tuned from gemma-2-2b-it, on HuggingFace$, enabling reproduction of our proposed framework using just a single 4090 GPU with 24GB memory, without any performance degradation.
Large language models (LLMs) excel in high-resource languages but struggle with low-resource languages (LRLs), particularly those spoken by minority communities in China, such as Tibetan, Uyghur, Kazakh, and Mongolian. To systematically track the progress in these languages, we introduce MiLiC-Eval, a benchmark designed for minority languages in China, featuring 24K instances across 9 tasks. MiLiC-Eval focuses on underrepresented writing systems and provides a fine-grained assessment of linguistic and problem-solving skills. Our evaluation reveals that LLMs perform poorly on syntax-intensive tasks and multi-script languages. We further demonstrate how MiLiC-Eval can help advance LRL research in handling diverse writing systems and understanding the process of language adaptation.
With the rise of large language models, neural text summarization has advanced significantly in recent years. However, even state-of-the-art models continue to rely heavily on high-quality human-annotated data for training and evaluation. Active learning is frequently used as an effective way to collect such datasets, especially when annotation resources are scarce. Active learning methods typically prioritize either uncertainty or diversity but have shown limited effectiveness in summarization, often being outperformed by random sampling. We present Diversity and Uncertainty Active Learning (DUAL), a novel algorithm that combines uncertainty and diversity to iteratively select and annotate samples that are both representative of the data distribution and challenging for the current model. DUAL addresses the selection of noisy samples in uncertainty-based methods and the limited exploration scope of diversity-based methods. Through extensive experiments with different summarization models and benchmark datasets, we demonstrate that DUAL consistently matches or outperforms the best performing strategies. Using visualizations and quantitative metrics, we provide valuable insights into the effectiveness and robustness of different active learning strategies, in an attempt to understand why these strategies haven't performed consistently in text summarization. Finally, we show that DUAL strikes a good balance between diversity and robustness.
Language model pretraining involves training on extensive corpora, where data quality plays a pivotal role. In this work, we aim to directly estimate the contribution of data during pretraining and select pretraining data in an efficient manner. Specifically, we draw inspiration from recent findings showing that compression efficiency (i.e., the normalized loss) of diverse models on certain text correlates strongly with their downstream performance, when the text domain aligns with the downstream benchmark (Huang et al., 2024). Building on this observation, we hypothesize that data on which model losses are predictive of downstream abilities also contribute effectively to learning. To leverage this insight, we introduce data selection based on data's Predictive strength (Preselect), a lightweight and efficient data selection method that requires training and deploying only a fastText-based scorer. Through comprehensive experiments with 1B and 3B parameter models, we demonstrate that models trained on 30B tokens selected with PreSelect surpasses the performance of a vanilla baseline trained on 300B tokens, achieving a 10x reduction in compute requirements. Furthermore, PreSelect significantly outperforms other competitive data selection baselines, such as DCLM and FineWeb-Edu on a scale of 3B models trained on 100B tokens. We open-source our trained data selection scorer along with the curated datasets at https://github.com/hkust-nlp/PreSelect.
Large language models (LLMs) exhibit exceptional performance across a wide range of tasks; however, their token-by-token autoregressive generation process significantly hinders inference speed. Speculative decoding presents a promising draft-then-verify framework that reduces generation latency while maintaining output distribution fidelity. Nevertheless, the draft model introduces additional computational overhead, becoming a performance bottleneck and increasing the time to first token (TTFT). Previous approaches to mitigate draft model overhead have primarily relied on heuristics and generally failed to match the quality of the draft language models. To address these challenges, we propose DuoDecoding, a novel approach that strategically deploys the draft and target models on the CPU and GPU respectively, enabling parallel decoding while preserving draft quality. Our method incorporates a hardware-aware optimal draft budget to minimize idle times and employs dynamic multi-sequence drafting to enhance draft quality. Extensive experiments across seven tasks show that DuoDecoding achieves up to 2.61x speedup in generation latency, while reducing TTFT to 83% of that in conventional speculative decoding. The Code is available at https://github.com/KaiLv69/DuoDecoding.
Personalized tool utilization is essential for aligning large language models (LLMs) with user preference in interaction scenarios with various tools. However, most of the current benchmarks primarily focus on either personalization of text generation or direct tool-utilizing, without considering both. In this work, we introduce a novel benchmark ETAPP for evaluating personalized tool invocation, establishing a sandbox environment, and a comprehensive dataset of 800 testing cases covering diverse user profiles. To improve the accuracy of our evaluation, we propose a key-point-based LLM evaluation method, mitigating biases in the LLM-as-a-judge system by manually annotating key points for each test case and providing them to LLM as the reference. Additionally, we evaluate the excellent LLMs and provide an in-depth analysis. Furthermore, we investigate the impact of different tool-invoking strategies on LLMs' personalization performance and the effects of fine-tuning in our task. The effectiveness of our preference-setting and key-point-based evaluation method is also validated. Our findings offer insights into improving personalized LLM agents. Our Code is available at https://github.com/hypasd-art/ETAPP.
Tool-Augmented Language Models (TALMs) leverage external APIs to answer user queries across various domains. However, existing benchmark datasets for TALM research often feature simplistic dialogues that do not reflect real-world scenarios, such as the need for models to ask clarifying questions or proactively call additional APIs when essential information is missing. To address these limitations, we construct and release ToolDial, a dataset comprising 11,111 multi-turn dialogues, with an average of 8.95 turns per dialogue, based on APIs from RapidAPI. ToolDial has two key characteristics. First, the dialogues incorporate 16 user and system actions (e.g., "Request", "Clarify", "Fail inform") to capture the rich dynamics of real-world interactions. Second, we simulate dialogues where the system requests necessary information from the user based on API documentation and seeks additional APIs if the user fails to provide the required information. To facilitate this process, we introduce a method for generating an API graph that represents input and output compatibility between APIs. Using ToolDial, we evaluate a suite of language models on their ability to predict correct actions and extract input parameter values for API calls from the dialogue history. Modern language models achieve accuracy scores below 70%, indicating substantial room for improvement. We release our dataset and code at https://github.com/holi-lab/ToolDial.
In this paper, we present our submission to SemEval-2025 Task 8: Question Answering over Tabular Data. This task, evaluated on the DataBench dataset, assesses Large Language Models' (LLMs) ability to answer natural language questions over structured data while addressing topic diversity and table size limitations in previous benchmarks. We propose a system that employs effective LLM prompting to translate natural language queries into executable code, enabling accurate responses, error correction, and interpretability. Our approach ranks first in both subtasks of the competition in the proprietary model category, significantly outperforming the organizer's baseline.
Recent advancements in Large Language Models (LLMs) have expanded their context windows to unprecedented lengths, sparking debates about the necessity of Retrieval-Augmented Generation (RAG). To address the fragmented evaluation paradigms and limited cases in existing Needle-in-a-Haystack (NIAH), this paper introduces U-NIAH, a unified framework that systematically compares LLMs and RAG methods in controlled long context settings. Our framework extends beyond traditional NIAH by incorporating multi-needle, long-needle, and needle-in-needle configurations, along with different retrieval settings, while leveraging the synthetic Starlight Academy dataset-a fictional magical universe-to eliminate biases from pre-trained knowledge. Through extensive experiments, we investigate three research questions: (1) performance trade-offs between LLMs and RAG, (2) error patterns in RAG, and (3) RAG's limitations in complex settings. Our findings show that RAG significantly enhances smaller LLMs by mitigating the "lost-in-the-middle" effect and improving robustness, achieving an 82.58% win-rate over LLMs. However, we observe that retrieval noise and reverse chunk ordering degrade performance, while surprisingly, advanced reasoning LLMs exhibit reduced RAG compatibility due to sensitivity to semantic distractors. We identify typical error patterns including omission due to noise, hallucination under high noise critical condition, and self-doubt behaviors. Our work not only highlights the complementary roles of RAG and LLMs, but also provides actionable insights for optimizing deployments. Code: https://github.com/Tongji-KGLLM/U-NIAH.
How human cognitive abilities are formed has long captivated researchers. However, a significant challenge lies in developing meaningful methods to measure these complex processes. With the advent of large language models (LLMs), which now rival human capabilities in various domains, we are presented with a unique testbed to investigate human cognition through a new lens. Among the many facets of cognition, one particularly crucial aspect is the concept of semantic size, the perceived magnitude of both abstract and concrete words or concepts. This study seeks to investigate whether LLMs exhibit similar tendencies in understanding semantic size, thereby providing insights into the underlying mechanisms of human cognition. We begin by exploring metaphorical reasoning, comparing how LLMs and humans associate abstract words with concrete objects of varying sizes. Next, we examine LLMs' internal representations to evaluate their alignment with human cognitive processes. Our findings reveal that multi-modal training is crucial for LLMs to achieve more human-like understanding, suggesting that real-world, multi-modal experiences are similarly vital for human cognitive development. Lastly, we examine whether LLMs are influenced by attention-grabbing headlines with larger semantic sizes in a real-world web shopping scenario. The results show that multi-modal LLMs are more emotionally engaged in decision-making, but this also introduces potential biases, such as the risk of manipulation through clickbait headlines. Ultimately, this study offers a novel perspective on how LLMs interpret and internalize language, from the smallest concrete objects to the most profound abstract concepts like love. The insights gained not only improve our understanding of LLMs but also provide new avenues for exploring the cognitive abilities that define human intelligence.
We report a peculiar observation that LLMs can assign hidden meanings to sequences that seem visually incomprehensible to humans: for example, a nonsensical phrase consisting of Byzantine musical symbols is recognized by gpt-4o as "say abracadabra". Moreover, some models can communicate using these sequences. Some of these meanings are hypothesized to partly originate in the massive spurious correlations due to BPE tokenization. We systematically evaluate the presence of such abilities in a wide range of models: Claude-3.5 Haiku, Claude-3.5 Sonnet (New and Old), Claude-3.7 Sonnet, gpt-4o mini, gpt-4o, o1-mini, Llama-3.3 70B, DeepSeek-R1-Distill-Lllama 70B, Qwen2.5 1.5B, Qwen2.5 32B, Phi-3.5 mini, GigaChat-Max, Vikhr-Llama-3.2 1B. We argue that this observation might have far-reaching consequences for both safety and security of the modern and future LLMs and systems that employ them. As an illustration, we show that applying this method in combination with simple templates is sufficient to jailbreak previous generation models, with ASR = 0.4 on gpt-4o mini. Our code and data artifacts are available at https://github.com/L3G5/llm-hidden-meanings
Typical evaluations of Large Language Models (LLMs) report a single metric per dataset, often representing the model's best-case performance under carefully selected settings. Unfortunately, this approach overlooks model robustness and reliability in real-world applications. For instance, simple paraphrasing of prompts on the MMLU-Pro dataset causes accuracy fluctuations of up to 10\%, while reordering answer choices in the AGIEval dataset results in accuracy differences of up to 6.1\%. While some studies discuss issues with LLM robustness, there is no unified or centralized framework for evaluating the robustness of language models. To address this gap and consolidate existing research on model robustness, we present SCORE ($\mathbf{S}$ystematic $\mathbf{CO}$nsistency and $\mathbf{R}$obustness $\mathbf{E}$valuation), a comprehensive framework for non-adversarial evaluation of LLMs. The SCORE framework evaluates models by repeatedly testing them on the same benchmarks in various setups to give a realistic estimate of their accuracy and consistency. We release the code publicly and start an LLM robustness leaderboard to facilitate further development and research.
Speculative Decoding (SD) enforces strict distributional equivalence to the target model, limiting potential speed ups as distributions of near-equivalence achieve comparable outcomes in many cases. Furthermore, enforcing distributional equivalence means that users are unable to trade deviations from the target model distribution for further inference speed gains. To address these limitations, we introduce Fuzzy Speculative Decoding (FSD) - a decoding algorithm that generalizes SD by accepting candidate tokens purely based on the divergences between the target and draft model distributions. By allowing for controlled divergence from the target model, FSD enables users to flexibly trade generation quality for inference speed. Across several benchmarks, our method is able to achieve significant runtime improvements of over 5 tokens per second faster than SD at only an approximate 2% absolute reduction in benchmark accuracy. In many cases, FSD is even able to match SD benchmark accuracy at over 2 tokens per second faster, demonstrating that distributional equivalence is not necessary to maintain target model performance.
Graph neural networks (GNNs) are conventionally trained on a per-domain, per-task basis. It creates a significant barrier in transferring the acquired knowledge to different, heterogeneous data setups. This paper introduces GraphBridge, a novel framework to enable knowledge transfer across disparate tasks and domains in GNNs, circumventing the need for modifications to task configurations or graph structures. Specifically, GraphBridge allows for the augmentation of any pre-trained GNN with prediction heads and a bridging network that connects the input to the output layer. This architecture not only preserves the intrinsic knowledge of the original model but also supports outputs of arbitrary dimensions. To mitigate the negative transfer problem, GraphBridge merges the source model with a concurrently trained model, thereby reducing the source bias when applied to the target domain. Our method is thoroughly evaluated across diverse transfer learning scenarios, including Graph2Graph, Node2Node, Graph2Node, and graph2point-cloud. Empirical validation, conducted over 16 datasets representative of these scenarios, confirms the framework's capacity for task- and domain-agnostic transfer learning within graph-like data, marking a significant advancement in the field of GNNs. Code is available at https://github.com/jujulili888/GraphBridge.
Intelligence is a crucial trait for species to find solutions within a limited number of trial-and-error attempts. Building on this idea, we introduce Survival Game as a framework to evaluate intelligence based on the number of failed attempts in a trial-and-error process. Fewer failures indicate higher intelligence. When the expectation and variance of failure counts are both finite, it signals the ability to consistently find solutions to new challenges, which we define as the Autonomous Level of intelligence. Using Survival Game, we comprehensively evaluate existing AI systems. Our results show that while AI systems achieve the Autonomous Level in simple tasks, they are still far from it in more complex tasks, such as vision, search, recommendation, and language. While scaling current AI technologies might help, this would come at an astronomical cost. Projections suggest that achieving the Autonomous Level for general tasks would require $10^{26}$ parameters. To put this into perspective, loading such a massive model requires so many H100 GPUs that their total value is $10^{7}$ times that of Apple Inc.'s market value. Even with Moore's Law, supporting such a parameter scale would take $70$ years. This staggering cost highlights the complexity of human tasks and the inadequacies of current AI technologies. To further investigate this phenomenon, we conduct a theoretical analysis of Survival Game and its experimental results. Our findings suggest that human tasks possess a criticality property. As a result, Autonomous Level requires a deep understanding of the task's underlying mechanisms. Current AI systems, however, do not fully grasp these mechanisms and instead rely on superficial mimicry, making it difficult for them to reach an autonomous level. We believe Survival Game can not only guide the future development of AI but also offer profound insights into human intelligence.