Publications and Preprints

2026

1. Preprint

   Beyond Monolingual Deep Research: Evaluating Agents and Retrievers with Cross-Lingual BrowseComp-Plus

   Yuheng Lu*, Qingcheng Zeng*, Heli Qi, Puxuan Yu, Fuheng Zhao, Rui Yang, Hitomi Yanaka, Naoto Yokoya, and Weihao Xuan

   2026
2. Preprint

   Knowledge Index of Noah’s Ark

   Sheng Jin, Minghao Liu, Yunze Xiao, Zeqi Zhou, Heli Qi, Yifan Yao, Meishu Song, Kaijing Ma, Xuan Zhang, Sicong Jiang, Yizhe Li, Ningshan Ma, Jie Wei, Ziniu Li, Minglai Yang, Bangya Liu, Yiming Liang, Xiao Fang, Qingcheng Zeng, Jiarui Liu, and 7 more authors

   2026
3. Preprint

   Dual-View Training for Instruction-Following Information Retrieval

   Qingcheng Zeng, Puxuan Yu, Aman Mehta, Fuheng Zhao, and Rajhans Samdani

   2026
4. Preprint

   Farther the Shift, Sparser the Representation: Analyzing OOD Mechanisms in LLMs

   Mingyu Jin, Yutong Yin, Jingcheng Niu, Qingcheng Zeng, Wujiang Xu, Mengnan Du, Wei Cheng, Zhaoran Wang, Tianlong Chen, and Dimitris N. Metaxas

   2026
5. ACL Findings

   Verified Critical Step Optimization for LLM Agents

   Mukai Li, Qingcheng Zeng, Tianqing Fang, Zhenwen Liang, Linfeng Song, Qi Liu, Haitao Mi, and Dong Yu

   In Findings of the Association for Computational Linguistics: ACL 2026, Jul 2026

   Abs

   As large language model agents tackle increasingly complex long-horizon tasks, effective post-training becomes critical. Prior work faces fundamental challenges: outcome-only rewards fail to precisely attribute credit to intermediate steps, estimated step-level rewards introduce systematic noise, and Monte Carlo sampling approaches for step reward estimation incur prohibitive computational cost. Inspired by findings that only a small fraction of high-entropy tokens drive effective RL for reasoning, we propose Critical Step Optimization (CSO), which focuses preference learning on verified critical steps—decision points where alternate actions demonstrably flip task outcomes from failure to success. Crucially, our method starts from failed policy trajectories rather than expert demonstrations, directly targeting the policy model’s weaknesses. We use a process reward model (PRM) to identify candidate critical steps, leverage expert models to propose high-quality alternatives, then continue execution from these alternatives using the policy model itself until task completion. Only alternatives that the policy successfully executes to correct outcomes are verified and used as DPO training data, ensuring both quality and policy reachability. This yields fine-grained, verifiable supervision at critical decisions while avoiding trajectory-level coarseness and step-level noise. Experiments on GAIA-Text-103 and XBench-DeepSearch show that CSO achieves 37% and 26% relative improvement over the SFT baseline and substantially outperforms other post-training methods, while requiring supervision at only 16% of trajectory steps. This demonstrates the effectiveness of selective verification-based learning for agent post-training.

6. ACL

   How to Improve LLMs’ Performance on Specific Languages: A Perspective on LLM-Derived Language Similarity

   Xinhe Shi, Qingcheng Zeng, Weihao Xuan, and Linchao Zhu

   In Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Jul 2026

   Abs

   Large language models (LLMs) exhibit uneven performance across languages. In language-specific applications, practitioners often rely on target-language corpora or cross-lingual transfer to achieve better performance. However, traditional linguistic typology, commonly used as a transfer language selection strategy in previous studies, may not align with LLM’s perception of language similarity. This work proposes **LLM-based language similarity** as a novel perspective for selecting effective fine-tuning languages. We construct a framework to quantify the similarity within each language pair through both the lenses of **language-specific performance patterns** and **cross-lingual transferability**, ultimately deriving three similarity score matrices. Moreover, we observe a counter-intuitive phenomenon: **super-additive transfer effect**, where fine-tuning on a certain language yields higher performance than fine-tuning directly on the target language. Additionally, due to the absence of an existing dataset meeting our experimental requirements, we construct and release **M4CQ-Pro** dataset, which features domain-diverse distribution of **135** tasks and content consistency across **31** languages (including over 20 medium- and low-resource languages), with 61518 manually reviewed high-quality questions per language. We evaluate our approach on representative multilingual LLMs and results show that all three LLM-based similarity measures effectively guide fine-tuning language selection, outperforming traditional linguistic similarity, with the integrated measure achieving the best results. Our approach provides not only **a novel perspective on language similarity**, but also **practical baselines for selecting fine-tuning languages**.

7. ACL

   On the Effect of Hyperparameters in Language Modeling for Computational Linguistics

   Ruoxi Ning, Yongpeng Zhu, Qingcheng Zeng, Tatsuki Kuribayashi, and Freda Shi

   In Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Jul 2026

   Abs

   Training language models and examining their linguistic behaviors have been a common protocol in computational linguistics for studying linguistic phenomena and modeling human language processing. However, work in this area is often limited to proof-of-concept demonstrations with arbitrary model configurations, without considering hyperparameter sensitivity, an important source of variation in model performance. In this work, we replicate three prior studies (Chang and Bergen, 2022; Hu et al., 2020b; Kuribayashi et al., 2024) with hyperparameters varied within a practical range, and show that modest hyperparameter changes can alter some qualitative conclusions about models’ linguistic abilities and even reverse the ranking of model performance. Our results highlight the risk that prior work may have reflected optimization artifacts rather than the genuine inductive biases of model classes, and that hyperparameter sensitivity should receive more attention as a factor that can meaningfully influence model behavior. We suggest future work to report the variation of performance across the configuration space to enhance the reliability and generalizability of conclusions. Code: https://github.com/compling-wat/tune-linguistic-lms.

8. ACL

   The Confidence Dichotomy: Analyzing and Mitigating Miscalibration in Tool-Use Agents

   Weihao Xuan*, Qingcheng Zeng*, Heli Qi, Yunze Xiao, Junjue Wang, and Naoto Yokoya

   In Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Jul 2026

   Abs

   Autonomous agents based on large language models (LLMs) are rapidly evolving to handle multi-turn tasks, but ensuring their trustworthiness remains a critical challenge. A fundamental pillar of this trustworthiness is calibration, which refers to an agent’s ability to express confidence that reliably reflects its actual performance. While calibration is well-established for static models, its dynamics in tool-integrated agentic workflows remain under-explored. In this work, we systematically investigate verbalized calibration in tool-use agents, revealing a fundamental confidence dichotomy driven by tool type. Specifically, our pilot study identifies that evidence tools (e.g., web search) systematically induce severe overconfidence due to inherent noise in retrieved information, while verification tools (e.g., code interpreters) can ground reasoning through deterministic feedback and mitigate miscalibration. To robustly improve calibration across tool types, we propose a reinforcement learning (RL) fine-tuning framework that jointly optimizes task accuracy and calibration, supported by a holistic benchmark of reward designs. We demonstrate that our trained agents not only achieve superior calibration but also exhibit robust generalization from local training environments to noisy web settings and to distinct domains such as mathematical reasoning. Our results highlight the necessity of domain-specific calibration strategies for tool-use agents. More broadly, this work establishes a foundation for building self-aware agents that can reliably communicate uncertainty in high-stakes, real-world deployments.

9. ACL Findings

   Code-Switching Information Retrieval: Benchmarks, Analysis, and the Limits of Current Retrievers

   Qingcheng Zeng*, Yuheng Lu*, Zeqi Zhou, Heli Qi, Puxuan Yu, Fuheng Zhao, Hitomi Yanaka, Weihao Xuan, and Naoto Yokoya

   In Findings of the Association for Computational Linguistics: ACL 2026, Jul 2026

   Abs

   Code-switching is a pervasive linguistic phenomenon in global communication, yet modern information retrieval systems remain predominantly designed for, and evaluated within, monolingual contexts. To bridge this critical disconnect, we present a holistic study dedicated to code-switching IR. We introduce CSR-L (Code-Switching Retrieval benchmark-Lite), constructing a dataset via human annotation to capture the authentic naturalness of mixed-language queries. Our evaluation across statistical, dense, and late-interaction paradigms reveals that code-switching acts as a fundamental performance bottleneck, degrading the effectiveness of even robust multilingual models. We demonstrate that this failure stems from substantial divergence in the embedding space between pure and code-switched text. Scaling this investigation, we propose CS-MTEB, a comprehensive benchmark covering 11 diverse tasks, where we observe performance declines of up to 27%. Finally, we show that standard multilingual techniques like vocabulary expansion are insufficient to resolve these deficits completely. These findings underscore the fragility of current systems and establish code-switching as a crucial frontier for future IR optimization.

10. Preprint

    RAPTOR: Ridge-Adaptive Logistic Probes

    Ziqi Gao, Yaotian Zhu, Qingcheng Zeng, Xu Zhao, Ziqing Wang, Feng Ruan, and Kaize Ding

    Jul 2026
11. EACL

    The Pragmatic Mind of Machines: Tracing the Emergence of Pragmatic Competence in Large Language Models

    Kefan Yu*, Qingcheng Zeng*†, Weihao Xuan, Wanxin Li, Jingyi Wu, and Rob Voigt

    In Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers), Mar 2026

    Abs

    Current large language models (LLMs) have demonstrated emerging capabilities in social intelligence tasks, including implicature resolution and theory-of-mind reasoning, both of which require substantial pragmatic understanding. However, how LLMs acquire this pragmatic competence throughout the training process remains poorly understood. In this work, we introduce ALTPRAG, a dataset grounded in the pragmatic concept of alternatives, to evaluate whether LLMs at different training stages can accurately infer nuanced speaker intentions. Each instance pairs two equally plausible yet pragmatically divergent continuations and requires the model to (i) infer the speaker’s intended meaning and (ii) explain when and why a speaker would choose one utterance over its alternative, thus directly probing pragmatic competence through contrastive reasoning. We systematically evaluate 22 LLMs across three key training stages: after pre-training, supervised fine-tuning (SFT), and preference optimization, to examine the development of pragmatic competence. Our results show that even base models exhibit notable sensitivity to pragmatic cues, which improves consistently with increases in model and data scale. Additionally, SFT and RLHF contribute further gains, particularly in cognitive-pragmatic scenarios. These findings highlight pragmatic competence as an emergent and compositional property of LLM training and offer new insights for aligning models with human communicative norms.

12. EACL Findings

    DeepSieve: Information Sieving via LLM-as-a-Knowledge-Router

    Minghao Guo, Qingcheng Zeng, Xujiang Zhao, Yanchi Liu, Wenchao Yu, Mengnan Du, Haifeng Chen, and Wei Cheng

    In Findings of the Association for Computational Linguistics: EACL 2026, Mar 2026

    Abs

    Large Language Models (LLMs) excel at many reasoning tasks but struggle with knowledge-intensive queries due to their inability to dynamically access up-to-date or domain-specific information. Retrieval-Augmented Generation (RAG) has emerged as a promising solution, enabling LLMs to ground their responses in external sources. However, existing RAG methods lack fine-grained control over both the query and source sides, often resulting in noisy retrieval and shallow reasoning. In this work, we introduce \textit\textbfDeepSieve, an agentic RAG framework that incorporates information sieving via LLM-as-a-knowledge-router. DeepSieve decomposes complex queries into structured sub-questions and recursively routes each to the most suitable knowledge source, filtering irrelevant information through a multi-stage distillation process. Our design emphasizes modularity, transparency, and adaptability, leveraging recent advances in agentic system design. Experiments on multi-hop QA tasks across heterogeneous sources demonstrate improved reasoning depth, retrieval precision, and interpretability over conventional RAG approaches.

13. Preprint

    NOVA: NOise-aware Verbal Confidence CAlibration for Robust Large Language Models in RAG Systems

    Jiayu Liu, Rui Wang, Qing Zong, Yumeng Wang, Cheng Qian, Qingcheng Zeng, Tianshi Zheng, Haochen Shi, Dadi Guo, Baixuan Xu, Chunyang Li, and Yangqiu Song

    Mar 2026
14. Preprint

    Toward Global Large Language Models in Medicine

    Rui Yang, Huitao Li, Weihao Xuan, Heli Qi, Xin Li, Kunyu Yu, Yingjian Chen, Rongrong Wang, Jacques Behmoaras, Tianxi Cai, Bibhas Chakraborty, Qingyu Chen, Lionel Tim-Ee Cheng, Marie-Louise Damwanza, Chido Dzinotyiwei, Aosong Feng, Chuan Hong, Yusuke Iwasawa, Yuhe Ke, Linah Kitala, and 30 more authors

    Mar 2026

2025

1. EMNLP

   Thinking Out Loud: Do Reasoning Models Know When They’re Right?

   Qingcheng Zeng*†, Weihao Xuan*, Leyang Cui, and Rob Voigt

   In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, Nov 2025

   Abs

   Large reasoning models (LRMs) have recently demonstrated impressive capabilities in complex reasoning tasks by leveraging increased test-time computation and exhibiting behaviors reminiscent of human-like self-reflection. While LRMs show a clear capacity for valuable self-reflection, how this ability interacts with other model behaviors remains underexplored. We investigate this connection by analyzing verbalized confidence, how models articulate their certainty, as a lens into the nature of self-reflection in LRMs. We find that supervised fine-tuning on reasoning traces (i.e., distillation) and reinforcement learning can improve verbalized calibration in reasoning-intensive settings in a progressive, laddered fashion. However, our results also indicate that reasoning models may possess a diminished awareness of their own knowledge boundaries, as evidenced by significantly lower "I don’t know" response rates on factuality benchmarks. Moreover, we examine the relationship between verbalized confidence and reasoning chains, finding that models tend to express higher confidence when providing shorter or less elaborate reasoning. Our findings highlight how reasoning-oriented training can enhance performance in reasoning-centric tasks while potentially incurring a reasoning tax, a cost reflected in the model’s reduced ability to accurately recognize the limits of its own knowledge in small-scale models. More broadly, our work showcases how this erosion of knowledge boundaries can compromise model faithfulness, as models grow more confident without a commensurate understanding of when they should abstain.

2. EMNLP

   Seeing is Believing, but How Much? A Comprehensive Analysis of Verbalized Calibration in Vision-Language Models

   Weihao Xuan*, Qingcheng Zeng*, Heli Qi, Junjue Wang, and Naoto Yokoya†

   In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, Nov 2025

   Abs

   Uncertainty quantification is essential for assessing the reliability and trustworthiness of modern AI systems. Among existing approaches, verbalized uncertainty, where models express their confidence through natural language, has emerged as a lightweight and interpretable solution in large language models (LLMs). However, its effectiveness in vision-language models (VLMs) remains insufficiently studied. In this work, we conduct a comprehensive evaluation of verbalized confidence in VLMs, spanning three model categories, four task domains, and three evaluation scenarios. Our results show that current VLMs often display notable miscalibration across diverse tasks and settings. Notably, visual reasoning models (i.e., thinking with images) consistently exhibit better calibration, suggesting that modality-specific reasoning is critical for reliable uncertainty estimation. To further address calibration challenges, we introduce Visual Confidence-Aware Prompting, a two-stage prompting strategy that improves confidence alignment in multimodal settings. Overall, our study highlights the inherent miscalibration in VLMs across modalities. More broadly, our findings underscore the fundamental importance of modality alignment and model faithfulness in advancing reliable multimodal systems.

3. EMNLP

   MMLU-ProX: A Multilingual Benchmark for Advanced Large Language Model Evaluation

   Weihao Xuan, Rui Yang, Heli Qi, Qingcheng Zeng, Yunze Xiao, Aosong Feng, Dairui Liu, Yun Xing, Junjue Wang, Fan Gao, Jinghui Lu, Yuang Jiang, Huitao Li, Xin Li, Kunyu Yu, Ruihai Dong, Shangding Gu, Yuekang Li, Xiaofei Xie, Felix Juefei-Xu, and 12 more authors

   In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, Nov 2025

   Abs

   Existing large language model (LLM) evaluation benchmarks primarily focus on English, while current multilingual tasks lack parallel questions that specifically assess cross-lingual reasoning abilities. This dual limitation makes it challenging to assess LLMs’ performance in the multilingual setting comprehensively. To fill this gap, we introduce MMLU-ProX, a comprehensive benchmark covering 29 languages, built on an English benchmark. Each language version consists of 11,829 identical questions, enabling direct cross-lingual comparisons. Additionally, to meet efficient evaluation needs, we provide a lite version containing 658 questions per language. To ensure the high quality of MMLU-ProX, we employ a rigorous development process that involves multiple powerful LLMs for translation, followed by expert review to ensure accurate expression, consistent terminology, and cultural relevance. Building on this, we systematically evaluate 36 state-of-the-art LLMs, including reasoning-enhanced and multilingual-optimized LLMs. The results reveal significant disparities in the multilingual capabilities of LLMs: While they perform well in high-resource languages, their performance declines markedly in low-resource languages, particularly for African languages. Through MMLU-ProX, we aim to advance the development of more inclusive AI systems and promote equitable access to technology across global contexts.

4. EMNLP

   Good Intentions Beyond ACL: Who Does NLP for Social Good, and Where?

   Grace LeFevre, Qingcheng Zeng, Adam Leif, Jason Jewell, Denis Peskoff, and Rob Voigt

   In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, Nov 2025

   Abs

   The social impact of Natural Language Processing (NLP) is increasingly important, with a rising community focus on initiatives related to NLP for Social Good (NLP4SG). Indeed, in recent years, almost 20% of all papers in the ACL Anthology address topics related to social good as defined by the UN Sustainable Development Goals (Aduato et al. 2023). In this study, we take an author- and venue-level perspective to map the landscape of NLP4SG, quantifying the proportion of work addressing social good concerns both within and beyond the ACL community, by both core ACL contributors and non-ACL authors. With this approach we discover two surprising facts about the landscape of NLP4SG. First, ACL authors are dramatically more likely to do work addressing social good concerns when publishing in venues outside of ACL. Second, the vast majority of publications using NLP techniques to address concerns of social good are done by non-ACL authors in venues outside of ACL. We discuss the implications of these findings on agenda-setting considerations for the ACL community related to NLP4SG.

5. CIKM

   Uncertainty Quantification for Multiple-Choice Questions is Just One-Token Deep

   Qingcheng Zeng*, Mingyu Jin*, Qinkai Yu, Zhenting Wang, Wenyue Hua, Guangyan Sun, Yanda Meng, Shiqing Ma, Qifan Wang, Felix Juefei-Xu, Fan Yang, Kaize Ding†, Ruixiang Tang, and Yongfeng Zhang

   In Proceedings of the 34th ACM International Conference on Information and Knowledge Management, Nov 2025

   Abs

   Multiple-choice question (MCQ) benchmarks such as MMLU and GPQA are widely used to assess the capabilities of large language models (LLMs). While accuracy remains the standard evaluation metric, recent work has introduced uncertainty quantification (UQ) methods, such as entropy, conformal prediction, and verbalized confidence, as complementary measures of model reliability and calibration. However, we find that these UQ methods, when applied to MCQ tasks, are unexpectedly fragile. Specifically, we show that fine-tuning a model on just 1,000 examples to adjust the probability of the first generated token, under the common prompting setup where the model is instructed to output only a single answer choice, can systematically distort a broad range of UQ methods across models, prompts, and domains, all while leaving answer accuracy unchanged. We validate this phenomenon through extensive experiments on five instruction-tuned LLMs, tested under standard prompting, zero-shot chain-of-thought reasoning, and a biomedical question answering setting. In all cases, models retain similar accuracy but exhibit significantly degraded calibration. These results suggest that current UQ practices for MCQs are "one-token deep", driven more by first-token decoding behavior than by any deeper representation of uncertainty, and are easily manipulated through minimal interventions. Our findings call for more robust and interpretable approaches to uncertainty estimation, particularly in structured formats like MCQs, where confidence signals are often reduced to token-level heuristics.

6. CIKM

   Fact or Facsimile? Evaluating the Factual Robustness of Modern Retrievers

   Haoyu Wu*, Qingcheng Zeng*, and Kaize Ding†

   In Proceedings of the 34th ACM International Conference on Information and Knowledge Management, Nov 2025

   Abs

   Dense retrievers and rerankers are central to retrieval-augmented generation (RAG) pipelines, where accurately retrieving factual information is crucial for maintaining system trustworthiness and defending against RAG poisoning. However, little is known about how much factual competence these components inherit or lose from the large language models (LLMs) they are based on. We pair 12 publicly released embedding checkpoints with their original base LLMs and evaluate both sets on a factuality benchmark. Across every model evaluated, the embedding variants achieve markedly lower accuracy than their bases, with absolute drops ranging from 12 to 43 percentage points (median 28 pts) and typical retriever accuracies collapsing into the 25-35% band versus the 60-70% attained by the generative models. This degradation intensifies under a more demanding condition: when the candidate pool per question is expanded from four options to one thousand, the strongest retriever’s top-1 accuracy falls from 33% to 26%, revealing acute sensitivity to distractor volume. Statistical tests further show that, for every embedding model, cosine-similarity scores between queries and correct completions are significantly higher than those for incorrect ones (p < 0.01), indicating decisions driven largely by surface-level semantic proximity rather than factual reasoning. To probe this weakness, we employed GPT-4.1 to paraphrase each correct completion, creating a rewritten test set that preserved factual truth while masking lexical cues, and observed that over two-thirds of previously correct predictions flipped to wrong, reducing overall accuracy to roughly one-third of its original level. Taken together, these findings reveal a systematic trade-off introduced by contrastive learning for retrievers: gains in semantic retrieval are paid for with losses in parametric factual knowledge, and the resulting models remain highly vulnerable to adversarial or even benign rephrasings. Our study underscores the need for retrieval objectives that balance similarity with factual fidelity to safeguard next-generation RAG systems against both misinformation and targeted attacks.

7. ACL

   Leveraging Human Production-Interpretation Asymmetries to Test LLM Cognitive Plausibility

   Suet-Ying Lam*, Qingcheng Zeng*†, Jingyi Wu, and Rob Voigt

   In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), Jul 2025

   Abs

   Whether large language models (LLMs) process language similarly to humans has been the subject of much theoretical and practical debate. We examine this question through the lens of the production-interpretation distinction found in human sentence processing and evaluate the extent to which instruction-tuned LLMs replicate this distinction. Using an empirically documented asymmetry between pronoun production and interpretation in humans for implicit causality verbs as a testbed, we find that some LLMs do quantitatively and qualitatively reflect human-like asymmetries between production and interpretation. We demonstrate that whether this behavior holds depends upon both model size-with larger models more likely to reflect human-like patterns and the choice of meta-linguistic prompts used to elicit the behavior. Our codes and results are available here.

8. XLLM

   Exploring Multilingual Probing in Large Language Models: A Cross-Language Analysis

   Daoyang Li, Haiyan Zhao, Qingcheng Zeng, and Mengnan Du

   In Proceedings of the 1st Joint Workshop on Large Language Models and Structure Modeling (XLLM 2025), Aug 2025

   Abs

   Probing techniques for large language models (LLMs) have primarily focused on English, overlooking the vast majority of other world’s languages. In this paper, we extend these probing methods to a multilingual context, investigating how LLMs encode linguistic structures across diverse languages. We conduct experiments on several open-source LLM models, analyzing probing accuracy, trends across layers, and similarities between probing vectors for multiple languages. Our key findings reveal: (1) a consistent performance gap between high-resource and low-resource languages, with high-resource languages achieving significantly higher probing accuracy; (2) divergent layer-wise accuracy trends, where high-resource languages show substantial improvement in deeper layers similar to English; and (3) higher representational similarities among high-resource languages, with low-resource languages demonstrating lower similarities both among themselves and with high-resource languages. These results provide insights into how linguistic structures are represented differently across languages in LLMs and emphasize the need for improved structure modeling for low-resource languages.

9. COLING

   Exploring Concept Depth: How Large Language Models Acquire Knowledge and Concept at Different Layers?

   Mingyu Jin*, Qinkai Yu*, Jingyuan Huang*, Qingcheng Zeng*, Zhenting Wang, Wenyue Hua, Haiyan Zhao, Kai Mei, Yanda Meng, Kaize Ding, Fan Yang, Mengnan Du, and Yongfeng Zhang

   In Proceedings of the 31st International Conference on Computational Linguistics, Jan 2025

   Abs

   Large language models (LLMs) have shown remarkable performances across a wide range of tasks. However, the mechanisms by which these models encode tasks of varying complexities remain poorly understood. In this paper, we explore the hypothesis that LLMs process concepts of varying complexities in different layers, introducing the idea of "Concept Depth" to suggest that more complex concepts are typically acquired in deeper layers. Specifically, we categorize concepts based on their level of abstraction, defining them in the order of increasing complexity within factual, emotional, and inferential tasks. We conduct extensive probing experiments using layer-wise representations across various LLM families (Gemma, LLaMA, Qwen) on various datasets spanning the three domains of tasks. Our findings reveal that models could efficiently conduct probing for simpler tasks in shallow layers, and more complex tasks typically necessitate deeper layers for accurate understanding. Additionally, we examine how external factors, such as adding noise to the input and quantizing the model weights, might affect layer-wise representations. Our findings suggest that these factors can impede the development of a conceptual understanding of LLMs until deeper layers are explored. We hope that our proposed concept and experimental insights will enhance the understanding of the mechanisms underlying LLMs. Our codes are available at https://github.com/Luckfort/CD.

10. Autism

    Pre-trained artificial intelligence language model represents pragmatic language variability central to autism and genetically related phenotypes

    Joseph CY Lau, Emily Landau, Qingcheng Zeng, Ruichun Zhang, Stephanie Crawford, Rob Voigt, and Molly Losh

    Autism, Jan 2025

    Abs

    Many individuals with autism experience challenges using language in social contexts (i.e., pragmatic language). Characterizing and understanding pragmatic variability is important to inform intervention strategies and the etiology of communication challenges in autism; however, current manual coding-based methods are often time and labor intensive, and not readily applied in ample sample sizes. This proof-of-concept methodological study employed an artificial intelligence pre-trained language model, Bidirectional Encoder Representations from Transformers, as a tool to address such challenges. We applied Bidirectional Encoder Representations from Transformers to computationally index pragmatic-related variability in autism and in genetically related phenotypes displaying pragmatic differences, namely, in parents of autistic individuals, fragile X syndrome, and FMR1 premutation. Findings suggest that without model fine-tuning, Bidirectional Encoder Representations from Transformers’s Next Sentence Prediction module was able to derive estimates that differentiate autistic from non-autistic groups. Moreover, such computational estimates correlated with manually coded characterization of pragmatic abilities that contribute to conversational coherence, not only in autism but also in the other genetically related phenotypes. This study represents a step forward in evaluating the efficacy of artificial intelligence language models for capturing clinically important pragmatic differences and variability related to autism, showcasing the potential of artificial intelligence to provide automatized, efficient, and objective tools for pragmatic characterization to help advance the field.

11. AACL Findings

    Sympathy over Polarization: A Computational Discourse Analysis of Social Media Posts about the July 2024 Trump Assassination Attempt

    Qingcheng Zeng*†, Guanhong Liu*, Zhaoqian Xue, Diego Ford, Rob Voigt, Loni Hagen, and Lingyao Li†

    In Proceedings of the 14th International Joint Conference on Natural Language Processing and the 4th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics, Dec 2025

    Abs

    On July 13, 2024, an assassination attempt was made on Republican presidential candidate Donald Trump during a rally in Pennsylvania. This event triggered widespread discourses on social media platforms. In this study, we analyze posts from X (formerly Twitter) collected during the week preceding and following the incident to examine the short-term impact of this political shock on public opinion and discourse. Our investigation is guided by three central research questions. First (RQ1), we assess how public stance toward Donald Trump evolved over time and varied across geographic regions. Second (RQ2), we apply causal inference methods to determine whether the assassination attempt itself significantly influenced public attitudes, independent of pre-existing political alignments. Third (RQ3), we conduct topic modeling to identify shifts in dominant themes of online discussions before and after the event. Integrating large language model-based stance detection, difference-in-differences estimation, and topic modeling, our findings reveal a marked surge in sympathetic responses toward Trump in the immediate aftermath of the attempt, suggesting a unifying effect that temporarily transcended ideological and regional divides.

12. SSRN

    A Century of Inflation Narratives

    Mourad Heddaya, Chenhao Tan, Rob Voigt, Qingcheng Zeng, and Alexander Zentefis

    Available at SSRN 5259107, Dec 2025
13. Preprint

    Position: The Hidden Costs and Measurement Gaps of Reinforcement Learning with Verifiable Rewards

    Aaron Tu, Weihao Xuan, Heli Qi, Xu Huang, Qingcheng Zeng, Shayan Talaei, Yijia Xiao, Peng Xia, Xiangru Tang, Yuchen Zhuang, Bing Hu, Hanqun Cao, Wenqi Shi, Tianang Leng, Rui Yang, Yingjian Chen, Ziqi Wang, Irene Li, Nan Liu, Huaxiu Yao, and 7 more authors

    Dec 2025
14. NeurIPS

    CARES: Comprehensive Evaluation of Safety and Adversarial Robustness in Medical LLMs

    Sijia Chen, Xiaomin Li, Mengxue Zhang, Eric Hanchen Jiang, Qingcheng Zeng, and Chen-Hsiang Yu

    Dec 2025
15. Preprint

    Large Language Models Are More Persuasive Than Incentivized Human Persuaders

    Philipp Schoenegger, Francesco Salvi, Jiacheng Liu, Xiaoli Nan, Ramit Debnath, Barbara Fasolo, Evelina Leivada, Gabriel Recchia, Fritz Günther, Ali Zarifhonarvar, Joe Kwon, Zahoor Ul Islam, Marco Dehnert, Daryl Y. H. Lee, Madeline G. Reinecke, David G. Kamper, Mert Kobaş, Adam Sandford, Jonas Kgomo, Luke Hewitt, and 20 more authors

    Dec 2025
16. NeurIPS

    ThinkBench: Dynamic Out-of-Distribution Evaluation for Robust LLM Reasoning

    Shulin Huang, Linyi Yang, Yan Song, Shuang Chen, Leyang Cui, Ziyu Wan, Qingcheng Zeng, Ying Wen, Kun Shao, Weinan Zhang, Jun Wang, and Yue Zhang

    Dec 2025
17. Preprint

    VeriGUI: Verifiable Long-Chain GUI Dataset

    Shunyu Liu, Minghao Liu, Huichi Zhou, Zhenyu Cui, Yang Zhou, Yuhao Zhou, Wendong Fan, Ge Zhang, Jiajun Shi, Weihao Xuan, Jiaxing Huang, Shuang Luo, Fang Wu, Heli Qi, Qingcheng Zeng, Ziqi Ren, Jialiang Gao, Jindi Lv, Junjie Wang, Aosong Feng, and 12 more authors

    Dec 2025
18. Preprint

    Toward Equitable Access: Leveraging Crowdsourced Reviews to Investigate Public Perceptions of Health Resource Accessibility

    Zhaoqian Xue, Guanhong Liu, Chong Zhang, Kai Wei, Qingcheng Zeng, Songhua Hu, Wenyue Hua, Lizhou Fan, Yongfeng Zhang, and Lingyao Li

    Dec 2025

2024

1. EMNLP

   Adaptive Axes: A Pipeline for In-domain Social Stereotype Analysis

   Qingcheng Zeng, Mingyu Jin, and Rob Voigt

   In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, Nov 2024

   Abs

   Prior work has explored the possibility of using the semantic information obtained from embedding representations to quantify social stereotypes, leveraging techniques such as word embeddings combined with a list of traits (Garg et al., 2018; Charlesworth et al., 2022) or semantic axes (An et al., 2018; Lucy et al., 2022). However, these approaches have struggled to fully capture the variability in stereotypes across different conceptual domains for the same social group (e.g., black in science, health, and art), in part because the identity of a word and the associations formed during pre-training can dominate its contextual representation (Field and Tsvetkov, 2019). This study explores the ability to recover stereotypes from the contexts surrounding targeted entities by utilizing state-of-the-art text embedding models and adaptive semantic axes enhanced by large language models (LLMs). Our results indicate that the proposed pipeline not only surpasses token-based methods in capturing in-domain framing but also effectively tracks stereotypes over time and along domain-specific semantic axes for in-domain texts. Our research highlights the potential of employing text embedding models to achieve a deeper understanding of nuanced social stereotypes.

2. WNU

   Causal Micro-Narratives

   Mourad Heddaya, Qingcheng Zeng, Alexander Zentefis, Rob Voigt, and Chenhao Tan

   In Proceedings of the 6th Workshop on Narrative Understanding, Nov 2024

   Abs

   We present a novel approach to classify causal micro-narratives from text. These narratives are sentence-level explanations of the cause(s) and/or effect(s) of a target subject. The approach requires only a subject-specific ontology of causes and effects, and we demonstrate it with an application to inflation narratives. Using a human-annotated dataset spanning historical and contemporary US news articles for training, we evaluate several large language models (LLMs) on this multi-label classification task. The best-performing model—a fine-tuned Llama 3.1 8B—achieves F1 scores of 0.87 on narrative detection and 0.71 on narrative classification. Comprehensive error analysis reveals challenges arising from linguistic ambiguity and highlights how model errors often mirror human annotator disagreements. This research establishes a framework for extracting causal micro-narratives from real-world data, with wide-ranging applications to social science research.

3. BioNLP

   KG-Rank: Enhancing Large Language Models for Medical QA with Knowledge Graphs and Ranking Techniques

   Rui Yang, Haoran Liu, Edison Marrese-Taylor, Qingcheng Zeng, Yuhe Ke, Wanxin Li, Lechao Cheng, Qingyu Chen, James Caverlee, Yutaka Matsuo, and Irene Li

   In Proceedings of the 23rd Workshop on Biomedical Natural Language Processing, Aug 2024

   Abs

   Large Language Models (LLMs) have significantly advanced healthcare innovation on generation capabilities. However, their application in real clinical settings is challenging due to potential deviations from medical facts and inherent biases. In this work, we develop an augmented LLM framework, KG-Rank, which leverages a medical knowledge graph (KG) with ranking and re-ranking techniques, aiming to improve free-text question-answering (QA) in the medical domain. Specifically, upon receiving a question, we initially retrieve triplets from a medical KG to gather factual information. Subsequently, we innovatively apply ranking methods to refine the ordering of these triplets, aiming to yield more precise answers. To the best of our knowledge, KG-Rank is the first application of ranking models combined with KG in medical QA specifically for generating long answers. Evaluation of four selected medical QA datasets shows that KG-Rank achieves an improvement of over 18% in the ROUGE-L score. Moreover, we extend KG-Rank to open domains, where it realizes a 14% improvement in ROUGE-L, showing the effectiveness and potential of KG-Rank.

4. ACL Findings

   Evaluating Large Language Models on Wikipedia-Style Survey Generation

   Fan Gao, Hang Jiang, Rui Yang, Qingcheng Zeng, Jinghui Lu, Moritz Blum, Tianwei She, Yuang Jiang, and Irene Li

   In Findings of the Association for Computational Linguistics: ACL 2024, Aug 2024

   Abs

   Educational materials such as survey articles in specialized fields like computer science traditionally require tremendous expert inputs and are therefore expensive to create and update. Recently, Large Language Models (LLMs) have achieved significant success across various general tasks. However, their effectiveness and limitations in the education domain are yet to be fully explored. In this work, we examine the proficiency of LLMs in generating succinct survey articles specific to the niche field of NLP in computer science, focusing on a curated list of 99 topics. Automated benchmarks reveal that GPT-4 surpasses its predecessors, inluding GPT-3.5, PaLM2, and LLaMa2 by margins ranging from 2% to 20% in comparison to the established ground truth. We compare both human and GPT-based evaluation scores and provide in-depth analysis. While our findings suggest that GPT-created surveys are more contemporary and accessible than human-authored ones, certain limitations were observed. Notably, GPT-4, despite often delivering outstanding content, occasionally exhibited lapses like missing details or factual errors. At last, we compared the rating behavior between humans and GPT-4 and found systematic bias in using GPT evaluation.

5. NEJM AI

   Clinical Text Datasets for Medical Artificial Intelligence and Large Language Models — A Systematic Review

   Jiageng Wu, Xiaocong Liu, Minghui Li, Wanxin Li, Zichang Su, Shixu Lin, Lucas Garay, Zhiyun Zhang, Yujie Zhang, Qingcheng Zeng, Jie Shen, Changzheng Yuan, and Jie Yang

   NEJM AI, Aug 2024

   Abs

   Privacy and ethical considerations limit access to large-scale clinical datasets, particularly clinical text data, which contain extensive and diverse information and serve as the foundation for building clinical large language models (LLMs). The limited accessibility of clinical text data impedes the development of clinical artificial intelligence systems and hampers research participation from resource-poor regions and medical institutions, thereby exacerbating health care disparities. In this review, we conduct a global review to identify publicly available clinical text datasets and elaborate on their accessibility, diversity, and usability for clinical LLMs.

6. JMIR

   Ascle—A Python Natural Language Processing Toolkit for Medical Text Generation: Development and Evaluation Study

   Rui Yang*, Qingcheng Zeng*, Keen You, Yujie Qiao, Lucas Huang, Chia-Chun Hsieh, Benjamin Rosand, Jeremy Goldwasser, Amisha Dave, Tiarnan Keenan, Yuhe Ke, Chuan Hong, Nan Liu, Emily Chew, Dragomir Radev, Zhiyong Lu, Hua Xu, Qingyu Chen, and Irene Li†

   J Med Internet Res, Oct 2024

   Abs

   This study aims to describe the development and preliminary evaluation of Ascle. Ascle is tailored for biomedical researchers and clinical staff with an easy-to-use, all-in-one solution that requires minimal programming expertise. For the first time, Ascle provides 4 advanced and challenging generative functions: question-answering, text summarization, text simplification, and machine translation. In addition, Ascle integrates 12 essential NLP functions, along with query and search capabilities for clinical databases.

2023

1. IJCAI

   GreenPLM: Cross-lingual Transfer of Monolingual Pre-trained Language Models at Almost No Cost

   Qingcheng Zeng*, Lucas Garay*, Peilin Zhou*, Dading Chong, Yining Hua, Jiageng Wu, Yikang Pan, Han Zhou, Rob Voigt, and Jie Yang†

   In Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence, Oct 2023

   Abs

   Large pre-trained models have revolutionized natural language processing (NLP) research and applications, but high training costs and limited data resources have prevented their benefits from being shared equally amongst speakers of all the world’s languages. To address issues of cross-linguistic access to such models and reduce energy consumption for sustainability during large-scale model training, this study proposes an effective and energy-efficient framework called Green-PLM that uses bilingual lexicons to directly "translate" pre-trained language models of one language into another at almost no additional cost. We validate this approach in 18 languages’ BERT models and show that this framework is comparable to, if not better than, other heuristics with high training costs. In addition, given lightweight continued pretraining on limited data where available, this framework outperforms the original monolingual language models in six out of seven tested languages with up to 200x less pre-training efforts. Aiming at the Leave No One Behind Principle (LNOB), our approach manages to reduce inequalities between languages and energy consumption greatly. We make our codes and models publicly available at https://github.com/qcznlp/GreenPLMs.

2. ACL Findings

   Large Language Models Are Partially Primed in Pronoun Interpretation

   Suet-Ying Lam*, Qingcheng Zeng*†, Kexun Zhang*, Chenyu You, and Rob Voigt

   In Findings of the Association for Computational Linguistics: ACL 2023, Jul 2023

   Abs

   While a large body of literature suggests that large language models (LLMs) acquire rich linguistic representations, little is known about whether they adapt to linguistic biases in a human-like way. The present study probes this question by asking whether LLMs display human-like referential biases using stimuli and procedures from real psycholinguistic experiments. Recent psycholinguistic studies suggest that humans adapt their referential biases with recent exposure to referential patterns; closely replicating three relevant psycholinguistic experiments from Johnson & Arnold (2022) in an in-context learning (ICL) framework, we found that InstructGPT adapts its pronominal interpretations in response to the frequency of referential patterns in the local discourse, though in a limited fashion: adaptation was only observed relative to syntactic but not semantic biases. By contrast, FLAN-UL2 fails to generate meaningful patterns. Our results provide further evidence that contemporary LLMs discourse representations are sensitive to syntactic patterns in the local context but less so to semantic patterns. Our data and code are available at https://github.com/zkx06111/llm_priming.

3. ICASSP

   Masked Spectrogram Prediction for Self-Supervised Audio Pre-Training

   Dading Chong, Helin Wang, Peilin Zhou, and Qingcheng Zeng

   In ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Jul 2023

2022

1. COLING

   A Survey in Automatic Irony Processing: Linguistic, Cognitive, and Multi-X Perspectives

   Qingcheng Zeng, and An-Ran Li

   In Proceedings of the 29th International Conference on Computational Linguistics, Oct 2022

   Abs

   Irony is a ubiquitous figurative language in daily communication. Previously, many researchers have approached irony from linguistic, cognitive science, and computational aspects. Recently, some progress have been witnessed in automatic irony processing due to the rapid development in deep neural models in natural language processing (NLP). In this paper, we will provide a comprehensive overview of computational irony, insights from linguisic theory and cognitive science, as well as its interactions with downstream NLP tasks and newly proposed multi-X irony processing perspectives.

2. Interspeech

   Low-resource Accent Classification in Geographically-proximate Settings: A Forensic and Sociophonetics Perspective

   Qingcheng Zeng, Dading Chong, Peilin Zhou, and Jie Yang

   In Interspeech 2022, Oct 2022
3. Interspeech

   Calibrate and Refine! A Novel and Agile Framework for ASR Error Robust Intent Detection

   Peilin Zhou, Dading Chong, Helin Wang, and Qingcheng Zeng

   In Interspeech 2022, Oct 2022
4. Preprint

   Filter and Evolve: Progressive Pseudo Label Refining for Semi-Supervised Automatic Speech Recognition

   Zezhong Jin, Dading Zhong, Xiao Song, Zhaoyi Liu, Naipeng Ye, and Qingcheng Zeng

   Oct 2022