Caiqi Zhang

Ruosong Ye, Caiqi Zhang, Runhui Wang et al. · cambridge

The emergence of large-scale pre-trained language models has revolutionized various AI research domains. Transformers-based Large Language Models (LLMs) have gradually replaced CNNs and RNNs to unify fields of computer vision and natural language processing. Compared with independent data samples such as images, videos or texts, graphs usually contain rich structural and relational information. Meanwhile, language, especially natural language, being one of the most expressive mediums, excels in describing complex structures. However, existing work on incorporating graph problems into the generative language modeling framework remains very limited. Considering the rising prominence of LLMs, it becomes essential to explore whether LLMs can also replace GNNs as the foundation model for graphs. In this paper, we propose InstructGLM (Instruction-finetuned Graph Language Model) with highly scalable prompts based on natural language instructions. We use natural language to describe multi-scale geometric structure of the graph and then instruction finetune an LLM to perform graph tasks, which enables Generative Graph Learning. Our method surpasses all GNN baselines on ogbn-arxiv, Cora and PubMed datasets, underscoring its effectiveness and sheds light on generative LLMs as new foundation model for graph machine learning. Our code is open-sourced at https://github.com/agiresearch/InstructGLM.

Lekang Jiang, Caiqi Zhang, Farimah Poursafaei et al. · cambridge

Recently, Graph Neural Networks (GNNs) have shown promising performance in tasks on dynamic graphs such as node classification, link prediction and graph regression. However, few work has studied the temporal edge regression task which has important real-world applications. In this paper, we explore the application of GNNs to edge regression tasks in both static and dynamic settings, focusing on predicting food and agriculture trade values between nations. We introduce three simple yet strong baselines and comprehensively evaluate one static and three dynamic GNN models using the UN Trade dataset. Our experimental results reveal that the baselines exhibit remarkably strong performance across various settings, highlighting the inadequacy of existing GNNs. We also find that TGN outperforms other GNN models, suggesting TGN is a more appropriate choice for edge regression tasks. Moreover, we note that the proportion of negative edges in the training samples significantly affects the test performance. The companion source code can be found at: https://github.com/scylj1/GNN_Edge_Regression.

Te-Lin Wu, Caiqi Zhang, Qingyuan Hu et al. · cambridge

The ability to infer pre- and postconditions of an action is vital for comprehending complex instructions, and is essential for applications such as autonomous instruction-guided agents and assistive AI that supports humans to perform physical tasks. In this work, we propose a task dubbed action condition inference, and collecting a high-quality, human annotated dataset of preconditions and postconditions of actions in instructional manuals. We propose a weakly supervised approach to automatically construct large-scale training instances from online instructional manuals, and curate a densely human-annotated and validated dataset to study how well the current NLP models can infer action-condition dependencies in the instruction texts. We design two types of models differ by whether contextualized and global information is leveraged, as well as various combinations of heuristics to construct the weak supervisions. Our experimental results show a >20% F1-score improvement with considering the entire instruction contexts and a >6% F1-score benefit with the proposed heuristics.

Beiduo Chen, Tiancheng Hu, Caiqi Zhang et al. · cambridge

Reasoning-tuned LLMs utilizing long Chain-of-Thought (CoT) excel at single-answer tasks, yet their ability to model Human Label Variation--which requires capturing probabilistic ambiguity rather than resolving it--remains underexplored. We investigate this through systematic disentanglement experiments on distribution-based tasks, employing Cross-CoT experiments to isolate the effect of reasoning text from intrinsic model priors. We observe a distinct "decoupled mechanism": while CoT improves distributional alignment, final accuracy is dictated by CoT content (99% variance contribution), whereas distributional ranking is governed by model priors (over 80%). Step-wise analysis further shows that while CoT's influence on accuracy grows monotonically during the reasoning process, distributional structure is largely determined by LLM's intrinsic priors. These findings suggest that long CoT serves as a decisive LLM decision-maker for the top option but fails to function as a granular distribution calibrator for ambiguous tasks.

Xiaochen Zhu, Caiqi Zhang, Yizhou Chi et al. · cambridge

Multi-agent debate (MAD) is widely used to improve large language model (LLM) performance through test-time scaling, yet recent work shows that vanilla MAD often underperforms simple majority vote despite higher computational cost. Studies show that, under homogeneous agents and uniform belief updates, debate preserves expected correctness and therefore cannot reliably improve outcomes. Drawing on findings from human deliberation and collective decision-making, we identify two key mechanisms missing from vanilla MAD: (i) diversity of initial viewpoints and (ii) explicit, calibrated confidence communication. We propose two lightweight interventions. First, a diversity-aware initialisation that selects a more diverse pool of candidate answers, increasing the likelihood that a correct hypothesis is present at the start of debate. Second, a confidence-modulated debate protocol in which agents express calibrated confidence and condition their updates on others' confidence. We show theoretically that diversity-aware initialisation improves the prior probability of MAD success without changing the underlying update dynamics, while confidence-modulated updates enable debate to systematically drift to the correct hypothesis. Empirically, across six reasoning-oriented QA benchmarks, our methods consistently outperform vanilla MAD and majority vote. Our results connect human deliberation with LLM-based debate and demonstrate that simple, principled modifications can substantially enhance debate effectiveness.

Caiqi Zhang, Ruihan Yang, Xiaochen Zhu et al.

While confidence estimation is a promising direction for mitigating hallucinations in Large Language Models (LLMs), current research dominantly focuses on single-turn settings. The dynamics of model confidence in multi-turn conversations, where context accumulates and ambiguity is progressively resolved, remain largely unexplored. Reliable confidence estimation in multi-turn settings is critical for many downstream applications, such as autonomous agents and human-in-the-loop systems. This work presents the first systematic study of confidence estimation in multi-turn interactions, establishing a formal evaluation framework grounded in two key desiderata: per-turn calibration and monotonicity of confidence as more information becomes available. To facilitate this, we introduce novel metrics, including a length-normalized Expected Calibration Error (InfoECE), and a new "Hinter-Guesser" paradigm for generating controlled evaluation datasets. Our experiments reveal that widely-used confidence techniques struggle with calibration and monotonicity in multi-turn dialogues. We propose P(Sufficient), a logit-based probe that achieves comparatively better performance, although the task remains far from solved. Our work provides a foundational methodology for developing more reliable and trustworthy conversational agents.

Yuan He, Bailan He, Zifeng Ding et al.

Understanding and mitigating hallucinations in Large Language Models (LLMs) is crucial for ensuring reliable content generation. While previous research has primarily focused on "when" LLMs hallucinate, our work explains "why" and directly links model behaviour to the pre-training data that forms their prior knowledge. Specifically, we demonstrate that an asymmetry exists in the recognition of logically equivalent facts, which can be attributed to frequency discrepancies of entities appearing as subjects versus objects. Given that most pre-training datasets are inaccessible, we leverage the fully open-source OLMo series by indexing its Dolma dataset to estimate entity frequencies. Using relational facts (represented as triples) from Wikidata5M, we construct probing datasets to isolate this effect. Our experiments reveal that facts with a high-frequency subject and a low-frequency object are better recognised than their inverse, despite their logical equivalence. The pattern reverses in low-to-high frequency settings, and no statistically significant asymmetry emerges when both entities are high-frequency. These findings highlight the influential role of pre-training data in shaping model predictions and provide insights for inferring the characteristics of pre-training data in closed or partially closed LLMs.

Runyang You, Hongru Cai, Caiqi Zhang et al.

LLM-as-a-Judge has revolutionized AI evaluation by leveraging large language models for scalable assessments. However, as evaluands become increasingly complex, specialized, and multi-step, the reliability of LLM-as-a-Judge has become constrained by inherent biases, shallow single-pass reasoning, and the inability to verify assessments against real-world observations. This has catalyzed the transition to Agent-as-a-Judge, where agentic judges employ planning, tool-augmented verification, multi-agent collaboration, and persistent memory to enable more robust, verifiable, and nuanced evaluations. Despite the rapid proliferation of agentic evaluation systems, the field lacks a unified framework to navigate this shifting landscape. To bridge this gap, we present the first comprehensive survey tracing this evolution. Specifically, we identify key dimensions that characterize this paradigm shift and establish a developmental taxonomy. We organize core methodologies and survey applications across general and professional domains. Furthermore, we analyze frontier challenges and identify promising research directions, ultimately providing a clear roadmap for the next generation of agentic evaluation.

Caiqi Zhang, Fangyu Liu, Marco Basaldella et al. · cambridge

Large Language Models (LLMs) have demonstrated remarkable capability in a variety of NLP tasks. However, LLMs are also prone to generate nonfactual content. Uncertainty Quantification (UQ) is pivotal in enhancing our understanding of a model's confidence on its generation, thereby aiding in the mitigation of nonfactual outputs. Existing research on UQ predominantly targets short text generation, typically yielding brief, word-limited responses. However, real-world applications frequently necessitate much longer responses. Our study first highlights the limitations of current UQ methods in handling long text generation. We then introduce \textsc{Luq} and its two variations, a series of novel sampling-based UQ approaches specifically designed for long text. Our findings reveal that \textsc{Luq} outperforms existing baseline methods in correlating with the model's factuality scores (negative coefficient of -0.85 observed for Gemini Pro). To further improve the factuality of LLM responses, we propose \textsc{Luq-Ensemble}, a method that ensembles responses from multiple models and selects the response with the lowest uncertainty. The ensembling method greatly improves the response factuality upon the best standalone LLM.

Caiqi Zhang, Zhijiang Guo, Andreas Vlachos · cambridge

This paper investigates the potential benefits of language-specific fact-checking models, focusing on the case of Chinese. We first demonstrate the limitations of translation-based methods and multilingual large language models (e.g., GPT-4), highlighting the need for language-specific systems. We further propose a Chinese fact-checking system that can better retrieve evidence from a document by incorporating context information. To better analyze token-level biases in different systems, we construct an adversarial dataset based on the CHEF dataset, where each instance has large word overlap with the original one but holds the opposite veracity label. Experimental results on the CHEF dataset and our adversarial dataset show that our proposed method outperforms translation-based methods and multilingual LLMs and is more robust toward biases, while there is still large room for improvement, emphasizing the importance of language-specific fact-checking systems.

Ruihan Yang, Caiqi Zhang, Zhisong Zhang et al. · cambridge

While Large Language Models (LLMs) demonstrate impressive capabilities, they still struggle with generating factually incorrect content (i.e., hallucinations). A promising approach to mitigate this issue is enabling models to express uncertainty when unsure. Previous research on uncertainty modeling has primarily focused on short-form QA, but realworld applications often require much longer responses. In this work, we introduce the task of Long-form Generation with Uncertainty(LoGU). We identify two key challenges: Uncertainty Suppression, where models hesitate to express uncertainty, and Uncertainty Misalignment, where models convey uncertainty inaccurately. To tackle these challenges, we propose a refinement-based data collection framework and a two-stage training pipeline. Our framework adopts a divide-and-conquer strategy, refining uncertainty based on atomic claims. The collected data are then used in training through supervised fine-tuning (SFT) and direct preference optimization (DPO) to enhance uncertainty expression. Extensive experiments on three long-form instruction following datasets show that our method significantly improves accuracy, reduces hallucinations, and maintains the comprehensiveness of responses.

Caiqi Zhang, Ruihan Yang, Zhisong Zhang et al. · cambridge

Large language models (LLMs) often suffer from hallucinations, posing significant challenges for real-world applications. Confidence calibration, as an effective indicator of hallucination, is thus essential to enhance the trustworthiness of LLMs. Prior work mainly focuses on short-form tasks using a single response-level score (macro calibration), which is insufficient for long-form outputs that may contain both accurate and inaccurate claims. In this work, we systematically study atomic calibration, which evaluates factuality calibration at a fine-grained level by decomposing long responses into atomic claims. We further categorize existing confidence elicitation methods into discriminative and generative types, and propose two new confidence fusion strategies to improve calibration. Our experiments demonstrate that LLMs exhibit poorer calibration at the atomic level during long-form generation. More importantly, atomic calibration uncovers insightful patterns regarding the alignment of confidence methods and the changes of confidence throughout generation. This sheds light on future research directions for confidence estimation in long-form generation.

Caiqi Zhang, Xiaochen Zhu, Chengzu Li et al. · cambridge

Hallucination remains a major challenge for the safe and trustworthy deployment of large language models (LLMs) in factual content generation. Prior work has explored confidence estimation as an effective approach to hallucination detection, but often relies on post-hoc self-consistency methods that require computationally expensive sampling. Verbalized confidence offers a more efficient alternative, but existing approaches are largely limited to short-form question answering (QA) tasks and do not generalize well to open-ended generation. In this paper, we propose LoVeC (Long-form Verbalized Confidence), an on-the-fly verbalized confidence estimation method for long-form generation. Specifically, we use reinforcement learning (RL) to train LLMs to append numerical confidence scores to each generated statement, serving as a direct and interpretable signal of the factuality of generation. Our experiments consider both on-policy and off-policy RL methods, including DPO, ORPO, and GRPO, to enhance the model calibration. We introduce two novel evaluation settings, free-form tagging and iterative tagging, to assess different verbalized confidence estimation methods. Experiments on three long-form QA datasets show that our RL-trained models achieve better calibration and generalize robustly across domains. Also, our method is highly efficient, as it only requires adding a few tokens to the output being decoded.

Artem Shelmanov, Ekaterina Fadeeva, Akim Tsvigun et al.

Large Language Models (LLMs) have the tendency to hallucinate, i.e., to sporadically generate false or fabricated information. This presents a major challenge, as hallucinations often appear highly convincing and users generally lack the tools to detect them. Uncertainty quantification (UQ) provides a framework for assessing the reliability of model outputs, aiding in the identification of potential hallucinations. In this work, we introduce pre-trained UQ heads: supervised auxiliary modules for LLMs that substantially enhance their ability to capture uncertainty compared to unsupervised UQ methods. Their strong performance stems from the powerful Transformer architecture in their design and informative features derived from LLM attention maps. Experimental evaluation shows that these heads are highly robust and achieve state-of-the-art performance in claim-level hallucination detection across both in-domain and out-of-domain prompts. Moreover, these modules demonstrate strong generalization to languages they were not explicitly trained on. We pre-train a collection of UQ heads for popular LLM series, including Mistral, Llama, and Gemma 2. We publicly release both the code and the pre-trained heads.

Ruihan Yang, Caiqi Zhang, Zhisong Zhang et al. · cambridge

Large Language Models (LLMs) are prone to hallucination, particularly in long-form generations. A promising direction to mitigate hallucination is to teach LLMs to express uncertainty explicitly when they lack sufficient knowledge. However, existing work lacks direct and fair evaluation of LLMs' ability to express uncertainty effectively in long-form generation. To address this gap, we first introduce UNCLE, a benchmark designed to evaluate uncertainty expression in both long- and short-form question answering (QA). UNCLE covers five domains and includes more than 1,000 entities, each with paired short- and long-form QA items. Our dataset is the first to directly link short- and long-form QA through aligned questions and gold-standard answers. Along with UNCLE, we propose a suite of new metrics to assess the models' capabilities to selectively express uncertainty. We then demonstrate that current models fail to convey uncertainty appropriately in long-form generation. We further explore both prompt-based and training-based methods to improve models' performance, with the training-based methods yielding greater gains. Further analysis of alignment gaps between short- and long-form uncertainty expression highlights promising directions for future research using UNCLE.

Xiaochen Zhu, Caiqi Zhang, Tom Stafford et al. · cambridge

The conformity effect describes the tendency of individuals to align their responses with the majority. Studying this bias in large language models (LLMs) is crucial, as LLMs are increasingly used in various information-seeking and decision-making tasks as conversation partners to improve productivity. Thus, conformity to incorrect responses can compromise their effectiveness. In this paper, we adapt psychological experiments to examine the extent of conformity in popular LLMs. Our findings reveal that all tested models exhibit varying levels of conformity toward the majority, regardless of their initial choice or correctness, across different knowledge domains. Notably, we are the first to show that LLMs are more likely to conform when they are more uncertain in their own prediction. We further explore factors that influence conformity, such as training paradigms and input characteristics, finding that instruction-tuned models are less susceptible to conformity, while increasing the naturalness of majority tones amplifies conformity. Finally, we propose two interventions, Devil's Advocate and Question Distillation, to mitigate conformity, providing insights into building more robust language models.

Wenyan Li, Raphael Tang, Chengzu Li et al. · cambridge

Vision--language models (VLMs) often process visual inputs through a pretrained vision encoder, followed by a projection into the language model's embedding space via a connector component. While crucial for modality fusion, the potential information loss induced by this projection step and its direct impact on model capabilities remain understudied. We introduce two complementary approaches to examine and quantify this loss by analyzing the latent representation space. First, we evaluate semantic information preservation by analyzing changes in k-nearest neighbor relationships between image representations, before and after projection. Second, we directly measure information loss by reconstructing visual embeddings from the projected representation, localizing loss at an image patch level. Experiments reveal that connectors substantially distort the local geometry of visual representations, with k-nearest neighbors diverging by 40--60\% post-projection, correlating with degradation in retrieval performance. The patch-level embedding reconstruction provides interpretable insights for model behavior on visually grounded question-answering tasks, finding that areas of high information loss reliably predict instances where models struggle.

Ej Zhou, Caiqi Zhang, Tiancheng Hu et al. · cambridge

Confidence calibration, the alignment of a model's predicted confidence with its actual accuracy, is crucial for the reliable deployment of Large Language Models (LLMs). However, this critical property remains largely under-explored in multilingual contexts. In this work, we conduct the first large-scale, systematic studies of multilingual calibration across six model families and over 100 languages, revealing that non-English languages suffer from systematically worse calibration. To diagnose this, we investigate the model's internal representations and find that the final layer, biased by English-centric training, provides a poor signal for multilingual confidence. In contrast, our layer-wise analysis uncovers a key insight that late-intermediate layers consistently offer a more reliable and better-calibrated signal. Building on this, we introduce a suite of training-free methods, including Language-Aware Confidence Ensemble (LACE), which adaptively selects an optimal ensemble of layers for each specific language. Our study highlights the hidden costs of English-centric alignment and offer a new path toward building more globally equitable and trustworthy LLMs by looking beyond the final layer.

Sanhanat Sivapiromrat, Caiqi Zhang, Marco Basaldella et al. · cambridge

Recent studies have shown that Large Language Models (LLMs) are vulnerable to data poisoning attacks, where malicious training examples embed hidden behaviours triggered by specific input patterns. However, most existing works assume a phrase and focus on the attack's effectiveness, offering limited understanding of trigger mechanisms and how multiple triggers interact within the model. In this paper, we present a framework for studying poisoning in LLMs. We show that multiple distinct backdoor triggers can coexist within a single model without interfering with each other, enabling adversaries to embed several triggers concurrently. Using multiple triggers with high embedding similarity, we demonstrate that poisoned triggers can achieve robust activation even when tokens are substituted or separated by long token spans. Our findings expose a broader and more persistent vulnerability surface in LLMs. To mitigate this threat, we propose a post hoc recovery method that selectively retrains specific model components based on a layer-wise weight difference analysis. Our method effectively removes the trigger behaviour with minimal parameter updates, presenting a practical and efficient defence against multi-trigger poisoning.

Caiqi Zhang, Chang Shu, Ehsan Shareghi et al. · cambridge

Confidence estimation is essential for the reliable deployment of large language models (LLMs). Existing methods are primarily designed for factual QA tasks and often fail to generalize to reasoning tasks. To address this gap, we propose a set of training-free, graph-based confidence estimation methods tailored to reasoning tasks. Our approach models reasoning paths as directed graphs and estimates confidence by exploiting graph properties such as centrality, path convergence, and path weighting. Experiments with two LLMs on three reasoning datasets demonstrate improved confidence estimation and enhanced performance on two downstream tasks.

Lekang Jiang, Caiqi Zhang, Pascal A Scherz et al.

Large language models (LLMs) have shown exceptional performance across various text generation tasks but remain under-explored in the patent domain, which offers highly structured and precise language. This paper constructs a dataset to investigate the performance of current LLMs in patent claim generation. Our results demonstrate that generating claims based on patent descriptions outperforms previous research relying on abstracts. Interestingly, current patent-specific LLMs perform much worse than state-of-the-art general LLMs, highlighting the necessity for future research on in-domain LLMs. We also find that LLMs can produce high-quality first independent claims, but their performances markedly decrease for subsequent dependent claims. Moreover, fine-tuning can enhance the completeness of inventions' features, conceptual clarity, and feature linkage. Among the tested LLMs, GPT-4 demonstrates the best performance in comprehensive human evaluations by patent experts, with better feature coverage, conceptual clarity, and technical coherence. Despite these capabilities, comprehensive revision and modification are still necessary to pass rigorous patent scrutiny and ensure legal robustness.

Chengzu Li, Caiqi Zhang, Han Zhou et al.

Top-view perspective denotes a typical way in which humans read and reason over different types of maps, and it is vital for localization and navigation of humans as well as of `non-human' agents, such as the ones backed by large Vision-Language Models (VLMs). Nonetheless, spatial reasoning capabilities of modern VLMs remain unattested and underexplored. In this work, we thus study their capability to understand and reason over spatial relations from the top view. The focus on top view also enables controlled evaluations at different granularity of spatial reasoning; we clearly disentangle different abilities (e.g., recognizing particular objects versus understanding their relative positions). We introduce the TopViewRS (Top-View Reasoning in Space) dataset, consisting of 11,384 multiple-choice questions with either realistic or semantic top-view map as visual input. We then use it to study and evaluate VLMs across 4 perception and reasoning tasks with different levels of complexity. Evaluation of 10 representative open- and closed-source VLMs reveals the gap of more than 50% compared to average human performance, and it is even lower than the random baseline in some cases. Although additional experiments show that Chain-of-Thought reasoning can boost model capabilities by 5.82% on average, the overall performance of VLMs remains limited. Our findings underscore the critical need for enhanced model capability in top-view spatial reasoning and set a foundation for further research towards human-level proficiency of VLMs in real-world multimodal tasks.