Xiaodan Zhu

Stephen Obadinma, Faiza Khan Khattak, Shirley Wang et al. · utoronto

Building Agent Assistants that can help improve customer service support requires inputs from industry users and their customers, as well as knowledge about state-of-the-art Natural Language Processing (NLP) technology. We combine expertise from academia and industry to bridge the gap and build task/domain-specific Neural Agent Assistants (NAA) with three high-level components for: (1) Intent Identification, (2) Context Retrieval, and (3) Response Generation. In this paper, we outline the pipeline of the NAA's core system and also present three case studies in which three industry partners successfully adapt the framework to find solutions to their unique challenges. Our findings suggest that a collaborative process is instrumental in spurring the development of emerging NLP models for Conversational AI tasks in industry. The full reference implementation code and results are available at \url{https://github.com/VectorInstitute/NAA}

Haritz Puerto, Tilek Chubakov, Xiaodan Zhu et al.

Requiring a large language model (LLM) to generate intermediary reasoning steps, known as Chain of Thought (CoT), has been shown to be an effective way of boosting performance. Previous approaches have focused on generating multiple independent CoTs, combining them through ensembling or other post-hoc strategies to enhance reasoning. In this work, we introduce a novel approach where LLMs are fine-tuned to generate a sequence of Diverse Chains of Thought (DCoT) within a single inference step, which is fundamentally different from prior work that primarily operate on parallel CoT generations. DCoT allows LLMs to gain the ability to perform within-inference refinement of reasoning chains without requiring external feedback. Through a rigorous set of experiments spanning a wide range of tasks that require various reasoning types, we show that fine-tuning on DCoT improves performance over the CoT baseline across model families and scales (1.3B to 70B). These improvements are particularly impactful for tasks with a large result state space, such as those involving numeric answers. Our work is also significant because both quantitative analyses and manual evaluations reveal the observed gains stem from the models' ability to refine an initial reasoning chain by generating a second, improved chain within the same inference step, demonstrating previously elusive self-improvement. Our code and data are publicly available at https://github.com/UKPLab/acl2025-diverse-cot.

Tianyu Yang, Xiaodan Zhu, Iryna Gurevych

Anonymizing text that contains sensitive information is crucial for a wide range of applications. Existing techniques face the emerging challenges of the re-identification ability of large language models (LLMs), which have shown advanced capability in memorizing detailed information and reasoning over dispersed pieces of patterns to draw conclusions. When defending against LLM-based re-identification, anonymization could jeopardize the utility of the resulting anonymized data in downstream tasks. In general, the interaction between anonymization and data utility requires a deeper understanding within the context of LLMs. In this paper, we propose a framework composed of three key LLM-based components: a privacy evaluator, a utility evaluator, and an optimization component, which work collaboratively to perform anonymization. Extensive experiments demonstrate that the proposed model outperforms existing baselines, showing robustness in reducing the risk of re-identification while preserving greater data utility in downstream tasks. We provide detailed studies on these core modules. To consider large-scale and real-time applications, we investigate the distillation of the anonymization capabilities into lightweight models. All of our code and datasets will be made publicly available at https://github.com/UKPLab/acl2025-rupta.

Yihao Fang, Xianzhi Li, Stephen W. Thomas et al.

Open intent detection, a crucial aspect of natural language understanding, involves the identification of previously unseen intents in user-generated text. Despite the progress made in this field, challenges persist in handling new combinations of language components, which is essential for compositional generalization. In this paper, we present a case study exploring the use of ChatGPT as a data augmentation technique to enhance compositional generalization in open intent detection tasks. We begin by discussing the limitations of existing benchmarks in evaluating this problem, highlighting the need for constructing datasets for addressing compositional generalization in open intent detection tasks. By incorporating synthetic data generated by ChatGPT into the training process, we demonstrate that our approach can effectively improve model performance. Rigorous evaluation of multiple benchmarks reveals that our method outperforms existing techniques and significantly enhances open intent detection capabilities. Our findings underscore the potential of large language models like ChatGPT for data augmentation in natural language understanding tasks.

Jonathan Li, Rohan Bhambhoria, Xiaodan Zhu

Seeking legal advice is often expensive. Recent advancements in machine learning for solving complex problems can be leveraged to help make legal services more accessible to the public. However, real-life applications encounter significant challenges. State-of-the-art language models are growing increasingly large, making parameter-efficient learning increasingly important. Unfortunately, parameter-efficient methods perform poorly with small amounts of data, which are common in the legal domain (where data labelling costs are high). To address these challenges, we propose parameter-efficient legal domain adaptation, which uses vast unsupervised legal data from public legal forums to perform legal pre-training. This method exceeds or matches the fewshot performance of existing models such as LEGAL-BERT on various legal tasks while tuning only approximately 0.1% of model parameters. Additionally, we show that our method can achieve calibration comparable to existing methods across several tasks. To the best of our knowledge, this work is among the first to explore parameter-efficient methods of tuning language models in the legal domain.

Yufei Feng, Xiaoyu Yang, Xiaodan Zhu et al.

We introduce a neuro-symbolic natural logic framework based on reinforcement learning with introspective revision. The model samples and rewards specific reasoning paths through policy gradient, in which the introspective revision algorithm modifies intermediate symbolic reasoning steps to discover reward-earning operations as well as leverages external knowledge to alleviate spurious reasoning and training inefficiency. The framework is supported by properly designed local relation models to avoid input entangling, which helps ensure the interpretability of the proof paths. The proposed model has built-in interpretability and shows superior capability in monotonicity inference, systematic generalization, and interpretability, compared to previous models on the existing datasets.

Jonathan Li, Rohan Bhambhoria, Samuel Dahan et al.

Generative AI models, such as the GPT and Llama series, have significant potential to assist laypeople in answering legal questions. However, little prior work focuses on the data sourcing, inference, and evaluation of these models in the context of laypersons. To this end, we propose a human-centric legal NLP pipeline, covering data sourcing, inference, and evaluation. We introduce and release a dataset, LegalQA, with real and specific legal questions spanning from employment law to criminal law, corresponding answers written by legal experts, and citations for each answer. We develop an automatic evaluation protocol for this dataset, then show that retrieval-augmented generation from only 850 citations in the train set can match or outperform internet-wide retrieval, despite containing 9 orders of magnitude less data. Finally, we propose future directions for open-sourced efforts, which fall behind closed-sourced models.

Chu Fei Luo, Samuel Dahan, Xiaodan Zhu

Test-time reasoning has become a significant field of study since the introduction of chain-of-thought reasoning in large language models (LLMs). However, the mechanisms of this reasoning process are still under-explored -- from the same input prompt, and even the same partial solution, LLMs can produce varied answers if sampled multiple times. We propose to leverage question-asking as an inference-time intervention that articulates information about the model's hidden state. To achieve that, we present a student-teacher setting where a student asks questions to a teacher. We train a probe on the student's hidden state before and after asking a question and find it is predictive of the trajectory's final correctness, even before generating the teacher's answer. This suggests there is a meaningful signal from the self-diagnosis that occurs during question generation rather than information transfer from the teacher. We then frame question-asking as a sequential decision problem, using this probe as a quality score, and define a gating policy to ask questions that maximize likelihood of correctness. We find that the success of question-asking as an intervention is largely dependent on the model's self-consistency. Our empirical results show a gap between detection and recovery; while our gating policy captures model correctness and uncertainty, interventions are equally likely to harm correct trajectories as they are to recover incorrect ones. This gap between diagnosis and correction has broader implications on language models' capacity for self-refinement under uncertainty.

Zi'ou Zheng, Xiaodan Zhu

Reasoning has been a central topic in artificial intelligence from the beginning. The recent progress made on distributed representation and neural networks continues to improve the state-of-the-art performance of natural language inference. However, it remains an open question whether the models perform real reasoning to reach their conclusions or rely on spurious correlations. Adversarial attacks have proven to be an important tool to help evaluate the Achilles' heel of the victim models. In this study, we explore the fundamental problem of developing attack models based on logic formalism. We propose NatLogAttack to perform systematic attacks centring around natural logic, a classical logic formalism that is traceable back to Aristotle's syllogism and has been closely developed for natural language inference. The proposed framework renders both label-preserving and label-flipping attacks. We show that compared to the existing attack models, NatLogAttack generates better adversarial examples with fewer visits to the victim models. The victim models are found to be more vulnerable under the label-flipping setting. NatLogAttack provides a tool to probe the existing and future NLI models' capacity from a key viewpoint and we hope more logic-based attacks will be further explored for understanding the desired property of reasoning.

Ethan Callanan, Amarachi Mbakwe, Antony Papadimitriou et al.

Large Language Models (LLMs) have demonstrated remarkable performance on a wide range of Natural Language Processing (NLP) tasks, often matching or even beating state-of-the-art task-specific models. This study aims at assessing the financial reasoning capabilities of LLMs. We leverage mock exam questions of the Chartered Financial Analyst (CFA) Program to conduct a comprehensive evaluation of ChatGPT and GPT-4 in financial analysis, considering Zero-Shot (ZS), Chain-of-Thought (CoT), and Few-Shot (FS) scenarios. We present an in-depth analysis of the models' performance and limitations, and estimate whether they would have a chance at passing the CFA exams. Finally, we outline insights into potential strategies and improvements to enhance the applicability of LLMs in finance. In this perspective, we hope this work paves the way for future studies to continue enhancing LLMs for financial reasoning through rigorous evaluation.

Stephen Obadinma, Hongyu Guo, Xiaodan Zhu

Recent work has demonstrated that using parameter efficient tuning techniques such as prefix tuning (or P-tuning) on pretrained language models can yield performance that is comparable or superior to fine-tuning while dramatically reducing trainable parameters. Nevertheless, the effectiveness of such methods under the context of data augmentation, a common strategy to improve learning under low data regimes, has not been fully explored. In this paper, we examine the effectiveness of several popular task-agnostic data augmentation techniques, i.e., EDA, Back Translation, and Mixup, when using two general parameter efficient tuning methods, P-tuning v2 and LoRA, under data scarcity. We show that data augmentation can be used to boost the performance of P-tuning and LoRA models, but the effectiveness of each technique varies and certain methods can lead to a notable degradation in performance, particularly when using larger models and on harder tasks. We further analyze the sentence representations of P-tuning compared to fine-tuning to help understand the above behaviour, and reveal how P-tuning generally presents a more limited ability to separate the sentence embeddings from different classes of augmented data. In addition, it displays poorer performance on heavily altered data. However, we demonstrate that by adding a simple contrastive loss function it can help mitigate such issues for prefix tuning, resulting in sizable improvements to augmented data performance.

Maede Ashofteh Barabadi, Xiaodan Zhu, Wai Yip Chan et al.

Understanding the progression of cancer is crucial for defining treatments for patients. The objective of this study is to automate the detection of metastatic liver disease from free-style computed tomography (CT) radiology reports. Our research demonstrates that transferring knowledge using three approaches can improve model performance. First, we utilize generic language models (LMs), pretrained in a self-supervised manner. Second, we use a semi-supervised approach to train our model by automatically annotating a large unlabeled dataset; this approach substantially enhances the model's performance. Finally, we transfer knowledge from related tasks by designing a multi-task transfer learning methodology. We leverage the recent advancement of parameter-efficient LM adaptation strategies to improve performance and training efficiency. Our dataset consists of CT reports collected at Memorial Sloan Kettering Cancer Center (MSKCC) over the course of 12 years. 2,641 reports were manually annotated by domain experts; among them, 841 reports have been annotated for the presence of liver metastases. Our best model achieved an F1-score of 73.8%, a precision of 84%, and a recall of 65.8%.

Mo Yu, Yi Gu, Xiaoxiao Guo et al.

Commonsense reasoning simulates the human ability to make presumptions about our physical world, and it is an essential cornerstone in building general AI systems. We propose a new commonsense reasoning dataset based on human's Interactive Fiction (IF) gameplay walkthroughs as human players demonstrate plentiful and diverse commonsense reasoning. The new dataset provides a natural mixture of various reasoning types and requires multi-hop reasoning. Moreover, the IF game-based construction procedure requires much less human interventions than previous ones. Different from existing benchmarks, our dataset focuses on the assessment of functional commonsense knowledge rules rather than factual knowledge. Hence, in order to achieve higher performance on our tasks, models need to effectively utilize such functional knowledge to infer the outcomes of actions, rather than relying solely on memorizing facts. Experiments show that the introduced dataset is challenging to previous machine reading models as well as the new large language models with a significant 20% performance gap compared to human experts.

Xianzhi Li, Will Aitken, Xiaodan Zhu et al.

With the recent surge of NLP technologies in the financial domain, banks and other financial entities have adopted virtual agents (VA) to assist customers. A challenging problem for VAs in this domain is determining a user's reason or intent for contacting the VA, especially when the intent was unseen or open during the VA's training. One method for handling open intents is adaptive decision boundary (ADB) post-processing, which learns tight decision boundaries from intent representations to separate known and open intents. We propose incorporating two methods for supervised pre-training of intent representations: prefix-tuning and fine-tuning just the last layer of a large language model (LLM). With this proposal, our accuracy is 1.63% - 2.07% higher than the prior state-of-the-art ADB method for open intent classification on the banking77 benchmark amongst others. Notably, we only supplement the original ADB model with 0.1% additional trainable parameters. Ablation studies also determine that our method yields better results than full fine-tuning the entire model. We hypothesize that our findings could stimulate a new optimal method of downstream tuning that combines parameter efficient tuning modules with fine-tuning a subset of the base model's layers.

Haishuo Fang, Xiaodan Zhu, Iryna Gurevych

Deploying LLM-based agents in real-life applications often faces a critical challenge: the misalignment between agents' behavior and user intent. Such misalignment may lead agents to unintentionally execute critical actions that carry negative outcomes (e.g., accidentally triggering a "buy-now" in web shopping), resulting in undesirable or even irreversible consequences. Although addressing these issues is crucial, the preemptive detection and correction of misaligned actions remains relatively underexplored. To fill this gap, we introduce InferAct, a novel approach that leverages the belief reasoning ability of LLMs, grounded in Theory-of-Mind, to detect misaligned actions before execution. Once the misalignment is detected, InferAct alerts users for timely correction, preventing adverse outcomes and enhancing the reliability of LLM agents' decision-making processes. Experiments on three widely used tasks demonstrate that InferAct achieves up to 20% improvements on Marco-F1 against baselines in misaligned action detection. An in-depth evaluation of misalignment correction further highlights InferAct's effectiveness in improving agent alignment.

Paul Quinlan, Jeremy Levasseur, Qingguo Li et al.

Real-world time series come with text: metadata, descriptions, news, reports. Yet time series foundation models process numerical sequences in isolation, and the multimodal text-and-time-series models that attempt to bridge the two all adapt a pretrained language model post hoc, inheriting representations shaped without ever seeing temporal data. These models are also evaluated almost exclusively against other multimodal baselines, not against the strongest unimodal foundation models in either domain, leaving open whether joint training is needed at all. We present Chronicle, a compact 324M-parameter decoder-only transformer trained from scratch on natural language and time series within a single unified architecture. Both modalities share the same transformer blocks, attention mechanism, and residual stream; the bulk of pretraining uses unimodal batches so cross-modal capability emerges purely from shared parameters, with a short alignment stage that interleaves the two. To our knowledge, Chronicle is the first model jointly pretrained on text and time series from scratch, and the first multimodal model evaluated against dedicated foundation models in both domains. It matches Gemma-3-270M-PT on 19 NLU tasks, sets a new bar for frozen-embedding time series classification on 24 UCR/UEA datasets, and produces multimodal forecasts on Time-MMD that beat every supervised fusion baseline, all from a single backbone.

Rohan Bhambhoria, Samuel Dahan, Jonathan Li et al.

This study evaluates the performance of general-purpose AI, like ChatGPT, in legal question-answering tasks, highlighting significant risks to legal professionals and clients. It suggests leveraging foundational models enhanced by domain-specific knowledge to overcome these issues. The paper advocates for creating open-source legal AI systems to improve accuracy, transparency, and narrative diversity, addressing general AI's shortcomings in legal contexts.

Behzad Shayegh, Yanshuai Cao, Xiaodan Zhu et al.

We investigate the unsupervised constituency parsing task, which organizes words and phrases of a sentence into a hierarchical structure without using linguistically annotated data. We observe that existing unsupervised parsers capture differing aspects of parsing structures, which can be leveraged to enhance unsupervised parsing performance. To this end, we propose a notion of "tree averaging," based on which we further propose a novel ensemble method for unsupervised parsing. To improve inference efficiency, we further distill the ensemble knowledge into a student model; such an ensemble-then-distill process is an effective approach to mitigate the over-smoothing problem existing in common multi-teacher distilling methods. Experiments show that our method surpasses all previous approaches, consistently demonstrating its effectiveness and robustness across various runs, with different ensemble components, and under domain-shift conditions.

Stephen Obadinma, Xiaodan Zhu, Hongyu Guo

In this work, we highlight and perform a comprehensive study on calibration attacks, a form of adversarial attacks that aim to trap victim models to be heavily miscalibrated without altering their predicted labels, hence endangering the trustworthiness of the models and follow-up decision making based on their confidence. We propose four typical forms of calibration attacks: underconfidence, overconfidence, maximum miscalibration, and random confidence attacks, conducted in both black-box and white-box setups. We demonstrate that the attacks are highly effective on both convolutional and attention-based models: with a small number of queries, they seriously skew confidence without changing the predictive performance. Given the potential danger, we further investigate the effectiveness of a wide range of adversarial defence and recalibration methods, including our proposed defences specifically designed for calibration attacks to mitigate the harm. From the ECE and KS scores, we observe that there are still significant limitations in handling calibration attacks. To the best of our knowledge, this is the first dedicated study that provides a comprehensive investigation on calibration-focused attacks. We hope this study helps attract more attention to these types of attacks and hence hamper their potential serious damages. To this end, this work also provides detailed analyses to understand the characteristics of the attacks. Our code is available at https://github.com/PhenetOs/CalibrationAttack

Adam Bayley, Xiaodan Zhu, Raquel Aoki et al.

The recent advancement of Large Language Models (LLMs) offers new opportunities to generate user preference data to warm-start bandits. Recent studies on contextual bandits with LLM initialization (CBLI) have shown that these synthetic priors can significantly lower early regret. However, these findings assume that LLM-generated choices are reasonably aligned with actual user preferences. In this paper, we systematically examine how LLM-generated preferences perform when random and label-flipping noise is injected into the synthetic training data. For aligned domains, we find that warm-starting remains effective up to 30% corruption, loses its advantage around 40%, and degrades performance beyond 50%. When there is systematic misalignment, even without added noise, LLM-generated priors can lead to higher regret than a cold-start bandit. To explain these behaviors, we develop a theoretical analysis that decomposes the effect of random label noise and systematic misalignment on the prior error driving the bandit's regret, and derive a sufficient condition under which LLM-based warm starts are provably better than a cold-start bandit. We validate these results across multiple conjoint datasets and LLMs, showing that estimated alignment reliably tracks when warm-starting improves or degrades recommendation quality.

Omkar Dige, Diljot Singh, Tsz Fung Yau et al.

Mitigating bias in language models (LMs) has become a critical problem due to the widespread deployment of LMs. Numerous approaches revolve around data pre-processing and fine-tuning of language models, tasks that can be both time-consuming and computationally demanding. Consequently, there is a growing interest in machine unlearning techniques given their capacity to induce the forgetting of undesired behaviors of the existing pre-trained or fine-tuned models with lower computational cost. In this work, we explore two unlearning methods, (1) Partitioned Contrastive Gradient Unlearning (PCGU) applied on decoder models and (2) Negation via Task Vector, to reduce social biases in state-of-the-art and open-source LMs such as LLaMA-2 and OPT. We also implement distributed PCGU for large models. It is empirically shown, through quantitative and qualitative analyses, that negation via Task Vector method outperforms PCGU in debiasing with minimum deterioration in performance and perplexity of the models. On LLaMA-27B, negation via Task Vector reduces the bias score by 11.8%

Md Imbesat Hassan Rizvi, Xiaodan Zhu, Iryna Gurevych

Spatial reasoning is a crucial component of both biological and artificial intelligence. In this work, we present a comprehensive study of the capability of current state-of-the-art large language models (LLMs) on spatial reasoning. To support our study, we created and contribute a novel Spatial Reasoning Characterization (SpaRC) framework and Spatial Reasoning Paths (SpaRP) datasets, to enable an in-depth understanding of the spatial relations and compositions as well as the usefulness of spatial reasoning chains. We found that all the state-of-the-art LLMs do not perform well on the datasets -- their performances are consistently low across different setups. The spatial reasoning capability improves substantially as model sizes scale up. Finetuning both large language models (e.g., Llama-2-70B) and smaller ones (e.g., Llama-2-13B) can significantly improve their F1-scores by 7--32 absolute points. We also found that the top proprietary LLMs still significantly outperform their open-source counterparts in topological spatial understanding and reasoning.

Boyuan Zheng, Xiaoyu Yang, Yu-Ping Ruan et al.

This paper introduces the SemEval-2021 shared task 4: Reading Comprehension of Abstract Meaning (ReCAM). This shared task is designed to help evaluate the ability of machines in representing and understanding abstract concepts. Given a passage and the corresponding question, a participating system is expected to choose the correct answer from five candidates of abstract concepts in a cloze-style machine reading comprehension setup. Based on two typical definitions of abstractness, i.e., the imperceptibility and nonspecificity, our task provides three subtasks to evaluate the participating models. Specifically, Subtask 1 aims to evaluate how well a system can model concepts that cannot be directly perceived in the physical world. Subtask 2 focuses on models' ability in comprehending nonspecific concepts located high in a hypernym hierarchy given the context of a passage. Subtask 3 aims to provide some insights into models' generalizability over the two types of abstractness. During the SemEval-2021 official evaluation period, we received 23 submissions to Subtask 1 and 28 to Subtask 2. The participating teams additionally made 29 submissions to Subtask 3. The leaderboard and competition website can be found at https://competitions.codalab.org/competitions/26153. The data and baseline code are available at https://github.com/boyuanzheng010/SemEval2021-Reading-Comprehension-of-Abstract-Meaning.

Cunxiang Wang, Shuailong Liang, Yili Jin et al.

In this paper, we present SemEval-2020 Task 4, Commonsense Validation and Explanation (ComVE), which includes three subtasks, aiming to evaluate whether a system can distinguish a natural language statement that makes sense to humans from one that does not, and provide the reasons. Specifically, in our first subtask, the participating systems are required to choose from two natural language statements of similar wording the one that makes sense and the one does not. The second subtask additionally asks a system to select the key reason from three options why a given statement does not make sense. In the third subtask, a participating system needs to generate the reason. We finally attracted 39 teams participating at least one of the three subtasks. For Subtask A and Subtask B, the performances of top-ranked systems are close to that of humans. However, for Subtask C, there is still a relatively large gap between systems and human performance. The dataset used in our task can be found at https://github.com/wangcunxiang/SemEval2020- Task4-Commonsense-Validation-and-Explanation; The leaderboard can be found at https://competitions.codalab.org/competitions/21080#results.

Yongfei Liu, Bo Wan, Xiaodan Zhu et al.

Visual grounding is a ubiquitous building block in many vision-language tasks and yet remains challenging due to large variations in visual and linguistic features of grounding entities, strong context effect and the resulting semantic ambiguities. Prior works typically focus on learning representations of individual phrases with limited context information. To address their limitations, this paper proposes a language-guided graph representation to capture the global context of grounding entities and their relations, and develop a cross-modal graph matching strategy for the multiple-phrase visual grounding task. In particular, we introduce a modular graph neural network to compute context-aware representations of phrases and object proposals respectively via message propagation, followed by a graph-based matching module to generate globally consistent localization of grounding phrases. We train the entire graph neural network jointly in a two-stage strategy and evaluate it on the Flickr30K Entities benchmark. Extensive experiments show that our method outperforms the prior state of the arts by a sizable margin, evidencing the efficacy of our grounding framework. Code is available at "https://github.com/youngfly11/LCMCG-PyTorch".

Ilia Kuznetsov, Osama Mohammed Afzal, Koen Dercksen et al.

The number of scientific articles produced every year is growing rapidly. Providing quality control over them is crucial for scientists and, ultimately, for the public good. In modern science, this process is largely delegated to peer review -- a distributed procedure in which each submission is evaluated by several independent experts in the field. Peer review is widely used, yet it is hard, time-consuming, and prone to error. Since the artifacts involved in peer review -- manuscripts, reviews, discussions -- are largely text-based, Natural Language Processing has great potential to improve reviewing. As the emergence of large language models (LLMs) has enabled NLP assistance for many new tasks, the discussion on machine-assisted peer review is picking up the pace. Yet, where exactly is help needed, where can NLP help, and where should it stand aside? The goal of our paper is to provide a foundation for the future efforts in NLP for peer-reviewing assistance. We discuss peer review as a general process, exemplified by reviewing at AI conferences. We detail each step of the process from manuscript submission to camera-ready revision, and discuss the associated challenges and opportunities for NLP assistance, illustrated by existing work. We then turn to the big challenges in NLP for peer review as a whole, including data acquisition and licensing, operationalization and experimentation, and ethical issues. To help consolidate community efforts, we create a companion repository that aggregates key datasets pertaining to peer review. Finally, we issue a detailed call for action for the scientific community, NLP and AI researchers, policymakers, and funding bodies to help bring the research in NLP for peer review forward. We hope that our work will help set the agenda for research in machine-assisted scientific quality control in the age of AI, within the NLP community and beyond.

Marcus Tantakoun, Xiaodan Zhu, Christian Muise

Large Language Models (LLMs) excel in various natural language tasks but often struggle with long-horizon planning problems requiring structured reasoning. This limitation has drawn interest in integrating neuro-symbolic approaches within the Automated Planning (AP) and Natural Language Processing (NLP) communities. However, identifying optimal AP deployment frameworks can be daunting and introduces new challenges. This paper aims to provide a timely survey of the current research with an in-depth analysis, positioning LLMs as tools for formalizing and refining planning specifications to support reliable off-the-shelf AP planners. By systematically reviewing the current state of research, we highlight methodologies, and identify critical challenges and future directions, hoping to contribute to the joint research on NLP and Automated Planning.

Yuanhao Shen, Xiaodan Zhu, Lei Chen

The tool-use ability of Large Language Models (LLMs) has a profound impact on a wide range of industrial applications. However, LLMs' self-control and calibration capability in appropriately using tools remains understudied. The problem is consequential as it raises potential risks of degraded performance and poses a threat to the trustworthiness of the models. In this paper, we conduct a study on a family of state-of-the-art LLMs on three datasets with two mainstream tool-use frameworks. Our study reveals the tool-abuse behavior of LLMs, a tendency for models to misuse tools with overconfidence. We also find that this is a common issue regardless of model capability. Accordingly, we propose a novel approach, \textit{SMARTCAL}, to mitigate the observed issues, and our results show an average of 8.6 percent increase in the QA performance and a 21.6 percent decrease in Expected Calibration Error (ECE) compared to baseline models.

Yihao Fang, Stephen W. Thomas, Xiaodan Zhu

With the widespread adoption of large language models (LLMs) in numerous applications, the challenge of factuality and the propensity for hallucinations has emerged as a significant concern. To address this issue, particularly in retrieval-augmented in-context learning, we introduce the hierarchical graph of thoughts (HGOT), a structured, multi-layered graph approach designed to enhance the retrieval of pertinent passages during in-context learning. The framework utilizes the emergent planning capabilities of LLMs, employing the divide-and-conquer strategy to break down complex queries into manageable sub-queries. It refines self-consistency majority voting for answer selection, which incorporates the recently proposed citation recall and precision metrics to assess the quality of thoughts, linking an answer's credibility intrinsically to the thought's quality. This methodology introduces a weighted system in majority voting, prioritizing answers based on the citation quality of their thoughts. Additionally, we propose a scoring mechanism for evaluating retrieved passages, considering factors such as citation frequency and quality, self-consistency confidence, and the retrieval module's ranking. Experiments indicate that HGOT excels as a versatile approach, outperforming competing models in FEVER by up to $7\%$ and matching leading models such as Retrieve-then-Read in Open-SQuAD, and DSP in HotPotQA, demonstrating its efficacy in enhancing LLMs' factuality.

Christopher Malon, Xiaodan Zhu

Self-consistency has emerged as a powerful method for improving the accuracy of short answers generated by large language models. As previously defined, it only concerns the accuracy of a final answer parsed from generated text. In this work, we extend the idea to open response generation, by integrating voting into the decoding method. Each output sentence is selected from among multiple samples, conditioning on the previous selections, based on a simple token overlap score. We compare this "Sample & Select" method to greedy decoding, beam search, nucleus sampling, and the recently introduced hallucination avoiding decoders of DoLA, P-CRR, and S-CRR. We show that Sample & Select improves factuality by a 30% relative margin against these decoders in NLI-based evaluation on the subsets of CNN/DM and XSum used in the FRANK benchmark, while maintaining comparable ROUGE-1 F1 scores against reference summaries. We collect human verifications of the generated summaries, confirming the factual superiority of our method.

Md Imbesat Hassan Rizvi, Xiaodan Zhu, Iryna Gurevych

Process or step-wise supervision has played a crucial role in advancing complex multi-step reasoning capabilities of Large Language Models (LLMs). However, efficient, high-quality automated process annotation remains a significant challenge. To address this, we introduce Single-Pass Annotation with Reference-Guided Evaluation (SPARE), a novel structured framework that enables efficient per-step annotation by jointly aligning solution steps to reference solutions and determine its accuracy with explicit reasoning in single generation. We demonstrate SPARE's effectiveness across four diverse datasets spanning mathematical reasoning (GSM8K, MATH), multi-hop question answering (MuSiQue-Ans), and spatial reasoning (SpaRP), showing consistent improvements in two applications: (1) training Process Reward Models (PRMs) for ranking and aggregating multiple generations, and (2) fine-tuning models via offline reinforcement learning for greedy decoding. On ProcessBench, SPARE demonstrates data-efficient out-of-distribution generalization, using only $\sim$16% of training samples compared to human-labeled and other synthetically trained baselines. Additionally, it achieves competitive performance with MCTS-based methods while offering 2.3$\times$ speedup in terms of total token count. Manual analysis reveals complementary precision-recall characteristics with MCTS approaches, suggesting potential for ensemble methods. These results establish SPARE as a practical and scalable solution for automatic process supervision in LLM reasoning.

Ali Asad, Stephen Obadinma, Radin Shayanfar et al.

We introduce RedDebate, a novel multi-agent debate framework that provides the foundation for Large Language Models (LLMs) to identify and mitigate their unsafe behaviours. Existing AI safety approaches often rely on costly human evaluation or isolated single-model assessment, both constrained by scalability and prone to oversight failures. RedDebate employs collaborative argumentation among multiple LLMs across diverse debate scenarios, enabling them to critically evaluate one another's reasoning and systematically uncover unsafe failure modes through fully automated red-teaming. We further integrate distinct long-term memory modules that preserve safety-relevant insights from debate interactions and leverage them during subsequent inference, facilitating continuous refinement of model behaviour. Empirical evaluation on safety benchmarks across a diverse set of models demonstrates that RedDebate substantially reduces unsafe outputs. While debate alone allows LLMs to refine their behaviour, the addition of memory yields further significant reductions. To the best of our knowledge, RedDebate is the first fully automated framework to unify multi-agent debate and red-teaming to progressively enhance LLM safety without human intervention.

Behzad Shayegh, Hobie H. -B. Lee, Xiaodan Zhu et al.

We address unsupervised dependency parsing by building an ensemble of diverse existing models through post hoc aggregation of their output dependency parse structures. We observe that these ensembles often suffer from low robustness against weak ensemble components due to error accumulation. To tackle this problem, we propose an efficient ensemble-selection approach that considers error diversity and avoids error accumulation. Results demonstrate that our approach outperforms each individual model as well as previous ensemble techniques. Additionally, our experiments show that the proposed ensemble-selection method significantly enhances the performance and robustness of our ensemble, surpassing previously proposed strategies, which have not accounted for error diversity.

Xianzhi Li, Ethan Callanan, Abdellah Ghassel et al.

Test-time compute methods can significantly improve the reasoning capabilities and problem-solving accuracy of large language models (LLMs). However, these approaches require substantially more computational resources, with most compute wasted on exploring low-diversity branches where the model already exhibits high confidence. We observe that a small subset of uncertain reasoning steps has a disproportionately large impact on final prediction accuracy, and branching at these critical junctures tends to yield more diverse and higher-quality candidate reasoning steps. We propose Entropy-Gated Branching (EGB), which branches only at high-uncertainty steps and prunes expansions with a lightweight verifier. On mathematical and financial reasoning benchmarks, EGB improves accuracy by 22.6% over standard inference while operating 31%-75% faster across math benchmarks than test-time beam search with higher performance. Our results show that dynamic resource allocation during inference can substantially improve both efficiency and effectiveness, offering a more scalable pathway to enhanced LLM reasoning capabilities.

Paul Quinlan, Qingguo Li, Xiaodan Zhu

Recent work on time-series models has leveraged self-supervised training to learn meaningful features and patterns in order to improve performance on downstream tasks and generalize to unseen modalities. While these pretraining methods have shown great promise in one-to-many scenarios, where a model is pre-trained on one dataset and fine-tuned on a downstream dataset, they have struggled to generalize to new datasets when more datasets are added during pre-training. This is a fundamental challenge in building foundation models for time-series data, as it limits the ability to develop models that can learn from a large variety of diverse datasets available. To address this challenge, we present a new pre-training paradigm for time-series data called ADAPT, which can efficiently align the physical properties of data in the time-series domain, enabling mixed-batch pre-training despite the extreme discrepancies in the input sizes and channel dimensions of pre-training data. We trained on 162 time-series classification datasets and set new state-of-the-art performance for classification benchmarks. We successfully train a model within the time-series domain on a wide range of datasets simultaneously, which is a major building block for building generalist foundation models in time-series domains.

Paul Quinlan, Qingguo Li, Xiaodan Zhu

Time-series analysis is critical for a wide range of fields such as healthcare, finance, transportation, and energy, among many others. The practical applications often involve analyzing time-series data alongside contextual information in the form of natural language to support informed decisions. However, current time-series models are limited in their ability to perform reasoning that involves both time-series and their textual content. In this work, we address this gap by introducing \textit{Chat-TS}, a large language model (LLM) based framework, designed to support reasoning over time series and textual data. Unlike traditional models, Chat-TS integrates time-series tokens into LLMs' vocabulary, enhancing its reasoning ability over both modalities without compromising the core natural language capabilities, enabling practical analysis and reasoning across modalities. To support learning and evaluation in this setup, we contribute new datasets: the \textit{TS Instruct Training Dataset} which pairs diverse time-series data with relevant text instructions and responses for instruction tuning, the \textit{TS Instruct Question and Answer (QA) Gold Dataset} which provides multiple-choice questions designed to evaluate multimodal reasoning, and a \textit{TS Instruct Quantitative Probing Set} which contains a small subset of the TS Instruct QA tasks alongside math and decision-making questions for LLM evaluation. We designed a training strategy to preserve the inherent reasoning capabilities of LLMs while augmenting them for time-series reasoning. Experiments show that Chat-TS achieves state-of-the-art performance in multi-modal reasoning tasks by maintaining strong natural language proficiency while improving time-series reasoning. ~\footnote{To ensure replicability and facilitate future research, all models, datasets, and code will be available at [\texttt{Github-URL}].}

Zi'ou Zheng, Christopher Malon, Martin Renqiang Min et al.

When performing complex multi-step reasoning tasks, the ability of Large Language Models (LLMs) to derive structured intermediate proof steps is important for ensuring that the models truly perform the desired reasoning and for improving models' explainability. This paper is centred around a focused study: whether the current state-of-the-art generalist LLMs can leverage the structures in a few examples to better construct the proof structures with \textit{in-context learning}. Our study specifically focuses on structure-aware demonstration and structure-aware pruning. We demonstrate that they both help improve performance. A detailed analysis is provided to help understand the results.

Chu Fei Luo, Ahmad Ghawanmeh, Bharat Bhimshetty et al.

Ensuring the safe deployment of AI systems is critical in industry settings where biased outputs can lead to significant operational, reputational, and regulatory risks. Thorough evaluation before deployment is essential to prevent these hazards. Red-teaming addresses this need by employing adversarial attacks to develop guardrails that detect and reject biased or harmful queries, enabling models to be retrained or steered away from harmful outputs. However, most red-teaming efforts focus on harmful or unethical instructions rather than addressing social bias, leaving this critical area under-explored despite its significant real-world impact, especially in customer-facing systems. We propose two bias-specific red-teaming methods, Emotional Bias Probe (EBP) and BiasKG, to evaluate how standard safety measures for harmful content affect bias. For BiasKG, we refactor natural language stereotypes into a knowledge graph. We use these attacking strategies to induce biased responses from several open- and closed-source language models. Unlike prior work, these methods specifically target social bias. We find our method increases bias in all models, even those trained with safety guardrails. Our work emphasizes uncovering societal bias in LLMs through rigorous evaluation, and recommends measures ensure AI safety in high-stakes industry deployments.

Chu Fei Luo, Samuel Dahan, Xiaodan Zhu

As language models have a greater impact on society, it is important to ensure they are aligned to a diverse range of perspectives and are able to reflect nuance in human values. However, the most popular training paradigms for modern language models often assume there is one optimal answer for every query, leading to generic responses and poor alignment. In this work, we aim to enhance pluralistic alignment of language models in a low-resource setting with two methods: pluralistic decoding and model steering. We empirically demonstrate that model steering offers consistent improvement over zero-shot and few-shot baselines with only 50 annotated samples. Our proposed methods decrease false positives in several high-stakes tasks such as hate speech detection and misinformation detection, and improves the distributional alignment to human values in GlobalOpinionQA. We hope our work highlights the importance of diversity and how language models can be adapted to consider nuanced perspectives.

Yuanhao Shen, Daniel Xavier de Sousa, Ricardo Marçal et al.

Recent advancements in Large Language Models (LLMs) have revealed their impressive ability to perform multi-step, logic-driven reasoning across complex domains, positioning them as powerful tools and collaborators in scientific discovery while challenging the long-held view that inspiration-driven ideation is uniquely human. However, the lack of a dedicated benchmark that evaluates LLMs' ability to develop ideas in Interdisciplinary Research (IDR) settings poses a critical barrier to fully understanding their strengths and limitations. To address this gap, we introduce IDRBench -- a pioneering benchmark featuring an expert annotated dataset and a suite of tasks tailored to evaluate LLMs' capabilities in proposing valuable research ideas from different scientific domains for interdisciplinary research. This benchmark aims to provide a systematic framework for assessing LLM performance in complex, cross-domain scientific research. Our dataset consists of scientific publications sourced from the ArXiv platform covering six distinct disciplines, and is annotated by domain experts with diverse academic backgrounds. To ensure high-quality annotations, we emphasize clearly defined dimensions that characterize authentic interdisciplinary research. The design of evaluation tasks in IDRBench follows a progressive, real-world perspective, reflecting the natural stages of interdisciplinary research development, including 1) IDR Paper Identification, 2) IDR Idea Integration, and 3) IDR Idea Recommendation. Using IDRBench, we construct baselines across 10 LLMs and observe that despite fostering some level of IDR awareness, LLMs still struggle to produce quality IDR ideas. These findings could not only spark new research directions, but also help to develop next-generation LLMs that excel in interdisciplinary research.

Stephen Obadinma, Xiaodan Zhu

Robust verbal confidence generated by large language models (LLMs) is crucial for the deployment of LLMs to ensure transparency, trust, and safety in human-AI interactions across many high-stakes applications. In this paper, we present the first comprehensive study on the robustness of verbal confidence under adversarial attacks. We introduce a novel framework for attacking verbal confidence scores through both perturbation and jailbreak-based methods, and show that these attacks can significantly jeopardize verbal confidence estimates and lead to frequent answer changes. We examine a variety of prompting strategies, model sizes, and application domains, revealing that current confidence elicitation methods are vulnerable and that commonly used defence techniques are largely ineffective or counterproductive. Our findings underscore the urgent need to design more robust mechanisms for confidence expression in LLMs, as even subtle semantic-preserving modifications can lead to misleading confidence in responses.

Radin Shayanfar, Chu Fei Luo, Rohan Bhambhoria et al.

Building Task-Oriented Dialogue (TOD) systems that generalize across different tasks remains a challenging problem. Data-driven approaches often struggle to transfer effectively to unseen tasks. While recent schema-based TOD frameworks improve generalization by decoupling task logic from language understanding, their reliance on neural or generative models often obscures how task schemas influence behaviour and hence impair interpretability. In this work, we introduce a novel framework, CoDial (Code for Dialogue), which converts a TOD task schema, represented as a novel structured heterogeneous graph, to programmatic LLM guardrailing code, such as NVIDIA's Colang, enabling interpretable and efficient alignment of dialogue policies during inference. We introduce two paradigms, $\text{CoDial}_{\text{free}}$ and $\text{CoDial}_{\text{structured}}$ for generating LLM guardrails, and propose a feedback mechanism that integrates human feedback to iteratively improve the generated code. Empirically, CoDial achieves state-of-the-art (SOTA) performance on the widely used STAR dataset and is on par with SOTA on the MultiWOZ dataset, while also providing interpretability. We additionally demonstrate CoDial's iterative improvement via manual and LLM-aided feedback, making it a practical tool for expert-guided alignment of LLMs in high-stakes domains.

Abdellah Ghassel, Xianzhi Li, Xiaodan Zhu

While Large Language Models (LLMs) are transforming numerous applications, their susceptibility to conversational breakdowns remains a critical challenge undermining user trust. This paper introduces a "Detect, Explain, Escalate" framework to manage dialogue breakdowns in LLM-powered agents, emphasizing low-carbon operation. Our approach integrates two key strategies: (1) We fine-tune a compact 8B-parameter model, augmented with teacher-generated reasoning traces, which serves as an efficient real-time breakdown 'detector' and 'explainer'. This model demonstrates robust classification and calibration on English and Japanese dialogues, and generalizes well to the BETOLD dataset, improving accuracy by 7% over its baseline. (2) We systematically evaluate frontier LLMs using advanced prompting (few-shot, chain-of-thought, analogical reasoning) for high-fidelity breakdown assessment. These are integrated into an 'escalation' architecture where our efficient detector defers to larger models only when necessary, substantially reducing operational costs and energy consumption. Our fine-tuned model and prompting strategies establish new state-of-the-art results on dialogue breakdown detection benchmarks, outperforming specialized classifiers and significantly narrowing the performance gap to larger proprietary models. The proposed monitor-escalate pipeline reduces inference costs by 54%, offering a scalable, efficient, and more interpretable solution for robust conversational AI in high-impact domains. Code and models will be publicly released.

Haishuo Fang, Xiaodan Zhu, Iryna Gurevych

Answering Questions over Knowledge Graphs (KGQA) is key to well-functioning autonomous language agents in various real-life applications. To improve the neural-symbolic reasoning capabilities of language agents powered by Large Language Models (LLMs) in KGQA, we propose the DecompositionAlignment-Reasoning Agent (DARA) framework. DARA effectively parses questions into formal queries through a dual mechanism: high-level iterative task decomposition and low-level task grounding. Importantly, DARA can be efficiently trained with a small number of high-quality reasoning trajectories. Our experimental results demonstrate that DARA fine-tuned on LLMs (e.g. Llama-2-7B, Mistral) outperforms both in-context learning-based agents with GPT-4 and alternative fine-tuned agents, across different benchmarks in zero-shot evaluation, making such models more accessible for real-life applications. We also show that DARA attains performance comparable to state-of-the-art enumerating-and-ranking-based methods for KGQA.

Haritz Puerto, Martin Tutek, Somak Aditya et al.

Reasoning is a fundamental component of language understanding. Recent prompting techniques, such as chain of thought, have consistently improved LLMs' performance on various reasoning tasks. Nevertheless, there is still little understanding of what triggers reasoning abilities in LLMs in the inference stage. In this paper, we introduce code prompting, a chain of prompts that transforms a natural language problem into code and directly prompts the LLM using the generated code without resorting to external code execution. We hypothesize that code prompts can elicit certain reasoning capabilities of LLMs trained on text and code and utilize the proposed method to improve conditional reasoning, the ability to infer different conclusions depending on the fulfillment of certain conditions. We find that code prompting exhibits a high-performance boost for multiple LLMs (up to 22.52 percentage points on GPT 3.5, 7.75 on Mixtral, and 16.78 on Mistral) across multiple conditional reasoning datasets. We then conduct comprehensive experiments to understand how code prompts trigger reasoning abilities and which capabilities are elicited in the underlying models. Our analysis of GPT 3.5 reveals that the code formatting of the input problem is essential for performance improvement. Furthermore, code prompts improve sample efficiency of in-context learning and facilitate state tracking of variables or entities.

Chu Fei Luo, Rohan Bhambhoria, Samuel Dahan et al.

Deep learning has made significant progress in the past decade, and demonstrates potential to solve problems with extensive social impact. In high-stakes decision making areas such as law, experts often require interpretability for automatic systems to be utilized in practical settings. In this work, we attempt to address these requirements applied to the important problem of legal citation prediction (LCP). We design the task with parallels to the thought-process of lawyers, i.e., with reference to both precedents and legislative provisions. After initial experimental results, we refine the target citation predictions with the feedback of legal experts. Additionally, we introduce a prototype architecture to add interpretability, achieving strong performance while adhering to decision parameters used by lawyers. Our study builds on and leverages the state-of-the-art language processing models for law, while addressing vital considerations for high-stakes tasks with practical societal impact.

Rohan Bhambhoria, Lei Chen, Xiaodan Zhu

In recent years, large language models (LLMs) have achieved strong performance on benchmark tasks, especially in zero or few-shot settings. However, these benchmarks often do not adequately address the challenges posed in the real-world, such as that of hierarchical classification. In order to address this challenge, we propose refactoring conventional tasks on hierarchical datasets into a more indicative long-tail prediction task. We observe LLMs are more prone to failure in these cases. To address these limitations, we propose the use of entailment-contradiction prediction in conjunction with LLMs, which allows for strong performance in a strict zero-shot setting. Importantly, our method does not require any parameter updates, a resource-intensive process and achieves strong performance across multiple datasets.

Chu Fei Luo, Rohan Bhambhoria, Xiaodan Zhu et al.

Hate speech causes widespread and deep-seated societal issues. Proper enforcement of hate speech laws is key for protecting groups of people against harmful and discriminatory language. However, determining what constitutes hate speech is a complex task that is highly open to subjective interpretations. Existing works do not align their systems with enforceable definitions of hate speech, which can make their outputs inconsistent with the goals of regulators. This research introduces a new perspective and task for enforceable hate speech detection centred around legal definitions, and a dataset annotated on violations of eleven possible definitions by legal experts. Given the challenge of identifying clear, legally enforceable instances of hate speech, we augment the dataset with expert-generated samples and an automatically mined challenge set. We experiment with grounding the model decision in these definitions using zero-shot and few-shot prompting. We then report results on several large language models (LLMs). With this task definition, automatic hate speech detection can be more closely aligned to enforceable laws, and hence assist in more rigorous enforcement of legal protections against harmful speech in public forums.

Jonathan Li, Will Aitken, Rohan Bhambhoria et al.

Parameter-efficient tuning aims to mitigate the large memory requirements of adapting pretrained language models for downstream tasks. For example, one popular method, prefix-tuning, prepends trainable tokens to sequences while freezing the rest of the model's parameters. Although such models attain comparable performance with fine-tuning when applied to sequences with short to moderate lengths, we show their inferior performance when modelling long sequences. To bridge this gap, we propose prefix-propagation, a simple but effective approach that conditions prefixes on previous hidden states. We empirically demonstrate that prefix-propagation outperforms prefix-tuning across long-document tasks, while using 50% fewer parameters. To further investigate the proposed architecture, we also show its advantage in calibration, and perform additional study on its relationship with kernel attention. To the best of our knowledge, this work is the first to focus on parameter-efficient learning for long-sequence language tasks.

Xianzhi Li, Samuel Chan, Xiaodan Zhu et al.

The most recent large language models(LLMs) such as ChatGPT and GPT-4 have shown exceptional capabilities of generalist models, achieving state-of-the-art performance on a wide range of NLP tasks with little or no adaptation. How effective are such models in the financial domain? Understanding this basic question would have a significant impact on many downstream financial analytical tasks. In this paper, we conduct an empirical study and provide experimental evidences of their performance on a wide variety of financial text analytical problems, using eight benchmark datasets from five categories of tasks. We report both the strengths and limitations of the current models by comparing them to the state-of-the-art fine-tuned approaches and the recently released domain-specific pretrained models. We hope our study can help understand the capability of the existing models in the financial domain and facilitate further improvements.