Jiazheng Li

Junru Lu, Jiazheng Li, Byron C. Wallace et al.

Accessing medical literature is difficult for laypeople as the content is written for specialists and contains medical jargon. Automated text simplification methods offer a potential means to address this issue. In this work, we propose a summarize-then-simplify two-stage strategy, which we call NapSS, identifying the relevant content to simplify while ensuring that the original narrative flow is preserved. In this approach, we first generate reference summaries via sentence matching between the original and the simplified abstracts. These summaries are then used to train an extractive summarizer, learning the most relevant content to be simplified. Then, to ensure the narrative consistency of the simplified text, we synthesize auxiliary narrative prompts combining key phrases derived from the syntactical analyses of the original text. Our model achieves results significantly better than the seq2seq baseline on an English medical corpus, yielding 3%~4% absolute improvements in terms of lexical similarity, and providing a further 1.1% improvement of SARI score when combined with the baseline. We also highlight shortcomings of existing evaluation methods, and introduce new metrics that take into account both lexical and high-level semantic similarity. A human evaluation conducted on a random sample of the test set further establishes the effectiveness of the proposed approach. Codes and models are released here: https://github.com/LuJunru/NapSS.

Jiazheng Li, Zhaoyue Sun, Bin Liang et al.

Text classifiers built on Pre-trained Language Models (PLMs) have achieved remarkable progress in various tasks including sentiment analysis, natural language inference, and question-answering. However, the occurrence of uncertain predictions by these classifiers poses a challenge to their reliability when deployed in practical applications. Much effort has been devoted to designing various probes in order to understand what PLMs capture. But few studies have delved into factors influencing PLM-based classifiers' predictive uncertainty. In this paper, we propose a novel framework, called CUE, which aims to interpret uncertainties inherent in the predictions of PLM-based models. In particular, we first map PLM-encoded representations to a latent space via a variational auto-encoder. We then generate text representations by perturbing the latent space which causes fluctuation in predictive uncertainty. By comparing the difference in predictive uncertainty between the perturbed and the original text representations, we are able to identify the latent dimensions responsible for uncertainty and subsequently trace back to the input features that contribute to such uncertainty. Our extensive experiments on four benchmark datasets encompassing linguistic acceptability classification, emotion classification, and natural language inference show the feasibility of our proposed framework. Our source code is available at: https://github.com/lijiazheng99/CUE.

Zhaoyue Sun, Jiazheng Li, Gabriele Pergola et al.

The primary goal of drug safety researchers and regulators is to promptly identify adverse drug reactions. Doing so may in turn prevent or reduce the harm to patients and ultimately improve public health. Evaluating and monitoring drug safety (i.e., pharmacovigilance) involves analyzing an ever growing collection of spontaneous reports from health professionals, physicians, and pharmacists, and information voluntarily submitted by patients. In this scenario, facilitating analysis of such reports via automation has the potential to rapidly identify safety signals. Unfortunately, public resources for developing natural language models for this task are scant. We present PHEE, a novel dataset for pharmacovigilance comprising over 5000 annotated events from medical case reports and biomedical literature, making it the largest such public dataset to date. We describe the hierarchical event schema designed to provide coarse and fine-grained information about patients' demographics, treatments and (side) effects. Along with the discussion of the dataset, we present a thorough experimental evaluation of current state-of-the-art approaches for biomedical event extraction, point out their limitations, and highlight open challenges to foster future research in this area.

Yuxiang Zhou, Jiazheng Li, Yanzheng Xiang et al.

Understanding in-context learning (ICL) capability that enables large language models (LLMs) to excel in proficiency through demonstration examples is of utmost importance. This importance stems not only from the better utilization of this capability across various tasks, but also from the proactive identification and mitigation of potential risks, including concerns regarding truthfulness, bias, and toxicity, that may arise alongside the capability. In this paper, we present a thorough survey on the interpretation and analysis of in-context learning. First, we provide a concise introduction to the background and definition of in-context learning. Then, we give an overview of advancements from two perspectives: 1) a theoretical perspective, emphasizing studies on mechanistic interpretability and delving into the mathematical foundations behind ICL; and 2) an empirical perspective, concerning studies that empirically analyze factors associated with ICL. We conclude by highlighting the challenges encountered and suggesting potential avenues for future research. We believe that our work establishes the basis for further exploration into the interpretation of in-context learning. Additionally, we have created a repository containing the resources referenced in our survey.

Jiazheng Li, Chi-Hao Wu, Yunze Liu et al.

Understanding ultra-long videos such as egocentric recordings, live streams, or surveillance footage spanning days to weeks, remains a challenge. For current multimodal LLMs: even with million-token context windows, frame budgets cover only tens of minutes of densely sampled video, and most evidence is discarded before inference begins. Memory-augmented and agentic approaches help with scale, but their retrieval remains fragmented across modalities and lacks long-range narrative summaries that span days or weeks. We propose \textbf{MAGIC-Video}, a training-free framework built around a multimodal memory graph with interleaved narrative chain: the graph unifies episodic, semantic, and visual content through six typed edges and supports cross-modal retrieval, while the chain distils long-horizon entity biographies and recurring activity events. At inference time, an agentic loop interleaves graph retrieval with narrative fact injection, covering both the modality and time dimensions of ultra-long video in a single retrieval pipeline. On EgoLifeQA, Ego-R1 and MM-Lifelong, MAGIC-Video consistently outperforms strong general-purpose, long-video, and agentic baselines, with gains of 10.1, 7.4, and 5.9 points over the prior best agentic system on each benchmark. Code is available at https://github.com/lijiazheng0917/MAGIC-video.

Runcong Zhao, Wenjia Zhang, Jiazheng Li et al.

In this paper, we introduce NarrativePlay, a novel system that allows users to role-play a fictional character and interact with other characters in narratives such as novels in an immersive environment. We leverage Large Language Models (LLMs) to generate human-like responses, guided by personality traits extracted from narratives. The system incorporates auto-generated visual display of narrative settings, character portraits, and character speech, greatly enhancing user experience. Our approach eschews predefined sandboxes, focusing instead on main storyline events extracted from narratives from the perspective of a user-selected character. NarrativePlay has been evaluated on two types of narratives, detective and adventure stories, where users can either explore the world or improve their favorability with the narrative characters through conversations.

Jiazheng Li, Emine Yilmaz, Bei Chen et al.

Large Language Model (LLM)-based Multi-Agent Systems (MAS) enable complex problem-solving but introduce significant debugging challenges, characterized by long interaction traces, inter-agent dependencies, and delayed error manifestation. Existing diagnostic approaches often rely on expensive expert annotation or ''LLM-as-a-judge'' paradigms, which struggle to pinpoint decisive error steps within extended contexts. In this paper, we introduce ErrorProbe, a self-improving framework for semantic failure attribution that identifies responsible agents and the originating error step. The framework operates via a three-stage pipeline: (1) operationalizing the MAS failure taxonomy to detect local anomalies, (2) performing symptom-driven backward tracing to prune irrelevant context, and (3) employing a specialized multi-agent team (Strategist, Investigator, Arbiter) to validate error hypotheses through tool-grounded execution. Crucially, ErrorProbe maintains a verified episodic memory that updates only when error patterns are confirmed by executable evidence, without the need for annotation. Experiments across the TracerTraj and Who&When benchmarks demonstrate that ErrorProbe significantly outperforms baselines, particularly in step-level localization, while the verified memory enables robust cross-domain transfer without retraining.

Ziming Li, Xiaoming Wu, Zehong Wang et al.

Graphs provide a natural representation of relational structure that arises across diverse domains. Despite this ubiquity, graph structure is typically learned in a modality- and task-isolated manner, where graph representations are constructed within individual task contexts and discarded thereafter. As a result, structural regularities across modalities and tasks are repeatedly reconstructed rather than accumulated at the level of intermediate graph representations. This motivates a representation-learning question: how should graph structure be organized so that it can persist and accumulate across heterogeneous modalities and tasks? We adopt a representation-centric perspective in which graph structure is treated as a structural substrate that persists across learning contexts. To instantiate this perspective, we propose G-Substrate, a graph substrate framework that organizes learning around shared graph structures. G-Substrate comprises two complementary mechanisms: a unified structural schema that ensures compatibility among graph representations across heterogeneous modalities and tasks, and an interleaved role-based training strategy that exposes the same graph structure to multiple functional roles during learning. Experiments across multiple domains, modalities, and tasks show that G-Substrate outperforms task-isolated and naive multi-task learning methods.

Safal Thapaliya, Zehong Wang, Jiazheng Li et al.

Graph-structured data exhibit substantial heterogeneity in where their predictive signals originate: in some domains, node-level semantics dominate, while in others, structural patterns play a central role. This structure-semantics heterogeneity implies that no graph learning model with a fixed inductive bias can generalize optimally across diverse graph domains. However, most existing methods address this challenge from the model side by incrementally injecting new inductive biases, which remains fundamentally limited given the open-ended diversity of real-world graphs. In this work, we take a data-centric perspective and treat node semantics as a task-adaptive variable. We propose a Data-Adaptive Semantic Refinement framework DAS for graph representation learning, which couples a fixed graph neural network (GNN) and a large language model (LLM) in a closed feedback loop. The GNN provides implicit supervisory signals to guide the semantic refinement of LLM, and the refined semantics are fed back to update the same graph learner. We evaluate our approach on both text-rich and text-free graphs. Results show consistent improvements on structure-dominated graphs while remaining competitive on semantics-rich graphs, demonstrating the effectiveness of data-centric semantic adaptation under structure-semantics heterogeneity.

Zhaoyue Sun, Jiazheng Li, Gabriele Pergola et al.

Predicting unknown drug-drug interactions (DDIs) is crucial for improving medication safety. Previous efforts in DDI prediction have typically focused on binary classification or predicting DDI categories, with the absence of explanatory insights that could enhance trust in these predictions. In this work, we propose to generate natural language explanations for DDI predictions, enabling the model to reveal the underlying pharmacodynamics and pharmacokinetics mechanisms simultaneously as making the prediction. To do this, we have collected DDI explanations from DDInter and DrugBank and developed various models for extensive experiments and analysis. Our models can provide accurate explanations for unknown DDIs between known drugs. This paper contributes new tools to the field of DDI prediction and lays a solid foundation for further research on generating explanations for DDI predictions.

Zehong Wang, Zheyuan Liu, Tianyi Ma et al.

Graph-structured data pervades domains such as social networks, biological systems, knowledge graphs, and recommender systems. While foundation models have transformed natural language processing, vision, and multimodal learning through large-scale pretraining and generalization, extending these capabilities to graphs -- characterized by non-Euclidean structures and complex relational semantics -- poses unique challenges and opens new opportunities. To this end, Graph Foundation Models (GFMs) aim to bring scalable, general-purpose intelligence to structured data, enabling broad transfer across graph-centric tasks and domains. This survey provides a comprehensive overview of GFMs, unifying diverse efforts under a modular framework comprising three key components: backbone architectures, pretraining strategies, and adaptation mechanisms. We categorize GFMs by their generalization scope -- universal, task-specific, and domain-specific -- and review representative methods, key innovations, and theoretical insights within each category. Beyond methodology, we examine theoretical foundations including transferability and emergent capabilities, and highlight key challenges such as structural alignment, heterogeneity, scalability, and evaluation. Positioned at the intersection of graph learning and general-purpose AI, GFMs are poised to become foundational infrastructure for open-ended reasoning over structured data. This survey consolidates current progress and outlines future directions to guide research in this rapidly evolving field. Resources are available at https://github.com/Zehong-Wang/Awesome-Foundation-Models-on-Graphs.

Zifan Chen, Xinyu Nan, Jiazheng Li et al.

Volumetric segmentation is important in medical imaging, but current methods face challenges like requiring lots of manual annotations and being tailored to specific tasks, which limits their versatility. General segmentation models used for natural images don't perform well with the unique features of medical images. There's a strong need for an adaptable approach that can effectively handle different 3D medical structures and imaging modalities. In this study, we present PAM (Propagating Anything Model), a segmentation approach that uses a 2D prompt, like a bounding box or sketch, to create a complete 3D segmentation of medical image volumes. PAM works by modeling relationships between slices, maintaining information flow across the 3D structure. It combines a CNN-based UNet for processing within slices and a Transformer-based attention module for propagating information between slices, leading to better generalizability across various imaging modalities. PAM significantly outperformed existing models like MedSAM and SegVol, with an average improvement of over 18.1% in dice similarity coefficient (DSC) across 44 medical datasets and various object types. It also showed stable performance despite prompt deviations and different propagation setups, and faster inference speeds compared to other models. PAM's one-view prompt design made it more efficient, reducing interaction time by about 63.6% compared to two-view prompts. Thanks to its focus on structural relationships, PAM handled unseen and complex objects well, showing a unique ability to generalize to new situations. PAM represents an advancement in medical image segmentation, effectively reducing the need for extensive manual work and specialized training. Its adaptability makes it a promising tool for more automated and reliable analysis in clinical settings.

Jiazheng Li, Hongzhou Lin, Hong Lu et al.

Reinforcement learning (RL) has emerged as a central paradigm for training large language models (LLMs) in reasoning tasks. Yet recent studies question RL's ability to incentivize reasoning capacity beyond the base model. This raises a key challenge: how can RL be adapted to solve harder reasoning problems more effectively? To address this challenge, we propose a simple yet effective strategy via Question Augmentation: introduce partial solutions during training to reduce problem difficulty and provide more informative learning signals. Our method, QuestA, when applied during RL training on math reasoning tasks, not only improves pass@1 but also pass@k-particularly on problems where standard RL struggles to make progress. This enables continual improvement over strong open-source models such as DeepScaleR and OpenMath Nemotron, further enhancing their reasoning capabilities. We achieve new state-of-the-art results on math benchmarks using 1.5B-parameter models: 72.50% (+10.73%) on AIME24, 62.29% (+12.79%) on AIME25, and 41.67% (+10.11%) on HMMT25. Code, data and model are available at https://github.com/foreverlasting1202/QuestA.

Jiazheng Li, Yuxiang Zhou, Junru Lu et al.

Although preference optimization methods have improved reasoning performance in Large Language Models (LLMs), they often lack transparency regarding why one reasoning outcome is preferred over another. This limitation is especially critical in Automated Student Answer Scoring (ASAS), where explainability is essential to justify assessment outcomes. Verbal reinforcement learning offers the potential to generate explicit reflection, but it tends to produce superficial critiques that can harm assessment performance. Existing LLMs also struggle to reliably detect subtle reasoning errors in ASAS tasks. Moreover, manually identifying intermediate reasoning errors is expensive and difficult to scale. To address these challenges, we introduce a contrastive reflection synthesis pipeline that generates precise verbal feedback by identifying discrepancies in structure reasoning graph paths. Leveraging these synthetic reflection data, we propose DARS, a Dual-model Reflective Scoring framework featuring a dedicated Critic model trained for effective reflection. DARS achieves strong performance and consistently outperforms existing ASAS baselines across all evaluation metrics. Extensive experiments further provide novel insights into the value of reflection data, framework design, and the scaling behavior of DARS. We release the DARS code at https://github.com/lijiazheng99/DARS.

Junru Lu, Jiazheng Li, Guodong Shen et al.

Role-playing is important for Large Language Models (LLMs) to follow diverse instructions while maintaining role identity and the role's pre-defined ability limits. Existing role-playing datasets mostly contribute to controlling role style and knowledge boundaries, but overlook role-playing in instruction-following scenarios. We introduce a fine-grained role-playing and instruction-following composite benchmark, named RoleMRC, including: (1) Multi-turn dialogues between ideal roles and humans, including free chats or discussions upon given passages; (2) Role-playing machine reading comprehension, involving response, refusal, and attempts according to passage answerability and role ability; (3) More complex scenarios with nested, multi-turn and prioritized instructions. The final RoleMRC features a 10.2k role profile meta-pool, 37.9k well-synthesized role-playing instructions, and 1.4k testing samples. We develop a pipeline to quantitatively evaluate the fine-grained role-playing and instruction-following capabilities of several mainstream LLMs, as well as models that are fine-tuned on our data. Moreover, cross-evaluation on external role-playing datasets confirms that models fine-tuned on RoleMRC enhances instruction-following without compromising general role-playing and reasoning capabilities. We also probe the neural-level activation maps of different capabilities over post-tuned LLMs. Access to our RoleMRC, RoleMRC-mix and Codes: https://github.com/LuJunru/RoleMRC.

Zhenyi Shen, Junru Lu, Lin Gui et al.

Sparse attention reduces the quadratic complexity of full self-attention but faces two challenges: (1) an attention gap, where applying sparse attention to full-attention-trained models causes performance degradation due to train-inference distribution mismatch, and (2) a capability gap, where models trained purely with sparse attention lack complete gradient flow, preventing them from matching full-attention performance. We propose SSA (Sparse Sparse Attention), a training framework that integrates both sparse and full attention with bidirectional attention-output alignment. We prove that the approximation error scales linearly with the attention mass dropped under sparse attention, and show that SSA's alignment objective substantially reduces this quantity compared to baselines. Experiments demonstrate that SSA achieves state-of-the-art performance under both inference modes, adapts smoothly to varying sparsity budgets, and demonstrates superior long-context capabilities. The code is available at https://github.com/zhenyi4/ssa.

Jiazheng Li, Hainiu Xu, Zhaoyue Sun et al.

Generating rationales that justify scoring decisions has been a promising way to facilitate explainability in automated scoring systems. However, existing methods do not match the accuracy of classifier-based methods. Plus, the generated rationales often contain hallucinated information. To address these issues, we propose a novel framework capable of generating more faithful rationales and, more importantly, matching performance with classifier-based black-box scoring systems. We first mimic the human assessment process by querying Large Language Models (LLMs) to generate a thought tree. We then summarise intermediate assessment decisions from each thought tree path for creating synthetic rationale data and rationale preference data. Finally, we utilise the generated synthetic data to calibrate LLMs through a two-step training process: supervised fine-tuning and preference optimization. Extensive experimental results demonstrate that our framework achieves a 38% assessment performance improvement in the QWK score compared to prior work while producing higher-quality rationales, as recognised by human evaluators and LLMs. Our work sheds light on the effectiveness of performing preference optimization using synthetic preference data obtained from thought tree paths. Data and code are available at https://github.com/lijiazheng99/thought_tree_assessment.

Junru Lu, Jiazheng Li, Siyu An et al.

Direct Preference Optimization (DPO) has emerged as a prominent algorithm for the direct and robust alignment of Large Language Models (LLMs) with human preferences, offering a more straightforward alternative to the complex Reinforcement Learning from Human Feedback (RLHF). Despite its promising efficacy, DPO faces a notable drawback: "verbosity", a common over-optimization phenomenon also observed in RLHF. While previous studies mainly attributed verbosity to biased labels within the data, we propose that the issue also stems from an inherent algorithmic length reliance in DPO. Specifically, we suggest that the discrepancy between sequence-level Kullback-Leibler (KL) divergences between chosen and rejected sequences, used in DPO, results in overestimated or underestimated rewards due to varying token lengths. Empirically, we utilize datasets with different label lengths to demonstrate the presence of biased rewards. We then introduce an effective downsampling approach, named SamPO, to eliminate potential length reliance. Our experimental evaluations, conducted across three LLMs of varying scales and a diverse array of conditional and open-ended benchmarks, highlight the efficacy of SamPO in mitigating verbosity, achieving improvements of 5% to 12% over DPO through debaised rewards. Our codes can be accessed at: https://github.com/LuJunru/SamPO/.

Jiazheng Li, Lin Gui, Yuxiang Zhou et al.

Providing explainable and faithful feedback is crucial for automated student answer assessment. In this paper, we introduce a novel framework that explores using ChatGPT, a cutting-edge large language model, for the concurrent tasks of student answer scoring and rationale generation. We identify the appropriate instructions by prompting ChatGPT with different templates to collect the rationales, where inconsistent rationales are refined to align with marking standards. The refined ChatGPT outputs enable us to fine-tune a smaller language model that simultaneously assesses student answers and provides rationales. Extensive experiments on the benchmark dataset show that the proposed method improves the overall QWK score by 11% compared to ChatGPT. Furthermore, our thorough analysis and human evaluation demonstrate that the rationales generated by our proposed method are comparable to those of ChatGPT. Our approach provides a viable solution to achieve explainable automated assessment in education. Code available at https://github.com/lijiazheng99/aera.

Jack Geraghty, Jiazheng Li, Alessandro Ragano et al.

Audio quality assessment has been widely researched in the signal processing area. Full-reference objective metrics (e.g., POLQA, ViSQOL) have been developed to estimate the audio quality relying only on human rating experiments. To evaluate the audio quality of novel audio processing techniques, researchers constantly need to compare objective quality metrics. Testing different implementations of the same metric and evaluating new datasets are fundamental and ongoing iterative activities. In this paper, we present AQP - an open-source, node-based, light-weight Python pipeline for audio quality assessment. AQP allows researchers to test and compare objective quality metrics helping to improve robustness, reproducibility and development speed. We introduce the platform, explain the motivations, and illustrate with examples how, using AQP, objective quality metrics can be (i) compared and benchmarked; (ii) prototyped and adapted in a modular fashion; (iii) visualised and checked for errors. The code has been shared on GitHub to encourage adoption and contributions from the community.

Runcong Zhao, Qinglin Zhu, Hainiu Xu et al.

Existing datasets for narrative understanding often fail to represent the complexity and uncertainty of relationships in real-life social scenarios. To address this gap, we introduce a new benchmark, Conan, designed for extracting and analysing intricate character relation graphs from detective narratives. Specifically, we designed hierarchical relationship categories and manually extracted and annotated role-oriented relationships from the perspectives of various characters, incorporating both public relationships known to most characters and secret ones known to only a few. Our experiments with advanced Large Language Models (LLMs) like GPT-3.5, GPT-4, and Llama2 reveal their limitations in inferencing complex relationships and handling longer narratives. The combination of the Conan dataset and our pipeline strategy is geared towards understanding the ability of LLMs to comprehend nuanced relational dynamics in narrative contexts.

Jing Xu, Jiazheng Li, Jingzhao Zhang

Model merging offers an effective way to integrate the capabilities of multiple fine-tuned models. However, the performance degradation of the merged model remains a challenge, particularly when none or few data are available. This paper first highlights the necessity of domain-specific data for model merging by proving that data-agnostic algorithms can have arbitrarily bad worst-case performance. Building on this theoretical insight, we explore the relationship between model merging and distillation, introducing a novel few-shot merging algorithm, ProDistill (Progressive Layer-wise Distillation). Unlike common belief that layer wise training hurts performance, we show that layer-wise teacher-student distillation not only enhances the scalability but also improves model merging performance. We conduct extensive experiments to show that compared to existing few-shot merging methods, ProDistill achieves state-of-the-art performance, with up to 6.14% and 6.61% improvements in vision and NLU tasks. Furthermore, we extend the experiments to models with over 10B parameters, showcasing the exceptional scalability of ProDistill.

Jiazheng Li, Jundong Li, Chuxu Zhang

Graph neural networks stand as the predominant technique for graph representation learning owing to their strong expressive power, yet the performance highly depends on the availability of high-quality labels in an end-to-end manner. Thus the pretraining and fine-tuning paradigm has been proposed to mitigate the label cost issue. Subsequently, the gap between the pretext tasks and downstream tasks has spurred the development of graph prompt learning which inserts a set of graph prompts into the original graph data with minimal parameters while preserving competitive performance. However, the current exploratory works are still limited since they all concentrate on learning fixed task-specific prompts which may not generalize well across the diverse instances that the task comprises. To tackle this challenge, we introduce Instance-Aware Graph Prompt Learning (IA-GPL) in this paper, aiming to generate distinct prompts tailored to different input instances. The process involves generating intermediate prompts for each instance using a lightweight architecture, quantizing these prompts through trainable codebook vectors, and employing the exponential moving average technique to ensure stable training. Extensive experiments conducted on multiple datasets and settings showcase the superior performance of IA-GPL compared to state-of-the-art baselines.

Hainiu Xu, Siya Qi, Jiazheng Li et al.

Theory-of-Mind (ToM), the ability to infer others' perceptions and mental states, is fundamental to human interaction but remains challenging for Large Language Models (LLMs). While existing ToM reasoning methods show promise with reasoning via perceptual perspective-taking, they often rely excessively on off-the-shelf LLMs, reducing their efficiency and limiting their applicability to high-order ToM reasoning. To address these issues, we present EnigmaToM, a novel neuro-symbolic framework that enhances ToM reasoning by integrating a Neural Knowledge Base of entity states (Enigma) for (1) a psychology-inspired iterative masking mechanism that facilitates accurate perspective-taking and (2) knowledge injection that elicits key entity information. Enigma generates structured knowledge of entity states to build spatial scene graphs for belief tracking across various ToM orders and enrich events with fine-grained entity state details. Experimental results on ToMi, HiToM, and FANToM benchmarks show that EnigmaToM significantly improves ToM reasoning across LLMs of varying sizes, particularly excelling in high-order reasoning scenarios.

Jiazheng Li, Artem Bobrov, David West et al.

In this demo, we present AERA Chat, an automated and explainable educational assessment system designed for interactive and visual evaluations of student responses. This system leverages large language models (LLMs) to generate automated marking and rationale explanations, addressing the challenge of limited explainability in automated educational assessment and the high costs associated with annotation. Our system allows users to input questions and student answers, providing educators and researchers with insights into assessment accuracy and the quality of LLM-assessed rationales. Additionally, it offers advanced visualization and robust evaluation tools, enhancing the usability for educational assessment and facilitating efficient rationale verification. Our demo video can be found at https://youtu.be/qUSjz-sxlBc.

Runcong Zhao, Artem Bobrov, Jiazheng Li et al.

Effective feedback is essential for student learning but is time-intensive for teachers. We present LearnLens, a modular, LLM-based system that generates personalised, curriculum-aligned feedback in science education. LearnLens comprises three components: (1) an error-aware assessment module that captures nuanced reasoning errors; (2) a curriculum-grounded generation module that uses a structured, topic-linked memory chain rather than traditional similarity-based retrieval, improving relevance and reducing noise; and (3) an educator-in-the-loop interface for customisation and oversight. LearnLens addresses key challenges in existing systems, offering scalable, high-quality feedback that empowers both teachers and students.

Xingbo Fu, Zehong Wang, Zihan Chen et al.

Graph learning models have demonstrated great prowess in learning expressive representations from large-scale graph data in a wide variety of real-world scenarios. As a prevalent strategy for training powerful graph learning models, the "pre-training, adaptation" scheme first pre-trains graph learning models on unlabeled graph data in a self-supervised manner and then adapts them to specific downstream tasks. During the adaptation phase, graph prompting emerges as a promising approach that learns trainable prompts while keeping the pre-trained graph learning models unchanged. In this paper, we present a systematic review of recent advancements in graph prompting. First, we introduce representative graph pre-training methods that serve as the foundation step of graph prompting. Next, we review mainstream techniques in graph prompting and elaborate on how they design learnable prompts for graph prompting. Furthermore, we summarize the real-world applications of graph prompting from different domains. Finally, we discuss several open challenges in existing studies with promising future directions in this field.

Xinran Gu, Kaifeng Lyu, Jiazheng Li et al. · tsinghua

Large Language Models (LLMs) are typically trained on data mixtures: most data come from web scrapes, while a small portion is curated from high-quality sources with dense domain-specific knowledge. In this paper, we show that when training LLMs on such data mixtures, knowledge acquisition from knowledge-dense datasets, unlike training exclusively on knowledge-dense data (arXiv:2404.05405), does not always follow a smooth scaling law but can exhibit phase transitions with respect to the mixing ratio and model size. Through controlled experiments on a synthetic biography dataset mixed with web-scraped data, we demonstrate that: (1) as we increase the model size to a critical value, the model suddenly transitions from memorizing very few to most of the biographies; (2) below a critical mixing ratio, the model memorizes almost nothing even with extensive training, but beyond this threshold, it rapidly memorizes more biographies. We attribute these phase transitions to a capacity allocation phenomenon: a model with bounded capacity must act like a knapsack problem solver to minimize the overall test loss, and the optimal allocation across datasets can change discontinuously as the model size or mixing ratio varies. We formalize this intuition in an information-theoretic framework and reveal that these phase transitions are predictable, with the critical mixing ratio following a power-law relationship with the model size. Our findings highlight a concrete case where a good mixing recipe for large models may not be optimal for small models, and vice versa.

Jiazheng Li, Lu Yu, Qing Cui et al.

High-quality data plays a critical role in the pretraining and fine-tuning of large language models (LLMs), even determining their performance ceiling to some degree. Consequently, numerous data selection methods have been proposed to identify subsets of data that can effectively and efficiently enhance model performance. However, most of these methods focus on general data selection and tend to overlook the specific nuances of domain-related data. In this paper, we introduce MASS, a \textbf{MA}thematical data \textbf{S}election framework using the \textbf{S}kill graph for pretraining LLMs in the mathematical reasoning domain. By taking into account the unique characteristics of mathematics and reasoning, we construct a skill graph that captures the mathematical skills and their interrelations from a reference dataset. This skill graph guides us in assigning quality scores to the target dataset, enabling us to select the top-ranked subset which is further used to pretrain LLMs. Experimental results demonstrate the efficiency and effectiveness of MASS across different model sizes (1B and 7B) and pretraining datasets (web data and synthetic data). Specifically, in terms of efficiency, models trained on subsets selected by MASS can achieve similar performance to models trained on the original datasets, with a significant reduction in the number of trained tokens - ranging from 50\% to 70\% fewer tokens. In terms of effectiveness, when trained on the same amount of tokens, models trained on the data selected by MASS outperform those trained on the original datasets by 3.3\% to 5.9\%. These results underscore the potential of MASS to improve both the efficiency and effectiveness of pretraining LLMs.

Jiazheng Li, Artem Bobrov, Runcong Zhao et al.

Explainability in automated student answer scoring systems is critical for building trust and enhancing usability among educators. Yet, generating high-quality assessment rationales remains challenging due to the scarcity of annotated data and the prohibitive cost of manual verification, prompting heavy reliance on rationales produced by large language models (LLMs), which are often noisy and unreliable. To address these limitations, we present AERA Chat, an interactive visualization platform designed for automated explainable student answer assessment. AERA Chat leverages multiple LLMs to concurrently score student answers and generate explanatory rationales, offering innovative visualization features that highlight critical answer components and rationale justifications. The platform also incorporates intuitive annotation and evaluation tools, supporting educators in marking tasks and researchers in evaluating rationale quality from different models. We demonstrate the effectiveness of our platform through evaluations of multiple rationale-generation methods on several datasets, showcasing its capability for facilitating robust rationale evaluation and comparative analysis.

Xiaoguang Guo, Zehong Wang, Jiazheng Li et al.

Deployed graph neural networks (GNNs) are frozen at deployment yet must fit clean data, generalize under distribution shifts, and remain stable to perturbations. We show that static inference induces a fundamental tradeoff: improving stability requires reducing reliance on shift-sensitive features, leaving an irreducible worst-case generalization floor. Instance-conditional routing can break this ceiling, but is fragile because shifts can mislead routing and perturbations can make routing fluctuate. We capture these effects via two decompositions separating coverage vs selection, and base sensitivity vs fluctuation amplification. Based on these insights, we propose STEM-GNN, a pretrain-then-finetune framework with a mixture-of-experts encoder for diverse computation paths, a vector-quantized token interface to stabilize encoder-to-head signals, and a Lipschitz-regularized head to bound output amplification. Across nine node, link, and graph benchmarks, STEM-GNN achieves a stronger three-way balance, improving robustness to degree/homophily shifts and to feature/edge corruptions while remaining competitive on clean graphs.

Jiazheng Li, Andreas Damianou, J Rosser et al.

Chain-of-thought (CoT) traces promise transparency for reasoning language models, but prior work shows they are not always faithful reflections of internal computation. This raises challenges for oversight: practitioners may misinterpret decorative reasoning as genuine. We introduce Concept Walk, a general framework for tracing how a model's internal stance evolves with respect to a concept direction during reasoning. Unlike surface text, Concept Walk operates in activation space, projecting each reasoning step onto the concept direction learned from contrastive data. This allows us to observe whether reasoning traces shape outcomes or are discarded. As a case study, we apply Concept Walk to the domain of Safety using Qwen 3-4B. We find that in 'easy' cases, perturbed CoTs are quickly ignored, indicating decorative reasoning, whereas in 'hard' cases, perturbations induce sustained shifts in internal activations, consistent with faithful reasoning. The contribution is methodological: Concept Walk provides a lens to re-examine faithfulness through concept-specific internal dynamics, helping identify when reasoning traces can be trusted and when they risk misleading practitioners.

Jiazheng Li, Yawei Wang, David Yan et al.

Large Language Models (LLMs) have demonstrated remarkable capabilities, enabling language agents to excel at single-turn tasks. However, their application to complex, multi-step, and long-horizon tasks remains challenging. While reinforcement learning (RL) offers a promising avenue for addressing these challenges, mainstream approaches typically rely solely on sparse, outcome-based rewards, a limitation that becomes especially problematic for group-based RL algorithms lacking critic models, such as Group Relative Policy Optimization (GRPO). In such methods, uniformly rewarding or penalizing all actions within a trajectory can lead to training instability and suboptimal policies, because beneficial and detrimental actions are often entangled across multi-step interactions. To address this challenge, we propose SALT, a novel and lightweight framework that provides a finer-grained advantage assignment, derived solely from outcome rewards. We achieve this by constructing a graph from trajectories of the same prompt, which allows us to quantify the quality of each step and assign advantages accordingly. Crucially, SALT is designed as a plug-and-play module that seamlessly integrates with existing group-based RL algorithms, requiring no modifications to the rollout procedure and introducing negligible computational overhead. Extensive experiments on the WebShop, ALFWorld, and AppWorld benchmarks with various model sizes demonstrate that SALT consistently improves performance. We also conduct a thorough analysis to validate the design choices behind SALT and offer actionable insights.

Yang Li, Hanjie Wang, Yuanzheng Li et al.

Wind power data often suffers from missing values due to sensor faults and unstable transmission at edge sites. While federated learning enables privacy-preserving collaboration without sharing raw data, it remains vulnerable to anomalous updates and privacy leakage during parameter exchange. These challenges are amplified in open industrial environments, necessitating zero-trust mechanisms where no participant is inherently trusted. To address these challenges, this work proposes ZTFed-MAS2S, a zero-trust federated learning framework that integrates a multi-head attention-based sequence-to-sequence imputation model. ZTFed integrates verifiable differential privacy with non-interactive zero-knowledge proofs and a confidentiality and integrity verification mechanism to ensure verifiable privacy preservation and secure model parameters transmission. A dynamic trust-aware aggregation mechanism is employed, where trust is propagated over similarity graphs to enhance robustness, and communication overhead is reduced via sparsity- and quantization-based compression. MAS2S captures long-term dependencies in wind power data for accurate imputation. Extensive experiments on real-world wind farm datasets validate the superiority of ZTFed-MAS2S in both federated learning performance and missing data imputation, demonstrating its effectiveness as a secure and efficient solution for practical applications in the energy sector.

Weizhi Fei, Hao Shi, Jing Xu et al. · tsinghua

Efficiently editing knowledge stored in large language models (LLMs) enables model updates without large-scale training. One possible solution is Locate-and-Edit (L\&E), allowing simultaneous modifications of a massive number of facts. However, such editing may compromise the general abilities of LLMs and even result in forgetting edited facts when scaling up to thousands of edits. In this paper, we model existing linear L\&E methods as querying a Key-Value (KV) database. From this perspective, we then propose NeuralDB, an editing framework that explicitly represents the edited facts as a neural KV database equipped with a non-linear gated retrieval module, % In particular, our gated module only operates when inference involves the edited facts, effectively preserving the general abilities of LLMs. Comprehensive experiments involving the editing of 10,000 facts were conducted on the ZsRE and CounterFacts datasets, using GPT2-XL, GPT-J (6B) and Llama-3 (8B). The results demonstrate that NeuralDB not only excels in editing efficacy, generalization, specificity, fluency, and consistency, but also preserves overall performance across six representative text understanding and generation tasks. Further experiments indicate that NeuralDB maintains its effectiveness even when scaled to 100,000 facts (\textbf{50x} more than in prior work).

Jian Zhou, Jiazheng Li, Sirui Zhuge et al.

This paper proposed an approach to automatically discovering subject dimension, action dimension, object dimension and adverbial dimension from texts to efficiently operate texts and support query in natural language. The high quality of trees guarantees that all subjects, actions, objects and adverbials and their subclass relations within texts can be represented. The independency of trees ensures that there is no redundant representation between trees. The expressiveness of trees ensures that the majority of sentences can be accessed from each tree and the rest of sentences can be accessed from at least one tree so that the tree-based search mechanism can support querying in natural language. Experiments show that the average precision, recall and F1-score of the abstraction trees constructed by the subclass relations of subject, action, object and adverbial are all greater than 80%. The application of the proposed approach to supporting query in natural language demonstrates that different types of question patterns for querying subject or object have high coverage of texts, and searching multiple trees on subject, action, object and adverbial according to the question pattern can quickly reduce search space to locate target sentences, which can support precise operation on texts.

Zifan Chen, Jiazheng Li, Jie Zhao et al.

**Background:** Accurate 3D CT scan segmentation of gastric tumors is pivotal for diagnosis and treatment. The challenges lie in the irregular shapes, blurred boundaries of tumors, and the inefficiency of existing methods. **Purpose:** We conducted a study to introduce a model, utilizing human-guided knowledge and unique modules, to address the challenges of 3D tumor segmentation. **Methods:** We developed the PropNet framework, propagating radiologists' knowledge from 2D annotations to the entire 3D space. This model consists of a proposing stage for coarse segmentation and a refining stage for improved segmentation, using two-way branches for enhanced performance and an up-down strategy for efficiency. **Results:** With 98 patient scans for training and 30 for validation, our method achieves a significant agreement with manual annotation (Dice of 0.803) and improves efficiency. The performance is comparable in different scenarios and with various radiologists' annotations (Dice between 0.785 and 0.803). Moreover, the model shows improved prognostic prediction performance (C-index of 0.620 vs. 0.576) on an independent validation set of 42 patients with advanced gastric cancer. **Conclusions:** Our model generates accurate tumor segmentation efficiently and stably, improving prognostic performance and reducing high-throughput image reading workload. This model can accelerate the quantitative analysis of gastric tumors and enhance downstream task performance.

Jiazheng Li, Runcong Zhao, Yongxin Yang et al.

The remarkable performance of pre-trained large language models has revolutionised various natural language processing applications. Due to huge parametersizes and extensive running costs, companies or organisations tend to transfer the models to the target task by zero-shot prompting techniques. However, the prohibitive costs of tokens and time have hindered their adoption in applications. We propose OverPrompt, leveraging the in-context learning capability of LLMs to handle multiple task inputs, thereby reducing token and time costs. This approach could potentially improve task performance during API queries due to better conditional distribution mapping. Evaluated across diverse classification datasets, our experiments show that OverPrompt can achieve cost-efficient zero-shot classification without causing significant detriment to task performance, and in some cases, even improving it. An ablation study conducted on various LLMs, along with an investigation into the robustness of our prompting strategy to different input ordering, offers valuable insights into the broader applicability of our method across diverse tasks. These findings also suggest a more seamless integration of our method with LLMs through an API.

Linyi Yang, Jiazheng Li, Ruihai Dong et al.

Financial forecasting has been an important and active area of machine learning research because of the challenges it presents and the potential rewards that even minor improvements in prediction accuracy or forecasting may entail. Traditionally, financial forecasting has heavily relied on quantitative indicators and metrics derived from structured financial statements. Earnings conference call data, including text and audio, is an important source of unstructured data that has been used for various prediction tasks using deep earning and related approaches. However, current deep learning-based methods are limited in the way that they deal with numeric data; numbers are typically treated as plain-text tokens without taking advantage of their underlying numeric structure. This paper describes a numeric-oriented hierarchical transformer model to predict stock returns, and financial risk using multi-modal aligned earnings calls data by taking advantage of the different categories of numbers (monetary, temporal, percentages etc.) and their magnitude. We present the results of a comprehensive evaluation of NumHTML against several state-of-the-art baselines using a real-world publicly available dataset. The results indicate that NumHTML significantly outperforms the current state-of-the-art across a variety of evaluation metrics and that it has the potential to offer significant financial gains in a practical trading context.

Yang Li, Jiazheng Li, Yi Wang

Scenario generation is a fundamental and crucial tool for decision-making in power systems with high-penetration renewables. Based on big historical data, a novel federated deep generative learning framework, called Fed-LSGAN, is proposed by integrating federated learning and least square generative adversarial networks (LSGANs) for renewable scenario generation. Specifically, federated learning learns a shared global model in a central server from renewable sites at network edges, which enables the Fed-LSGAN to generate scenarios in a privacy-preserving manner without sacrificing the generation quality by transferring model parameters, rather than all data. Meanwhile, the LSGANs-based deep generative model generates scenarios that conform to the distribution of historical data through fully capturing the spatial-temporal characteristics of renewable powers, which leverages the least squares loss function to improve the training stability and generation quality. The simulation results demonstrate that the proposal manages to generate high-quality renewable scenarios and outperforms the state-of-the-art centralized methods. Besides, an experiment with different federated learning settings is designed and conducted to verify the robustness of our method.

Linyi Yang, Jiazheng Li, Pádraig Cunningham et al.

While state-of-the-art NLP models have been achieving the excellent performance of a wide range of tasks in recent years, important questions are being raised about their robustness and their underlying sensitivity to systematic biases that may exist in their training and test data. Such issues come to be manifest in performance problems when faced with out-of-distribution data in the field. One recent solution has been to use counterfactually augmented datasets in order to reduce any reliance on spurious patterns that may exist in the original data. Producing high-quality augmented data can be costly and time-consuming as it usually needs to involve human feedback and crowdsourcing efforts. In this work, we propose an alternative by describing and evaluating an approach to automatically generating counterfactual data for data augmentation and explanation. A comprehensive evaluation on several different datasets and using a variety of state-of-the-art benchmarks demonstrate how our approach can achieve significant improvements in model performance when compared to models training on the original data and even when compared to models trained with the benefit of human-generated augmented data.