Zihao Li

Jui-Hui Chung, Ziyang Cai, Zihao Li et al.

We introduce Goedel-Architect, an agentic framework for formal theorem proving in Lean 4 centered on blueprint generation and refinement. A blueprint is a dependency graph of definitions and lemmas that builds up to the main theorem. First, Goedel-Architect generates a blueprint of formally stated definitions and lemmas, along with declared dependencies. This blueprint is optionally guided by a natural language proof. Then, a tool-equipped Lean prover component closes each open lemma node in parallel using relevant dependencies. Failed lemmas in turn drive refinement of the global blueprint. This strategy contrasts with other mainstream approaches which use recursive lemma decomposition, and can inefficiently loop on dead-end strategies. Using the open-weight DeepSeek-V4-Flash (284B-A13B) as the backbone, Goedel-Architect attains 99.2% pass@1 on MiniF2F-test and 75.6% pass@1 on PutnamBench. With an optional natural-language proof seeding the initial blueprint on the harder problems, we additionally close the remaining two MiniF2F-test problems (reaching 100%), lift PutnamBench to 88.8% (597/672), and solve 4/6 on IMO 2025, 11/12 on Putnam 2025, and 3/6 on USAMO 2026. This represents state-of-the-art performance for an open-source pipeline at a price point up to 500x less than comparable open-source pipelines.

Zihao Li, Kaifeng Jin, Yuanchen Bei et al.

Time series are often embedded in rich contexts that are essential for holistic modeling. Moreover, real-world practitioners often require end-to-end workflows for analyzing temporal dynamics, where widely studied tasks such as forecasting are only one step in a broader solution loop. While generalist AI agents offer a promising interface for such workflows under complex contexts, they still operate primarily in textual spaces that are not fully aligned with structured temporal signals. In this work, we introduce TimeClaw, an agentic harness framework for time series that equips generalist LLM agents with the time series-native runtime support needed for contextualized temporal reasoning. TimeClaw integrates executable temporal tools for grounded and auditable analysis, experience-driven capability evolution for creating reusable analytical routines, and episodic multimodal memory for retrieving relevant reasoning traces. Together, these components unlock harnessed open-ended temporal reasoning with contextual information. Extensive evaluation on multiple benchmarks covering diverse tasks across energy, finance, weather, traffic, and other real-world domains demonstrates improved performance of TimeClaw. Code is available at https://github.com/iDEA-iSAIL-Lab-UIUC/TimeClaw.

Abdelaziz M. A. Ibrahim, Zihao Li, Jörg Tiedemann et al.

Large-scale multilingual bitext often contains two distinct problems: non-parallel sentence pairs and low-quality translations. We decompose model-based assessment for such data into two independent components: parallelism assessment with multilingual embeddings and reference-free quality estimation (QE). For parallelism, we benchmark four embedding models on FLORES-200 and BOUQuET retrieval tasks, covering 6,654 source--target directions in our target language-pair inventory. For QE, we evaluate nine reference-free evaluators on professional FLORES-200 translations across 41,412 ordered source--target directions. Results show that no model is universally reliable across translation directions. Naive QE ensembles dilute strong model signals, while documented target-language coverage is strongly associated with higher QE scores. Overall, these findings suggest that multilingual parallel-data assessment is best approached as a direction-aware routing and calibration problem, where no single universal metric is expected to suffice across all languages.

Guangyuan Dong, Ziwei Hong, Shenghao Liu et al.

Although Multimodal Sentiment Analysis (MSA) effectively leverages rich information from language, visual, and acoustic modalities, existing methods still face two core challenges: 1) static conflict suppression mechanisms fail to adapt to dynamic variations across samples, and 2) the inherent sentimental bias within the language modality, which can misguide learning from other modalities, remains entangled. To this end, we propose a Dynamic Multimodal Causal Disentanglement and Adaptive Fusion Framework (MCAF). Its cornerstone is the Multi-Granularity Causal Dynamic Router and a Conditional Diffusion Denoising Module. First, we introduce a causal intervention module based on the information bottleneck principle, which builds a Structural Causal Model to disentangle sentimental bias from language features, yielding a "de-confounded" language representation as a pure guiding signal. Second, we devise a Dynamic Multimodal Router that evaluates the interaction states (complementary, conflicting, or redundant) among visual, acoustic, and de-confounded language signals in real-time across three levels: feature, temporal, and modality, then adaptively allocates weights and routes information flow for fine-grained regulation. Finally, a lightweight Conditional Diffusion Denoising Module performs iterative denoising on the fused joint representation to explicitly filter out residual irrelevant information, generating a robust hyper-modality representation. Extensive experiments on the CMU-MOSI and CMU-MOSEI benchmarks show that MCAF sets new state-of-the-art on key classification metrics, achieving an Acc-2/F1 of 86.52%/86.51% on MOSI and 86.72%/86.65% on MOSEI, while remaining highly competitive on others. Comprehensive analyses and visualizations further validate its efficacy in dynamically perceiving interactions, disentangling bias, and enhancing interpretability.

Zihao Li, Samuel Belkadi, Nicolo Micheletti et al.

Biomedical literature often uses complex language and inaccessible professional terminologies. That is why simplification plays an important role in improving public health literacy. Applying Natural Language Processing (NLP) models to automate such tasks allows for quick and direct accessibility for lay readers. In this work, we investigate the ability of state-of-the-art large language models (LLMs) on the task of biomedical abstract simplification, using the publicly available dataset for plain language adaptation of biomedical abstracts (\textbf{PLABA}). The methods applied include domain fine-tuning and prompt-based learning (PBL) on: 1) Encoder-decoder models (T5, SciFive, and BART), 2) Decoder-only GPT models (GPT-3.5 and GPT-4) from OpenAI and BioGPT, and 3) Control-token mechanisms on BART-based models. We used a range of automatic evaluation metrics, including BLEU, ROUGE, SARI, and BERTscore, and also conducted human evaluations. BART-Large with Control Token (BART-L-w-CT) mechanisms reported the highest SARI score of 46.54 and T5-base reported the highest BERTscore 72.62. In human evaluation, BART-L-w-CTs achieved a better simplicity score over T5-Base (2.9 vs. 2.2), while T5-Base achieved a better meaning preservation score over BART-L-w-CTs (3.1 vs. 2.6). We also categorised the system outputs with examples, hoping this will shed some light for future research on this task. Our code, fine-tuned models, and data splits are available at \url{https://github.com/HECTA-UoM/PLABA-MU} \begin{IEEEkeywords} Large Language Models, Text Simplification, Biomedical NLP, Control Mechanisms, Health Informatics \end{IEEEkeywords}

Zihao Li, Pan Gao, Hui Yuan et al.

Discovering inter-point connection for efficient high-dimensional feature extraction from point coordinate is a key challenge in processing point cloud. Most existing methods focus on designing efficient local feature extractors while ignoring global connection, or vice versa. In this paper, we design a new Inductive Bias-aided Transformer (IBT) method to learn 3D inter-point relations, which considers both local and global attentions. Specifically, considering local spatial coherence, local feature learning is performed through Relative Position Encoding and Attentive Feature Pooling. We incorporate the learned locality into the Transformer module. The local feature affects value component in Transformer to modulate the relationship between channels of each point, which can enhance self-attention mechanism with locality based channel interaction. We demonstrate its superiority experimentally on classification and segmentation tasks. The code is available at: https://github.com/jiamang/IBT

Zihao Li, Pan Gao, Hui Yuan et al.

Existing point cloud learning methods aggregate features from neighbouring points relying on constructing graph in the spatial domain, which results in feature update for each point based on spatially-fixed neighbours throughout layers. In this paper, we propose a dynamic feature aggregation (DFA) method that can transfer information by constructing local graphs in the feature domain without spatial constraints. By finding k-nearest neighbors in the feature domain, we perform relative position encoding and semantic feature encoding to explore latent position and feature similarity information, respectively, so that rich local features can be learned. At the same time, we also learn low-dimensional global features from the original point cloud for enhancing feature representation. Between DFA layers, we dynamically update the constructed local graph structure, so that we can learn richer information, which greatly improves adaptability and efficiency. We demonstrate the superiority of our method by conducting extensive experiments on point cloud classification and segmentation tasks. Implementation code is available: https://github.com/jiamang/DFA.

Stephan Oepen, Nikolay Arefev, Mikko Aulamo et al.

We present an ongoing initiative to provide open, very large, high-quality, and richly annotated textual datasets for almost 200 languages. At 30 trillion tokens, this is likely the largest generally available multilingual collection of LLM pre-training data. These datasets are derived from web crawls from different sources and accompanied with a complete, open-source pipeline for document selection from web archives, text extraction from HTML, language identification for noisy texts, exact and near-deduplication, annotation with, among others, register labels, text quality estimates, and personally identifiable information; and final selection and filtering. We report on data quality probes through contrastive and analytical statistics, through manual inspection of samples for 24 languages, and through end-to-end evaluation of various language model architectures trained on this data. For multilingual LLM evaluation, we provide a comprehensive collection of benchmarks for nine European languages, with special emphasis on natively created tasks, mechanisms to mitigate prompt sensitivity, and refined normalization and aggregation of scores. Additionally, we train and evaluate a family of 57 monolingual encoder-decoder models, as well as a handful of monolingual GPT-like reference models. Besides the monolingual data and models, we also present a very large collection of parallel texts automatically mined from this data, together with a novel parallel corpus synthesized via machine translation.

Hui Li, Zihao Li, Rui Ma et al.

Class activation map (CAM) has been widely studied for visual explanation of the internal working mechanism of convolutional neural networks. The key of existing CAM-based methods is to compute effective weights to combine activation maps in the target convolution layer. Existing gradient and score based weighting schemes have shown superiority in ensuring either the discriminability or faithfulness of the CAM, but they normally cannot excel in both properties. In this paper, we propose a novel CAM weighting scheme, named FD-CAM, to improve both the faithfulness and discriminability of the CAM-based CNN visual explanation. First, we improve the faithfulness and discriminability of the score-based weights by performing a grouped channel switching operation. Specifically, for each channel, we compute its similarity group and switch the group of channels on or off simultaneously to compute changes in the class prediction score as the weights. Then, we combine the improved score-based weights with the conventional gradient-based weights so that the discriminability of the final CAM can be further improved. We perform extensive comparisons with the state-of-the-art CAM algorithms. The quantitative and qualitative results show our FD-CAM can produce more faithful and more discriminative visual explanations of the CNNs. We also conduct experiments to verify the effectiveness of the proposed grouped channel switching and weight combination scheme on improving the results. Our code is available at https://github.com/crishhh1998/FD-CAM.

Xinhang Li, Zihao Li, Nan Yang et al.

The expansion of renewable energy could help realizing the goals of peaking carbon dioxide emissions and carbon neutralization. Some existing grid dispatching methods integrating short-term renewable energy prediction and reinforcement learning (RL) have been proved to alleviate the adverse impact of energy fluctuations risk. However, these methods omit the long-term output prediction, which leads to stability and security problems on the optimal power flow. This paper proposes a confidence estimation Transformer for long-term renewable energy forecasting in reinforcement learning-based power grid dispatching (Conformer-RLpatching). Conformer-RLpatching predicts long-term active output of each renewable energy generator with an enhanced Transformer to boost the performance of hybrid energy grid dispatching. Furthermore, a confidence estimation method is proposed to reduce the prediction error of renewable energy. Meanwhile, a dispatching necessity evaluation mechanism is put forward to decide whether the active output of a generator needs to be adjusted. Experiments carried out on the SG-126 power grid simulator show that Conformer-RLpatching achieves great improvement over the second best algorithm DDPG in security score by 25.8% and achieves a better total reward compared with the golden medal team in the power grid dispatching competition sponsored by State Grid Corporation of China under the same simulation environment. Codes are outsourced in https://github.com/buptlxh/Conformer-RLpatching.

Baoyu Jing, Sanhorn Chen, Lecheng Zheng et al.

Time series data are integral to critical applications across domains such as finance, healthcare, transportation, and environmental science. While recent work has begun to explore multi-task time series question answering (QA), current benchmarks remain limited to forecasting and anomaly detection tasks. We introduce TSAQA, a novel unified benchmark designed to broaden task coverage and evaluate diverse temporal analysis capabilities. TSAQA integrates six diverse tasks under a single framework ranging from conventional analysis, including anomaly detection and classification, to advanced analysis, such as characterization, comparison, data transformation, and temporal relationship analysis. Spanning 210k samples across 13 domains, the dataset employs diverse formats, including true-or-false (TF), multiple-choice (MC), and a novel puzzling (PZ), to comprehensively assess time series analysis. Zero-shot evaluation demonstrates that these tasks are challenging for current Large Language Models (LLMs): the best-performing commercial LLM, Gemini-2.5-Flash, achieves an average score of only 65.08. Although instruction tuning boosts open-source performance: the best-performing open-source model, LLaMA-3.1-8B, shows significant room for improvement, highlighting the complexity of temporal analysis for LLMs.

Shaoxiong Ji, Zihao Li, Jaakko Paavola et al.

In this work, we introduce EMMA-500, a large-scale multilingual language model continue-trained on texts across 546 languages designed for enhanced multilingual performance, focusing on improving language coverage for low-resource languages. To facilitate continual pre-training, we compile the MaLA corpus, a comprehensive multilingual dataset enriched with curated datasets across diverse domains. Leveraging this corpus, we conduct extensive continual pre-training of the Llama 2 7B model, resulting in EMMA-500, which demonstrates robust performance across a wide collection of benchmarks, including a comprehensive set of multilingual tasks. Our results highlight the effectiveness of continual pre-training in expanding large language models' language capacity, particularly for underrepresented languages, demonstrating significant gains in cross-lingual transfer, task generalization, and language adaptability. We release the MaLA corpus, EMMA-500 model weights, scripts, and model generations.

Chaoran Zhang, Lixin Zou, Dan Luo et al.

In recent years, Large Language Models (LLMs) have demonstrated remarkable capabilities across a wide array of text-centric tasks. However, their `large' scale introduces significant computational and storage challenges, particularly in managing the key-value states of the transformer, which limits their wider applicability. Therefore, we propose to adaptively release resources from caches and rebuild the necessary key-value states. Particularly, we accomplish this by a lightweight controller module to approximate an ideal top-$K$ sparse attention. This module retains the tokens with the highest top-$K$ attention weights and simultaneously rebuilds the discarded but necessary tokens, which may become essential for future decoding. Comprehensive experiments in natural language generation and modeling reveal that our method is not only competitive with full attention in terms of performance but also achieves a significant throughput improvement of up to 221.8%. The code for replication is available on the https://github.com/WHUIR/ADORE.

Zihao Li, Wenchuan Wu, Boming Zhang

For its high coefficient of performance and zero local emissions, the heat pump (HP) has recently become popular in North Europe and China. However, the integration of HPs may aggravate the daily peak-valley gap in distribution networks significantly.

Jiacheng Guo, Zihao Li, Huazheng Wang et al.

In this paper, we study representation learning in partially observable Markov Decision Processes (POMDPs), where the agent learns a decoder function that maps a series of high-dimensional raw observations to a compact representation and uses it for more efficient exploration and planning. We focus our attention on the sub-classes of \textit{$γ$-observable} and \textit{decodable POMDPs}, for which it has been shown that statistically tractable learning is possible, but there has not been any computationally efficient algorithm. We first present an algorithm for decodable POMDPs that combines maximum likelihood estimation (MLE) and optimism in the face of uncertainty (OFU) to perform representation learning and achieve efficient sample complexity, while only calling supervised learning computational oracles. We then show how to adapt this algorithm to also work in the broader class of $γ$-observable POMDPs.

Shengbo Gao, Ziji Zhang, Jiechao Ma et al.

Semi-supervised learning has become increasingly popular in medical image segmentation due to its ability to leverage large amounts of unlabeled data to extract additional information. However, most existing semi-supervised segmentation methods only focus on extracting information from unlabeled data, disregarding the potential of labeled data to further improve the performance of the model. In this paper, we propose a novel Correlation Aware Mutual Learning (CAML) framework that leverages labeled data to guide the extraction of information from unlabeled data. Our approach is based on a mutual learning strategy that incorporates two modules: the Cross-sample Mutual Attention Module (CMA) and the Omni-Correlation Consistency Module (OCC). The CMA module establishes dense cross-sample correlations among a group of samples, enabling the transfer of label prior knowledge to unlabeled data. The OCC module constructs omni-correlations between the unlabeled and labeled datasets and regularizes dual models by constraining the omni-correlation matrix of each sub-model to be consistent. Experiments on the Atrial Segmentation Challenge dataset demonstrate that our proposed approach outperforms state-of-the-art methods, highlighting the effectiveness of our framework in medical image segmentation tasks. The codes, pre-trained weights, and data are publicly available.

Zihao Li, Shaoxiong Ji, Timothee Mickus et al.

Pretrained language models (PLMs) display impressive performances and have captured the attention of the NLP community. Establishing best practices in pretraining has, therefore, become a major focus of NLP research, especially since insights gained from monolingual English models may not necessarily apply to more complex multilingual models. One significant caveat of the current state of the art is that different works are rarely comparable: they often discuss different parameter counts, training data, and evaluation methodology. This paper proposes a comparison of multilingual pretraining objectives in a controlled methodological environment. We ensure that training data and model architectures are comparable, and discuss the downstream performances across 6 languages that we observe in probing and fine-tuning scenarios. We make two key observations: (1) the architecture dictates which pretraining objective is optimal; (2) multilingual translation is a very effective pretraining objective under the right conditions. We make our code, data, and model weights available at \texttt{\url{https://github.com/Helsinki-NLP/lm-vs-mt}}.

Chris Kelly, Luhui Hu, Cindy Yang et al.

In the current landscape of artificial intelligence, foundation models serve as the bedrock for advancements in both language and vision domains. OpenAI GPT-4 has emerged as the pinnacle in large language models (LLMs), while the computer vision (CV) domain boasts a plethora of state-of-the-art (SOTA) models such as Meta's SAM and DINO, and YOLOS. However, the financial and computational burdens of training new models from scratch remain a significant barrier to progress. In response to this challenge, we introduce UnifiedVisionGPT, a novel framework designed to consolidate and automate the integration of SOTA vision models, thereby facilitating the development of vision-oriented AI. UnifiedVisionGPT distinguishes itself through four key features: (1) provides a versatile multimodal framework adaptable to a wide range of applications, building upon the strengths of multimodal foundation models; (2) seamlessly integrates various SOTA vision models to create a comprehensive multimodal platform, capitalizing on the best components of each model; (3) prioritizes vision-oriented AI, ensuring a more rapid progression in the CV domain compared to the current trajectory of LLMs; and (4) introduces automation in the selection of SOTA vision models, generating optimal results based on diverse multimodal inputs such as text prompts and images. This paper outlines the architecture and capabilities of UnifiedVisionGPT, demonstrating its potential to revolutionize the field of computer vision through enhanced efficiency, versatility, generalization, and performance. Our implementation, along with the unified multimodal framework and comprehensive dataset, is made publicly available at https://github.com/LHBuilder/SA-Segment-Anything.

Zihao Li

With the launch of ChatGPT, Large Language Models (LLMs) are shaking up our whole society, rapidly altering the way we think, create and live. For instance, the GPT integration in Bing has altered our approach to online searching. While nascent LLMs have many advantages, new legal and ethical risks are also emerging, stemming in particular from stochastic parrots and hallucination. The EU is the first and foremost jurisdiction that has focused on the regulation of AI models. However, the risks posed by the new LLMs are likely to be underestimated by the emerging EU regulatory paradigm. Therefore, this correspondence warns that the European AI regulatory paradigm must evolve further to mitigate such risks.

Zihao Li, Ji Guan, Mingsheng Ying

We present a tool QSeqSim, a Qiskit-integrated symbolic backend that fills the current gap of having no Qiskit-native support for simulating while-loop quantum programs and their induced sequential quantum circuits. QSeqSim takes Qiskit QuantumCircuit objects, translates them into OpenQASM 3 code, and organises the resulting program into a combination of combinational, dynamic, and sequential circuits, thereby assigning while-loops a precise sequential circuit semantics with explicit internal and external qubits. Building on this semantics, QSeqSim adopts a Binary Decision Diagram (BDD)-based symbolic representation and integrates weighted model counting to compute measurement probabilities efficiently by exploiting sharing in structured and sparse BDDs. On top of this Boolean backbone, it introduces dedicated symbolic operators for state composition and state retention, thereby enabling efficient symbolic execution of sequential quantum circuits. Our experiments demonstrate that QSeqSim scales to substantial while-induced sequential circuits; in particular, in the quantum random walk benchmark we successfully simulate circuits with over 1000 qubits for more than 10 loop iterations. QSeqSim is available at https://github.com/Veri-Q/QSeqSim.

Jinxiang Xie, Zihao Li, Wei He et al.

Text analysis of tabular data relies on two core operations: \emph{summarization} for corpus-level theme extraction and \emph{tagging} for row-level labeling. A critical limitation of employing large language models (LLMs) for these tasks is their inability to meet the high standards of output stability demanded by data analytics. To address this challenge, we introduce \textbf{CAST} (\textbf{C}onsistency via \textbf{A}lgorithmic Prompting and \textbf{S}table \textbf{T}hinking), a framework that enhances output stability by constraining the model's latent reasoning path. CAST combines (i) Algorithmic Prompting to impose a procedural scaffold over valid reasoning transitions and (ii) Thinking-before-Speaking to enforce explicit intermediate commitments before final generation. To measure progress, we introduce \textbf{CAST-S} and \textbf{CAST-T}, stability metrics for bulleted summarization and tagging, and validate their alignment with human judgments. Experiments across publicly available benchmarks on multiple LLM backbones show that CAST consistently achieves the best stability among all baselines, improving Stability Score by up to 16.2\%, while maintaining or improving output quality.

Boyan Xu, Liang Wen, Zihao Li et al.

Recent advancements in Large Language Models (LLMs) have sparked interest in their potential applications across various fields. This paper embarked on a pivotal inquiry: Can existing LLMs effectively serve as "water expert models" for water engineering and research tasks? This study was the first to evaluate LLMs' contributions across various water engineering and research tasks by establishing a domain-specific benchmark suite, namely, WaterER. Herein, we prepared 983 tasks related to water engineering and research, categorized into "wastewater treatment", "environmental restoration", "drinking water treatment and distribution", "sanitation", "anaerobic digestion" and "contaminants assessment". We evaluated the performance of seven LLMs (i.e., GPT-4, GPT-3.5, Gemini, GLM-4, ERNIE, QWEN and Llama3) on these tasks. We highlighted the strengths of GPT-4 in handling diverse and complex tasks of water engineering and water research, the specialized capabilities of Gemini in academic contexts, Llama3's strongest capacity to answer Chinese water engineering questions and the competitive performance of Chinese-oriented models like GLM-4, ERNIE and QWEN in some water engineering tasks. More specifically, current LLMs excelled particularly in generating precise research gaps for papers on "contaminants and related water quality monitoring and assessment". Additionally, they were more adept at creating appropriate titles for research papers on "treatment processes for wastewaters", "environmental restoration", and "drinking water treatment". Overall, this study pioneered evaluating LLMs in water engineering and research by introducing the WaterER benchmark to assess the trustworthiness of their predictions. This standardized evaluation framework would also drive future advancements in LLM technology by using targeting datasets, propelling these models towards becoming true "water expert".

Zihao Li, Samuel Belkadi, Nicolo Micheletti et al.

In this system report, we describe the models and methods we used for our participation in the PLABA2023 task on biomedical abstract simplification, part of the TAC 2023 tracks. The system outputs we submitted come from the following three categories: 1) domain fine-tuned T5-like models including Biomedical-T5 and Lay-SciFive; 2) fine-tuned BARTLarge model with controllable attributes (via tokens) BART-w-CTs; 3) ChatGPTprompting. We also present the work we carried out for this task on BioGPT finetuning. In the official automatic evaluation using SARI scores, BeeManc ranks 2nd among all teams and our model LaySciFive ranks 3rd among all 13 evaluated systems. In the official human evaluation, our model BART-w-CTs ranks 2nd on Sentence-Simplicity (score 92.84), 3rd on Term-Simplicity (score 82.33) among all 7 evaluated systems; It also produced a high score 91.57 on Fluency in comparison to the highest score 93.53. In the second round of submissions, our team using ChatGPT-prompting ranks the 2nd in several categories including simplified term accuracy score 92.26 and completeness score 96.58, and a very similar score on faithfulness score 95.3 to re-evaluated PLABA-base-1 (95.73) via human evaluations. Our codes, fine-tuned models, prompts, and data splits from the system development stage will be available at https://github.com/ HECTA-UoM/PLABA-MU

Keshu Wu, Chenchen Kuai, Zihao Li et al.

Retrieval-augmented generation (RAG) enhances large language models by grounding outputs in retrieved knowledge. However, existing RAG methods including graph- and hypergraph-based approaches treat retrieved evidence as an unordered set, implicitly assuming permutation invariance. This assumption is misaligned with many real-world reasoning tasks, where outcomes depend not only on which interactions occur, but also on the order in which they unfold. We propose Order-Aware Knowledge Hypergraph RAG (OKH-RAG), which treats order as a first-class structural property. OKH-RAG represents knowledge as higher-order interactions within a hypergraph augmented with precedence structure, and reformulates retrieval as sequence inference over hyperedges. Instead of selecting independent facts, it recovers coherent interaction trajectories that reflect underlying reasoning processes. A learned transition model infers precedence directly from data without requiring explicit temporal supervision. We evaluate OKH-RAG on order-sensitive question answering and explanation tasks, including tropical cyclone and port operation scenarios. OKH-RAG consistently outperforms permutation-invariant baselines, and ablations show that these gains arise specifically from modeling interaction order. These results highlight a key limitation of set-based retrieval: effective reasoning requires not only retrieving relevant evidence, but organizing it into structured sequences.

Zihao Li, Changkun Jiang, Jianqiang Li

Automatic protein function prediction (AFP) is classified as a large-scale multi-label classification problem aimed at automating protein enrichment analysis to eliminate the current reliance on labor-intensive wet-lab methods. Currently, popular methods primarily combine protein-related information and Gene Ontology (GO) terms to generate final functional predictions. For example, protein sequences, structural information, and protein-protein interaction networks are integrated as prior knowledge to fuse with GO term embeddings and generate the ultimate prediction results. However, these methods are limited by the difficulty in obtaining structural information or network topology information, as well as the accuracy of such data. Therefore, more and more methods that only use protein sequences for protein function prediction have been proposed, which is a more reliable and computationally cheaper approach. However, the existing methods fail to fully extract feature information from protein sequences or label data because they do not adequately consider the intrinsic characteristics of the data itself. Therefore, we propose a sequence-based hierarchical prediction method, DeepGATGO, which processes protein sequences and GO term labels hierarchically, and utilizes graph attention networks (GATs) and contrastive learning for protein function prediction. Specifically, we compute embeddings of the sequence and label data using pre-trained models to reduce computational costs and improve the embedding accuracy. Then, we use GATs to dynamically extract the structural information of non-Euclidean data, and learn general features of the label dataset with contrastive learning by constructing positive and negative example samples. Experimental results demonstrate that our proposed model exhibits better scalability in GO term enrichment analysis on large-scale datasets.

Zihao Li, Chunhua Wu, Shenglin Wu et al.

In recent years, some researchers focused on using a single image to obtain a large number of samples through multi-scale features. This study intends to a brand-new idea that requires only ten or even fewer samples to construct the low-rank structural-Hankel matrices-assisted score-based generative model (SHGM) for color image inpainting task. During the prior learning process, a certain amount of internal-middle patches are firstly extracted from several images and then the structural-Hankel matrices are constructed from these patches. To better apply the score-based generative model to learn the internal statistical distribution within patches, the large-scale Hankel matrices are finally folded into the higher dimensional tensors for prior learning. During the iterative inpainting process, SHGM views the inpainting problem as a conditional generation procedure in low-rank environment. As a result, the intermediate restored image is acquired by alternatively performing the stochastic differential equation solver, alternating direction method of multipliers, and data consistency steps. Experimental results demonstrated the remarkable performance and diversity of SHGM.

Xuying Ning, Katherine Tieu, Dongqi Fu et al.

Recent large language models (LLMs) have demonstrated strong capabilities in understanding and generating code, from competitive programming to repository-level software engineering. In emerging agentic systems, code is no longer only a target output. It increasingly serves as an operational substrate for agent reasoning, acting, environment modeling, and execution-based verification. We frame this shift through the lens of agent harnesses and introduce code as agent harness: a unified view that centers code as the basis for agent infrastructure. To systematically study this perspective, we organize the survey around three connected layers. First, we study the harness interface, where code connects agents to reasoning, action, and environment modeling. Second, we examine harness mechanisms: planning, memory, and tool use for long-horizon execution, together with feedback-driven control and optimization that make harness reliable and adaptive. Third, we discuss scaling the harness from single-agent systems to multi-agent settings, where shared code artifacts support multi-agent coordination, review, and verification. Across these layers, we summarize representative methods and practical applications of code as agent harness, spanning coding assistants, GUI/OS automation, embodied agents, scientific discovery, personalization and recommendation, DevOps, and enterprise workflows. We further outline open challenges for harness engineering, including evaluation beyond final task success, verification under incomplete feedback, regression-free harness improvement, consistent shared state across multiple agents, human oversight for safety-critical actions, and extensions to multimodal environments. By centering code as the harness of agentic AI, this survey provides a unified roadmap toward executable, verifiable, and stateful AI agent systems.

Zihao Li, Dongqi Fu, Mengting Ai et al.

Knowledge graphs (KGs), which store an extensive number of relational facts, serve various applications. Recently, personalized knowledge graphs (PKGs) have emerged as a solution to optimize storage costs by customizing their content to align with users' specific interests within particular domains. In the real world, on one hand, user queries and their underlying interests are inherently evolving, requiring PKGs to adapt continuously; on the other hand, the summarization is constantly expected to be as small as possible in terms of storage cost. However, the existing PKG summarization methods implicitly assume that the user's interests are constant and do not shift. Furthermore, when the size constraint of PKG is extremely small, the existing methods cannot distinguish which facts are more of immediate interest and guarantee the utility of the summarized PKG. To address these limitations, we propose APEX$^2$, a highly scalable PKG summarization framework designed with robust theoretical guarantees to excel in adaptive summarization tasks with extremely small size constraints. To be specific, after constructing an initial PKG, APEX$^2$ continuously tracks the interest shift and adjusts the previous summary. We evaluate APEX$^2$ under an evolving query setting on benchmark KGs containing up to 12 million triples, summarizing with compression ratios $\leq 0.1\%$. The experiments show that APEX outperforms state-of-the-art baselines in terms of both query-answering accuracy and efficiency. Code is available at https://github.com/iDEA-iSAIL-Lab-UIUC/APEX.

Zihao Li, Lecheng Zheng, Bowen Jin et al.

While great success has been achieved in building vision models with Contrastive Language-Image Pre-training (CLIP) over internet-scale image-text pairs, building transferable Graph Neural Networks (GNNs) with CLIP pipeline is challenging because of the scarcity of labeled data and text supervision, different levels of downstream tasks, and the conceptual gaps between domains. In this work, to address these issues, we propose a multi-modal prompt learning paradigm to effectively adapt pre-trained GNN to downstream tasks and data, given only a few semantically labeled samples, each with extremely weak text supervision. Our new paradigm embeds the graphs directly in the same space as the Large Language Models (LLMs) by learning both graph prompts and text prompts simultaneously. We demonstrate the superior performance of our paradigm in few-shot, multi-task-level, and cross-domain settings. Moreover, we build the first CLIP-style zero-shot classification prototype that can generalize GNNs to unseen classes with extremely weak text supervision. The code is available at https://github.com/Violet24K/Morpher.

Zihao Li, Yuyi Ao, Jingrui He

Knowledge graphs (KGs), which store an extensive number of relational facts (head, relation, tail), serve various applications. While many downstream tasks highly rely on the expressive modeling and predictive embedding of KGs, most of the current KG representation learning methods, where each entity is embedded as a vector in the Euclidean space and each relation is embedded as a transformation, follow an entity ranking protocol. On one hand, such an embedding design cannot capture many-to-many relations. On the other hand, in many retrieval cases, the users wish to get an exact set of answers without any ranking, especially when the results are expected to be precise, e.g., which genes cause an illness. Such scenarios are commonly referred to as "set retrieval". This work presents a pioneering study on the KG set retrieval problem. We show that the set retrieval highly depends on expressive modeling of many-to-many relations, and propose a new KG embedding model SpherE to address this problem. SpherE is based on rotational embedding methods, but each entity is embedded as a sphere instead of a vector. While inheriting the high interpretability of rotational-based models, our SpherE can more expressively model one-to-many, many-to-one, and many-to-many relations. Through extensive experiments, we show that our SpherE can well address the set retrieval problem while still having a good predictive ability to infer missing facts. The code is available at https://github.com/Violet24K/SpherE.

Lecheng Zheng, Baoyu Jing, Zihao Li et al.

Graph Self-Supervised Learning (SSL) has emerged as a pivotal area of research in recent years. By engaging in pretext tasks to learn the intricate topological structures and properties of graphs using unlabeled data, these graph SSL models achieve enhanced performance, improved generalization, and heightened robustness. Despite the remarkable achievements of these graph SSL methods, their current implementation poses significant challenges for beginners and practitioners due to the complex nature of graph structures, inconsistent evaluation metrics, and concerns regarding reproducibility hinder further progress in this field. Recognizing the growing interest within the research community, there is an urgent need for a comprehensive, beginner-friendly, and accessible toolkit consisting of the most representative graph SSL algorithms. To address these challenges, we present a Graph SSL toolkit named PyG-SSL, which is built upon PyTorch and is compatible with various deep learning and scientific computing backends. Within the toolkit, we offer a unified framework encompassing dataset loading, hyper-parameter configuration, model training, and comprehensive performance evaluation for diverse downstream tasks. Moreover, we provide beginner-friendly tutorials and the best hyper-parameters of each graph SSL algorithm on different graph datasets, facilitating the reproduction of results. The GitHub repository of the library is https://github.com/iDEA-iSAIL-Lab-UIUC/pyg-ssl.

Yikun Ban, Jiaru Zou, Zihao Li et al.

Link prediction is a critical problem in graph learning with broad applications such as recommender systems and knowledge graph completion. Numerous research efforts have been directed at solving this problem, including approaches based on similarity metrics and Graph Neural Networks (GNN). However, most existing solutions are still rooted in conventional supervised learning, which makes it challenging to adapt over time to changing customer interests and to address the inherent dilemma of exploitation versus exploration in link prediction. To tackle these challenges, this paper reformulates link prediction as a sequential decision-making process, where each link prediction interaction occurs sequentially. We propose a novel fusion algorithm, PRB (PageRank Bandits), which is the first to combine contextual bandits with PageRank for collaborative exploitation and exploration. We also introduce a new reward formulation and provide a theoretical performance guarantee for PRB. Finally, we extensively evaluate PRB in both online and offline settings, comparing it with bandit-based and graph-based methods. The empirical success of PRB demonstrates the value of the proposed fusion approach. Our code is released at https://github.com/jiaruzouu/PRB.

Zihao Li, Xiao Lin, Zhining Liu et al.

While many advances in time series models focus exclusively on numerical data, research on multimodal time series, particularly those involving contextual textual information commonly encountered in real-world scenarios, remains in its infancy. With recent progress in large language models and time series learning, we revisit the integration of paired texts with time series through the Platonic Representation Hypothesis, which posits that representations of different modalities converge to shared spaces. In this context, we identify that time-series-paired texts may naturally exhibit periodic properties that closely mirror those of the original time series. Building on this insight, we propose a novel framework, Texts as Time Series (TaTS), which considers the time-series-paired texts to be auxiliary variables of the time series. TaTS can be plugged into any existing numerical-only time series models and enable them to handle time series data with paired texts effectively. Through extensive experiments on both multimodal time series forecasting and imputation tasks across benchmark datasets with various existing time series models, we demonstrate that TaTS can enhance predictive performance without modifying model architectures. Code available at https://github.com/iDEA-iSAIL-Lab-UIUC/TaTS.

Chaolun Ma, Bruce Wang, Zihao Li et al.

In traffic signal control, flow-based (optimizing the overall flow) and pressure-based methods (equalizing and alleviating congestion) are commonly used but often considered separately. This study introduces a unified framework using Lyapunov control theory, defining specific Lyapunov functions respectively for these methods. We have found interesting results. For example, the well-recognized back-pressure method is equal to differential queue lengths weighted by intersection lane saturation flows. We further improve it by adding basic traffic flow theory. Rather than ensuring that the control system be stable, the system should be also capable of adaptive to various performance metrics. Building on insights from Lyapunov theory, this study designs a reward function for the Reinforcement Learning (RL)-based network signal control, whose agent is trained with Double Deep Q-Network (DDQN) for effective control over complex traffic networks. The proposed algorithm is compared with several traditional and RL-based methods under pure passenger car flow and heterogenous traffic flow including freight, respectively. The numerical tests demonstrate that the proposed method outperforms the alternative control methods across different traffic scenarios, covering corridor and general network situations each with varying traffic demands, in terms of the average network vehicle waiting time per vehicle.

Yang Zhao, Zihao Li, Zhiyu Jiang et al.

While Large Language Models (LLMs) form the cornerstone of sequential decision-making agent development, they have inherent limitations in high-frequency decision tasks. Existing research mainly focuses on discrete embodied decision scenarios with low-frequency and significant semantic differences in state space (e.g., household planning). These methods suffer from limited performance in high-frequency decision-making tasks, since high-precision numerical state information in such tasks undergoes frequent updates with minimal fluctuations, and exhibiting policy misalignment between the learned sub-tasks and composite tasks. To address these issues, this paper proposes Normalized Action Reward guided Consistency Policy Optimization (NAR-CP). 1) Our method first acquires predefined dense rewards from environmental feedback of candidate actions via reward functions, then completes reward shaping through normalization, and theoretically verifies action reward normalization does not impair optimal policy. 2) To reduce policy misalignment in composite tasks, we use LLMs to infer sub-observation candidate actions and generate joint policies, with consistency loss ensuring precise alignment between global semantic policies and sub-semantic policies. Experiments on UAV pursuit, a typical high-frequency task, show our method delivers superior performance on independent and composite tasks with excellent generalization to unseen tasks.

Jiaru Zou, Yikun Ban, Zihao Li et al.

Large language models are typically adapted to downstream tasks through supervised fine-tuning on domain-specific data. While standard fine-tuning focuses on minimizing generation loss to optimize model parameters, we take a deeper step by retaining and leveraging the model's own learning signals, analogous to how human learners reflect on past mistakes to improve future performance. We first introduce the concept of Mistake Log to systematically track the model's learning behavior and recurring errors throughout fine-tuning. Treating the original transformer-based model as the Pilot, we correspondingly design a Copilot model to refine the Pilot's inference performance via logits rectification. We name the overall Pilot-Copilot framework the Transformer Copilot, which introduces (i) a novel Copilot model design, (ii) a joint training paradigm where the Copilot continuously learns from the evolving Mistake Log alongside the Pilot, and (iii) a fused inference paradigm where the Copilot rectifies the Pilot's logits for enhanced generation. We provide both theoretical and empirical analyses on our new learning framework. Experiments on 12 benchmarks spanning commonsense, arithmetic, and recommendation tasks demonstrate that Transformer Copilot consistently improves performance by up to 34.5%, while introducing marginal computational overhead to Pilot models and exhibiting strong scalability and transferability. Our code is released at https://github.com/jiaruzouu/TransformerCopilot.

Keshu Wu, Sixu Li, Zihao Li et al.

Sensor networks increasingly govern modern infrastructure, yet the data they lose are rarely missing in the uniform-random patterns assumed by standard imputation benchmarks. Loop detectors go offline during calibration, roadside cabinets silence clusters of nearby sensors, and newly installed instruments provide no history. Such failures create structured absences whose values are constrained by higher-order relations among groups of sensors, not merely by pairwise proximity. Existing low-rank and graph-based methods often miss this collective structure and can fail when missingness becomes coherent. We introduce Multi-Scale Hypergraph Laplacians (MSHL), a two-stage framework for learning higher-order structure from incomplete spatiotemporal observations. The Discovery stage builds a multi-scale hypergraph from complementary topology and residual-correlation evidence, with an observation-only selector that adapts to the supported interaction scale. The Refinement stage adds a small hypergraph-conditioned residual network that is safe by construction: it learns nonlinear corrections where informative residual features exist and defers to the linear estimate where they do not. We prove that MSHL represents group-conservation patterns inaccessible to pairwise graph priors, adapts to the best fixed scale up to a logarithmic factor, transfers this advantage to held-out imputation error, and admits a one-sided refinement guarantee. On two real traffic networks evaluated across scattered cell missingness, contiguous block outages, and whole-sensor blackouts at five rates, MSHL improves over a pairwise-graph baseline whenever higher-order structure is identifiable and otherwise matches it within sampling noise. The results point to a broader principle for reliable infrastructure learning: missing data should be treated not as isolated entries to fill, but as evidence of structure to discover.

Zihao Li, Pan Gao, Kang You et al.

Previous studies have demonstrated the effectiveness of point-based neural models on the point cloud analysis task. However, there remains a crucial issue on producing the efficient input embedding for raw point coordinates. Moreover, another issue lies in the limited efficiency of neighboring aggregations, which is a critical component in the network stem. In this paper, we propose a Global Attention-guided Dual-domain Feature Learning network (GAD) to address the above-mentioned issues. We first devise the Contextual Position-enhanced Transformer (CPT) module, which is armed with an improved global attention mechanism, to produce a global-aware input embedding that serves as the guidance to subsequent aggregations. Then, the Dual-domain K-nearest neighbor Feature Fusion (DKFF) is cascaded to conduct effective feature aggregation through novel dual-domain feature learning which appreciates both local geometric relations and long-distance semantic connections. Extensive experiments on multiple point cloud analysis tasks (e.g., classification, part segmentation, and scene semantic segmentation) demonstrate the superior performance of the proposed method and the efficacy of the devised modules.

Zihao Li, Xiang Ji, Minshuo Chen et al.

A recently popular approach to solving reinforcement learning is with data from human preferences. In fact, human preference data are now used with classic reinforcement learning algorithms such as actor-critic methods, which involve evaluating an intermediate policy over a reward learned from human preference data with distribution shift, known as off-policy evaluation (OPE). Such algorithm includes (i) learning reward function from human preference dataset, and (ii) learning expected cumulative reward of a target policy. Despite the huge empirical success, existing OPE methods with preference data often lack theoretical understanding and rely heavily on heuristics. In this paper, we study the sample efficiency of OPE with human preference and establish a statistical guarantee for it. Specifically, we approach OPE by learning the value function by fitted-Q-evaluation with a deep neural network. By appropriately selecting the size of a ReLU network, we show that one can leverage any low-dimensional manifold structure in the Markov decision process and obtain a sample-efficient estimator without suffering from the curse of high data ambient dimensionality. Under the assumption of high reward smoothness, our results \textit{almost align with the classical OPE results with observable reward data}. To the best of our knowledge, this is the first result that establishes a \textit{provably efficient} guarantee for off-policy evaluation with RLHF.

Jiacheng Guo, Suozhi Huang, Zixin Yao et al.

This paper introduces CryptoBench, the first expert-curated, dynamic benchmark designed to rigorously evaluate the real-world capabilities of Large Language Model (LLM) agents in the uniquely demanding and fast-paced cryptocurrency domain. Unlike general-purpose agent benchmarks for search and prediction, professional crypto analysis presents specific challenges: \emph{extreme time-sensitivity}, \emph{a highly adversarial information environment}, and the critical need to synthesize data from \emph{diverse, specialized sources}, such as on-chain intelligence platforms and real-time Decentralized Finance (DeFi) dashboards. CryptoBench thus serves as a much more challenging and valuable scenario for LLM agent assessment. To address these challenges, we constructed a live, dynamic benchmark featuring 50 questions per month, expertly designed by crypto-native professionals to mirror actual analyst workflows. These tasks are rigorously categorized within a four-quadrant system: Simple Retrieval, Complex Retrieval, Simple Prediction, and Complex Prediction. This granular categorization enables a precise assessment of an LLM agent's foundational data-gathering capabilities alongside its advanced analytical and forecasting skills. Our evaluation of ten LLMs, both directly and within an agentic framework, reveals a performance hierarchy and uncovers a failure mode. We observe a \textit{retrieval-prediction imbalance}, where many leading models, despite being proficient at data retrieval, demonstrate a pronounced weakness in tasks requiring predictive analysis. This highlights a problematic tendency for agents to appear factually grounded while lacking the deeper analytical capabilities to synthesize information.

Shaoxiong Ji, Zihao Li, Jaakko Paavola et al.

This paper investigates a critical design decision in the practice of massively multilingual continual pre-training -- the inclusion of parallel data. Specifically, we study the impact of bilingual translation data for massively multilingual language adaptation of the Llama3 family of models to 500 languages. To this end, we construct the MaLA bilingual translation corpus, containing data from more than 2,500 language pairs. Subsequently, we develop the EMMA-500 Llama 3 suite of four massively multilingual models -- continually pre-trained from the Llama 3 family of base models extensively on diverse data mixes up to 671B tokens -- and explore the effect of continual pre-training with or without bilingual translation data. Comprehensive evaluation across 7 tasks and 12 benchmarks demonstrates that bilingual data tends to enhance language transfer and performance, particularly for low-resource languages. We open-source the MaLA corpus, EMMA-500 Llama 3 suite artefacts, code, and model generations.

Jiaru Zou, Dongqi Fu, Sirui Chen et al.

Retrieval-Augmented Generation (RAG) enhances Large Language Models (LLMs) by integrating them with an external knowledge base to improve the answer relevance and accuracy. In real-world scenarios, beyond pure text, a substantial amount of knowledge is stored in tables, and user questions often require retrieving answers that are distributed across multiple tables. Retrieving knowledge from a table corpora (i.e., various individual tables) for a question remains nascent, at least, for (i) how to understand intra- and inter-table knowledge effectively, (ii) how to filter unnecessary tables and how to retrieve the most relevant tables efficiently, (iii) how to prompt LLMs to infer over the retrieval, (iv) how to evaluate the corresponding performance in a realistic setting. Facing the above challenges, in this paper, we first propose a table-corpora-aware RAG framework, named T-RAG, which consists of the hierarchical memory index, multi-stage retrieval, and graph-aware prompting for effective and efficient table knowledge retrieval and inference. Further, we first develop a multi-table question answering benchmark named MultiTableQA, which spans 3 different task types, 57,193 tables, and 23,758 questions in total, and the sources are all from real-world scenarios. Based on MultiTableQA, we did the holistic comparison over table retrieval methods, RAG methods, and table-to-graph representation learning methods, where T-RAG shows the leading accuracy, recall, and running time performance. Also, under T-RAG, we evaluate the inference ability upgrade of different LLMs. Code and Data are available at https://github.com/jiaruzouu/T-RAG

Dongqi Fu, Yada Zhu, Zhining Liu et al.

Climate science studies the structure and dynamics of Earth's climate system and seeks to understand how climate changes over time, where the data is usually stored in the format of time series, recording the climate features, geolocation, time attributes, etc. Recently, much research attention has been paid to the climate benchmarks. In addition to the most common task of weather forecasting, several pioneering benchmark works are proposed for extending the modality, such as domain-specific applications like tropical cyclone intensity prediction and flash flood damage estimation, or climate statement and confidence level in the format of natural language. To further motivate the artificial general intelligence development for climate science, in this paper, we first contribute a multi-modal climate benchmark, i.e., ClimateBench-M, which aligns (1) the time series climate data from ERA5, (2) extreme weather events data from NOAA, and (3) satellite image data from NASA HLS based on a unified spatial-temporal granularity. Second, under each data modality, we also propose a simple but strong generative method that could produce competitive performance in weather forecasting, thunderstorm alerts, and crop segmentation tasks in the proposed ClimateBench-M. The data and code of ClimateBench-M are publicly available at https://github.com/iDEA-iSAIL-Lab-UIUC/ClimateBench-M.

Lecheng Zheng, Dongqi Fu, Zihao Li et al.

Graph data is informative to represent complex relationships such as transactions between accounts, communications between devices, and dependencies among machines or processes. Correspondingly, graph anomaly detection (GAD) plays a critical role in identifying anomalies across various domains, including finance, cybersecurity, manufacturing, etc. Facing the large-volume and multi-domain graph data, nascent efforts attempt to develop foundational generalist models capable of detecting anomalies in unseen graphs without retraining. To the best of our knowledge, the different feature semantics and dimensions of cross-domain graph data heavily hinder the development of the graph foundation model, leaving further in-depth continual learning and inference capabilities a quite open problem. Hence, we propose OWLEYE, a novel zero-shot GAD framework that learns transferable patterns of normal behavior from multiple graphs, with a threefold contribution. First, OWLEYE proposes a cross-domain feature alignment module to harmonize feature distributions, which preserves domain-specific semantics during alignment. Second, with aligned features, to enable continuous learning capabilities, OWLEYE designs the multi-domain multi-pattern dictionary learning to encode shared structural and attribute-based patterns. Third, for achieving the in-context learning ability, OWLEYE develops a truncated attention-based reconstruction module to robustly detect anomalies without requiring labeled data for unseen graph-structured data. Extensive experiments on real-world datasets demonstrate that OWLEYE achieves superior performance and generalizability compared to state-of-the-art baselines, establishing a strong foundation for scalable and label-efficient anomaly detection.

Chenchen Kuai, Chenhao Wu, Yang Zhou et al.

As tropical cyclones intensify and track forecasts become increasingly uncertain, U.S. ports face heightened supply-chain risk under extreme weather conditions. Port operators need to rapidly synthesize diverse multimodal forecast products, such as probabilistic wind maps, track cones, and official advisories, into clear, actionable guidance as cyclones approach. Multimodal large language models (MLLMs) offer a powerful means to integrate these heterogeneous data sources alongside broader contextual knowledge, yet their accuracy and reliability in the specific context of port cyclone preparedness have not been rigorously evaluated. To fill this gap, we introduce CyPortQA, the first multimodal benchmark tailored to port operations under cyclone threat. CyPortQA assembles 2,917 realworld disruption scenarios from 2015 through 2023, spanning 145 U.S. principal ports and 90 named storms. Each scenario fuses multisource data (i.e., tropical cyclone products, port operational impact records, and port condition bulletins) and is expanded through an automated pipeline into 117,178 structured question answer pairs. Using this benchmark, we conduct extensive experiments on diverse MLLMs, including both open-source and proprietary model. MLLMs demonstrate great potential in situation understanding but still face considerable challenges in reasoning tasks, including potential impact estimation and decision reasoning.

Zhining Liu, Zihao Li, Ze Yang et al.

Class-imbalanced learning (CIL) on tabular data is important in many real-world applications where the minority class holds the critical but rare outcomes. In this paper, we present CLIMB, a comprehensive benchmark for class-imbalanced learning on tabular data. CLIMB includes 73 real-world datasets across diverse domains and imbalance levels, along with unified implementations of 29 representative CIL algorithms. Built on a high-quality open-source Python package with unified API designs, detailed documentation, and rigorous code quality controls, CLIMB supports easy implementation and comparison between different CIL algorithms. Through extensive experiments, we provide practical insights on method accuracy and efficiency, highlighting the limitations of naive rebalancing, the effectiveness of ensembles, and the importance of data quality. Our code, documentation, and examples are available at https://github.com/ZhiningLiu1998/imbalanced-ensemble.

Kaitao Song, Xiaohua Wang, Xu Tan et al. · cmu, microsoft-research

In this paper, we propose a novel learning paradigm, termed Chain-of-Model (CoM), which incorporates the causal relationship into the hidden states of each layer as a chain style, thereby introducing great scaling efficiency in model training and inference flexibility in deployment. We introduce the concept of Chain-of-Representation (CoR), which formulates the hidden states at each layer as a combination of multiple sub-representations (i.e., chains) at the hidden dimension level. In each layer, each chain from the output representations can only view all of its preceding chains in the input representations. Consequently, the model built upon CoM framework can progressively scale up the model size by increasing the chains based on the previous models (i.e., chains), and offer multiple sub-models at varying sizes for elastic inference by using different chain numbers. Based on this principle, we devise Chain-of-Language-Model (CoLM), which incorporates the idea of CoM into each layer of Transformer architecture. Based on CoLM, we further introduce CoLM-Air by introducing a KV sharing mechanism, that computes all keys and values within the first chain and then shares across all chains. This design demonstrates additional extensibility, such as enabling seamless LM switching, prefilling acceleration and so on. Experimental results demonstrate our CoLM family can achieve comparable performance to the standard Transformer, while simultaneously enabling greater flexiblity, such as progressive scaling to improve training efficiency and offer multiple varying model sizes for elastic inference, paving a a new way toward building language models. Our code will be released in the future at: https://github.com/microsoft/CoLM.

Ming Deng, Sijin Sun, Zihao Li et al.

Camouflaged Object Detection (COD) is challenging due to the strong similarity between camouflaged objects and their surroundings, which complicates identification. Existing methods mainly rely on spatial local features, failing to capture global information, while Transformers increase computational costs. To address this, the Frequency-Assisted Mamba-Like Linear Attention Network (FMNet) is proposed, which leverages frequency-domain learning to efficiently capture global features and mitigate ambiguity between objects and the background. FMNet introduces the Multi-Scale Frequency-Assisted Mamba-Like Linear Attention (MFM) module, integrating frequency and spatial features through a multi-scale structure to handle scale variations while reducing computational complexity. Additionally, the Pyramidal Frequency Attention Extraction (PFAE) module and the Frequency Reverse Decoder (FRD) enhance semantics and reconstruct features. Experimental results demonstrate that FMNet outperforms existing methods on multiple COD datasets, showcasing its advantages in both performance and efficiency. Code available at https://github.com/Chranos/FMNet.

Zhidong Ling, Zihao Li, Pablo Romero et al.

This report is the system description of the MaLei team (Manchester and Leiden) for the shared task Plain Language Adaptation of Biomedical Abstracts (PLABA) 2024 (we had an earlier name BeeManc following last year), affiliated with TREC2024 (33rd Text REtrieval Conference https://ir.nist.gov/evalbase/conf/trec-2024). This report contains two sections corresponding to the two sub-tasks in PLABA-2024. In task one (term replacement), we applied fine-tuned ReBERTa-Base models to identify and classify the difficult terms, jargon, and acronyms in the biomedical abstracts and reported the F1 score (Task 1A and 1B). In task two (complete abstract adaptation), we leveraged Llamma3.1-70B-Instruct and GPT-4o with the one-shot prompts to complete the abstract adaptation and reported the scores in BLEU, SARI, BERTScore, LENS, and SALSA. From the official Evaluation from PLABA-2024 on Task 1A and 1B, our much smaller fine-tuned RoBERTa-Base model ranked 3rd and 2nd respectively on the two sub-tasks, and the 1st on averaged F1 scores across the two tasks from 9 evaluated systems. Our LLaMA-3.1-70B-instructed model achieved the highest Completeness score for Task 2. We share our source codes, fine-tuned models, and related resources at https://github.com/HECTA-UoM/PLABA2024

Natchaphon Leungbootnak, Zihao Li, Zihang Wei et al.

The ability of existing headway distributions to accurately reflect the diverse behaviors and characteristics in heterogeneous traffic (different types of vehicles) and mixed traffic (human-driven vehicles with autonomous vehicles) is limited, leading to unsatisfactory goodness of fit. To address these issues, we modified the exponential function to obtain a novel headway distribution. Rather than employing Euler's number (e) as the base of the exponential function, we utilized a real number base to provide greater flexibility in modeling the observed headway. However, the proposed is not a probability function. We normalize it to calculate the probability and derive the closed-form equation. In this study, we utilized a comprehensive experiment with five open datasets: highD, exiD, NGSIM, Waymo, and Lyft to evaluate the performance of the proposed distribution and compared its performance with six existing distributions under mixed and heterogeneous traffic flow. The results revealed that the proposed distribution not only captures the fundamental characteristics of headway distribution but also provides physically meaningful parameters that describe the distribution shape of observed headways. Under heterogeneous flow on highways (i.e., uninterrupted traffic flow), the proposed distribution outperforms other candidate distributions. Under urban road conditions (i.e., interrupted traffic flow), including heterogeneous and mixed traffic, the proposed distribution still achieves decent results.