Searcharxiv⌕ Search

arXiv subjects

Yada Zhu

Publications and source records attributed to Yada Zhu.

At least 19 recordsLinked to original sources
arXiv2026

SafeDream: Safety World Model for Proactive Early Jailbreak Detection

Multi-turn jailbreak attacks progressively erode LLM safety alignment across seemingly innocuous conversation turns, achieving success rates exceeding 90% against state-of-the-art models. Existing alignment-based and guardrail methods suffer from three key limitations: they require costly weight modification, evaluate each turn independently without modeling cumulative safety erosion, and detect attacks only after harmful content has been generated. To address these limitations, we first formulate the proactive early jailbreak detection problem with a new metric, detection lead, that measures how early an attack can be detected before the LLM complies. We then propose SAFEDREAM, a lightweight world-model-based framework that operates as an external module without modifying the LLM's weights. SAFEDREAM introduces three components: (1) a safety state world model that encodes LLM hidden states into a compact safety representation and predicts how it evolves across turns, (2) CUSUM detection that accumulates weak per-turn risk signals into reliable evidence, and (3) contrastive imagination that simultaneously rolls out attack and benign futures in latent space to issue early alarms before jailbreaks occur. On three multi-turn jailbreak benchmarks (XGuard-Train, SafeDialBench, SafeMTData) against 8 baselines, SAFEDREAM achieves the best detection timeliness across all benchmarks (1.06-1.20 turns before compliance) while maintaining competitive false positive rates and outperforming baselines in detection quality.

cs.CR↗
arXiv2026

MC-Search: Evaluating and Enhancing Multimodal Agentic Search with Structured Long Reasoning Chains

With the increasing demand for step-wise, cross-modal, and knowledge-grounded reasoning, multimodal large language models (MLLMs) are evolving beyond the traditional fixed retrieve-then-generate paradigm toward more sophisticated agentic multimodal retrieval-augmented generation (MM-RAG). Existing benchmarks, however, mainly focus on simplified QA with short retrieval chains, leaving adaptive planning and multimodal reasoning underexplored. We present MC-Search, the first benchmark for agentic MM-RAG with long, step-wise annotated reasoning chains spanning five representative reasoning structures. Each example specifies sub-questions, retrieval modalities, supporting facts, and intermediate answers, with fidelity ensured by HAVE (Hop-wise Attribution and Verification of Evidence), resulting in 3,333 high-quality examples averaging 3.7 hops. Beyond answer accuracy, MC-Search introduces new process-level metrics for reasoning quality, stepwise retrieval and planning accuracy. By developing a unified agentic MM-RAG pipeline, we benchmark six leading MLLMs and reveal systematic issues such as over- and under-retrieval and modality-misaligned planning. Finally, we introduce Search-Align, a process-supervised fine-tuning framework leveraging verified reasoning chains, showing that our data not only enables faithful evaluation but also improves planning and retrieval fidelity in open-source MLLMs.

cs.AI↗
arXiv2026

How Much Reasoning Do Retrieval-Augmented Models Add beyond LLMs? A Benchmarking Framework for Multi-Hop Inference over Hybrid Knowledge

Large language models (LLMs) continue to struggle with knowledge-intensive questions that require up-to-date information and multi-hop reasoning. Augmenting LLMs with hybrid external knowledge, such as unstructured text and structured knowledge graphs, offers a promising alternative to costly continual pretraining. As such, reliable evaluation of their retrieval and reasoning capabilities becomes critical. However, many existing benchmarks increasingly overlap with LLM pretraining data, which means answers or supporting knowledge may already be encoded in model parameters, making it difficult to distinguish genuine retrieval and reasoning from parametric recall. We introduce HybridRAG-Bench, a framework for constructing benchmarks to evaluate retrieval-intensive, multi-hop reasoning over hybrid knowledge. HybridRAG-Bench automatically couples unstructured text and structured knowledge graph representations derived from recent scientific literature on arXiv, and generates knowledge-intensive question-answer pairs grounded in explicit reasoning paths. The framework supports flexible domain and time-frame selection, enabling contamination-aware and customizable evaluation as models and knowledge evolve. Experiments across three domains (artificial intelligence, governance and policy, and bioinformatics) demonstrate that HybridRAG-Bench rewards genuine retrieval and reasoning rather than parametric recall, offering a principled testbed for evaluating hybrid knowledge-augmented reasoning systems. We release our code and data at github.com/junhongmit/HybridRAG-Bench.

cs.LG↗
arXiv2026

Mem-Gallery: Benchmarking Multimodal Long-Term Conversational Memory for MLLM Agents

Long-term memory is a critical capability for multimodal large language model (MLLM) agents, particularly in conversational settings where information accumulates and evolves over time. However, existing benchmarks either evaluate multi-session memory in text-only conversations or assess multimodal understanding within localized contexts, failing to evaluate how multimodal memory is preserved, organized, and evolved across long-term conversational trajectories. Thus, we introduce Mem-Gallery, a new benchmark for evaluating multimodal long-term conversational memory in MLLM agents. Mem-Gallery features high-quality multi-session conversations grounded in both visual and textual information, with long interaction horizons and rich multimodal dependencies. Building on this dataset, we propose a systematic evaluation framework that assesses key memory capabilities along three functional dimensions: memory extraction and test-time adaptation, memory reasoning, and memory knowledge management. Extensive benchmarking across thirteen memory systems reveals several key findings, highlighting the necessity of explicit multimodal information retention and memory organization, the persistent limitations in memory reasoning and knowledge management, as well as the efficiency bottleneck of current models.

cs.CL↗
arXiv2025

HeroFilter: Adaptive Spectral Graph Filter for Varying Heterophilic Relations

Graph heterophily, where connected nodes have different labels, has attracted significant interest recently. Most existing works adopt a simplified approach - using low-pass filters for homophilic graphs and high-pass filters for heterophilic graphs. However, we discover that the relationship between graph heterophily and spectral filters is more complex - the optimal filter response varies across frequency components and does not follow a strict monotonic correlation with heterophily degree. This finding challenges conventional fixed filter designs and suggests the need for adaptive filtering to preserve expressiveness in graph embeddings. Formally, natural questions arise: Given a heterophilic graph G, how and to what extent will the varying heterophily degree of G affect the performance of GNNs? How can we design adaptive filters to fit those varying heterophilic connections? Our theoretical analysis reveals that the average frequency response of GNNs and graph heterophily degree do not follow a strict monotonic correlation, necessitating adaptive graph filters to guarantee good generalization performance. Hence, we propose [METHOD NAME], a simple yet powerful GNN, which extracts information across the heterophily spectrum and combines salient representations through adaptive mixing. [METHOD NAME]'s superior performance achieves up to 9.2% accuracy improvement over leading baselines across homophilic and heterophilic graphs.

cs.LG↗
arXiv2025

Efficient and Transferable Agentic Knowledge Graph RAG via Reinforcement Learning

Knowledge-graph retrieval-augmented generation (KG-RAG) couples large language models (LLMs) with structured, verifiable knowledge graphs (KGs) to reduce hallucination and provide reasoning traces. However, current KG-RAG systems often rely on fixed pipelines of multiple LLM modules (e.g., planning, reasoning, and responding), which inflate inference costs and tie performance to specific graph schemas. To address this, we introduce KG-R1, an agentic framework that optimizes KG-RAG through reinforcement learning (RL). Unlike modular workflows, KG-R1 uses a single agent that interacts with KGs as its environment, learning to retrieve information at each step and incorporating it into its reasoning and generation in a unified process. Across Knowledge-Graph Question Answering (KGQA) benchmarks, KG-R1 demonstrates both efficiency and transferability-using Qwen 2.5-3B, KG-R1 improves answer accuracy with fewer generation tokens than prior multi-module workflow methods that use much larger foundation or fine-tuned models. Furthermore, KG-R1 exhibits strong plug-and-play capability: after training, maintaining accuracy on unseen KGs without retraining. These properties make KG-R1 a promising KG-RAG framework for real-world deployment. Our code is publicly available at github.com/junhongmit/KG-R1/.

cs.CL↗
arXiv2025

Temporal Reasoning with Large Language Models Augmented by Evolving Knowledge Graphs

Large language models (LLMs) excel at many language understanding tasks but struggle to reason over knowledge that evolves. To address this, recent work has explored augmenting LLMs with knowledge graphs (KGs) to provide structured, up-to-date information. However, many existing approaches assume a static snapshot of the KG and overlook the temporal dynamics and factual inconsistencies inherent in real-world data. To address the challenge of reasoning over temporally shifting knowledge, we propose EvoReasoner, a temporal-aware multi-hop reasoning algorithm that performs global-local entity grounding, multi-route decomposition, and temporally grounded scoring. To ensure that the underlying KG remains accurate and up-to-date, we introduce EvoKG, a noise-tolerant KG evolution module that incrementally updates the KG from unstructured documents through confidence-based contradiction resolution and temporal trend tracking. We evaluate our approach on temporal QA benchmarks and a novel end-to-end setting where the KG is dynamically updated from raw documents. Our method outperforms both prompting-based and KG-enhanced baselines, effectively narrowing the gap between small and large LLMs on dynamic question answering. Notably, an 8B-parameter model using our approach matches the performance of a 671B model prompted seven months later. These results highlight the importance of combining temporal reasoning with KG evolution for robust and up-to-date LLM performance. Our code is publicly available at github.com/junhongmit/TREK.

cs.LG↗
arXiv2025

Breaking Silos: Adaptive Model Fusion Unlocks Better Time Series Forecasting

Time-series forecasting plays a critical role in many real-world applications. Although increasingly powerful models have been developed and achieved superior results on benchmark datasets, through a fine-grained sample-level inspection, we find that (i) no single model consistently outperforms others across different test samples, but instead (ii) each model excels in specific cases. These findings prompt us to explore how to adaptively leverage the distinct strengths of various forecasting models for different samples. We introduce TimeFuse, a framework for collective time-series forecasting with sample-level adaptive fusion of heterogeneous models. TimeFuse utilizes meta-features to characterize input time series and trains a learnable fusor to predict optimal model fusion weights for any given input. The fusor can leverage samples from diverse datasets for joint training, allowing it to adapt to a wide variety of temporal patterns and thus generalize to new inputs, even from unseen datasets. Extensive experiments demonstrate the effectiveness of TimeFuse in various long-/short-term forecasting tasks, achieving near-universal improvement over the state-of-the-art individual models. Code is available at https://github.com/ZhiningLiu1998/TimeFuse.

cs.LG↗
arXiv2025

CLIMB: Class-imbalanced Learning Benchmark on Tabular Data

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.

cs.LG↗
arXiv2025

ClimateBench-M: A Multi-Modal Climate Data Benchmark with a Simple Generative Method

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.

cs.LG↗
arXiv2025

RAG over Tables: Hierarchical Memory Index, Multi-Stage Retrieval, and Benchmarking

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

cs.CL↗
arXiv2025

Reasoning of Large Language Models over Knowledge Graphs with Super-Relations

While large language models (LLMs) have made significant progress in processing and reasoning over knowledge graphs, current methods suffer from a high non-retrieval rate. This limitation reduces the accuracy of answering questions based on these graphs. Our analysis reveals that the combination of greedy search and forward reasoning is a major contributor to this issue. To overcome these challenges, we introduce the concept of super-relations, which enables both forward and backward reasoning by summarizing and connecting various relational paths within the graph. This holistic approach not only expands the search space, but also significantly improves retrieval efficiency. In this paper, we propose the ReKnoS framework, which aims to Reason over Knowledge Graphs with Super-Relations. Our framework's key advantages include the inclusion of multiple relation paths through super-relations, enhanced forward and backward reasoning capabilities, and increased efficiency in querying LLMs. These enhancements collectively lead to a substantial improvement in the successful retrieval rate and overall reasoning performance. We conduct extensive experiments on nine real-world datasets to evaluate ReKnoS, and the results demonstrate the superior performance of ReKnoS over existing state-of-the-art baselines, with an average accuracy gain of 2.92%.

cs.LG↗
arXiv2025

PLAY2PROMPT: Zero-shot Tool Instruction Optimization for LLM Agents via Tool Play

Large language models (LLMs) are increasingly integrated with specialized external tools, yet many tasks demand zero-shot tool usage with minimal or noisy documentation. Existing solutions rely on manual rewriting or labeled data for validation, making them inapplicable in true zero-shot settings. To address these challenges, we propose PLAY2PROMPT, an automated framework that systematically "plays" with each tool to explore its input-output behaviors. Through this iterative trial-and-error process, PLAY2PROMPT refines tool documentation and generates usage examples without any labeled data. These examples not only guide LLM inference but also serve as validation to further enhance tool utilization. Extensive experiments on real-world tasks demonstrate that PLAY2PROMPT significantly improves zero-shot tool performance across both open and closed models, offering a scalable and effective solution for domain-specific tool integration.

cs.CL↗
arXiv2025

Language in the Flow of Time: Time-Series-Paired Texts Weaved into a Unified Temporal Narrative

While many advances in time series models focus exclusively on numerical data, research on multimodal time series, particularly those involving contextual textual information, 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 effectively enable them to handle time series data with paired texts. 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 multimodal predictive performance without modifying model architectures. Our Code is available at https://github.com/iDEA-iSAIL-Lab-UIUC/TaTS.

cs.LG↗
arXiv2024

Enterprise Benchmarks for Large Language Model Evaluation

The advancement of large language models (LLMs) has led to a greater challenge of having a rigorous and systematic evaluation of complex tasks performed, especially in enterprise applications. Therefore, LLMs need to be able to benchmark enterprise datasets for various tasks. This work presents a systematic exploration of benchmarking strategies tailored to LLM evaluation, focusing on the utilization of domain-specific datasets and consisting of a variety of NLP tasks. The proposed evaluation framework encompasses 25 publicly available datasets from diverse enterprise domains like financial services, legal, cyber security, and climate and sustainability. The diverse performance of 13 models across different enterprise tasks highlights the importance of selecting the right model based on the specific requirements of each task. Code and prompts are available on GitHub.

cs.CL↗
arXiv2024

Generating Fine-Grained Causality in Climate Time Series Data for Forecasting and Anomaly Detection

Understanding the causal interaction of time series variables can contribute to time series data analysis for many real-world applications, such as climate forecasting and extreme weather alerts. However, causal relationships are difficult to be fully observed in real-world complex settings, such as spatial-temporal data from deployed sensor networks. Therefore, to capture fine-grained causal relations among spatial-temporal variables for further a more accurate and reliable time series analysis, we first design a conceptual fine-grained causal model named TBN Granger Causality, which adds time-respecting Bayesian Networks to the previous time-lagged Neural Granger Causality to offset the instantaneous effects. Second, we propose an end-to-end deep generative model called TacSas, which discovers TBN Granger Causality in a generative manner to help forecast time series data and detect possible anomalies during the forecast. For evaluations, besides the causality discovery benchmark Lorenz-96, we also test TacSas on climate benchmark ERA5 for climate forecasting and the extreme weather benchmark of NOAA for extreme weather alerts.

cs.LG↗
arXiv2024

When Heterophily Meets Heterogeneity: Challenges and a New Large-Scale Graph Benchmark

Graph mining has become crucial in fields such as social science, finance, and cybersecurity. Many large-scale real-world networks exhibit both heterogeneity, where multiple node and edge types exist in the graph, and heterophily, where connected nodes may have dissimilar labels and attributes. However, existing benchmarks primarily focus on either heterophilic homogeneous graphs or homophilic heterogeneous graphs, leaving a significant gap in understanding how models perform on graphs with both heterogeneity and heterophily. To bridge this gap, we introduce H2GB, a large-scale node-classification graph benchmark that brings together the complexities of both the heterophily and heterogeneity properties of real-world graphs. H2GB encompasses 9 real-world datasets spanning 5 diverse domains, 28 baseline models, and a unified benchmarking library with a standardized data loader, evaluator, unified modeling framework, and an extensible framework for reproducibility. We establish a standardized workflow supporting both model selection and development, enabling researchers to easily benchmark graph learning methods. Extensive experiments across 28 baselines reveal that current methods struggle with heterophilic and heterogeneous graphs, underscoring the need for improved approaches. Finally, we present a new variant of the model, H2G-former, developed following our standardized workflow, that excels at this challenging benchmark. Both the benchmark and the framework are publicly available at Github and PyPI, with documentation hosted at https://junhongmit.github.io/H2GB.

cs.LG↗
arXiv2024

AIM: Attributing, Interpreting, Mitigating Data Unfairness

Data collected in the real world often encapsulates historical discrimination against disadvantaged groups and individuals. Existing fair machine learning (FairML) research has predominantly focused on mitigating discriminative bias in the model prediction, with far less effort dedicated towards exploring how to trace biases present in the data, despite its importance for the transparency and interpretability of FairML. To fill this gap, we investigate a novel research problem: discovering samples that reflect biases/prejudices from the training data. Grounding on the existing fairness notions, we lay out a sample bias criterion and propose practical algorithms for measuring and countering sample bias. The derived bias score provides intuitive sample-level attribution and explanation of historical bias in data. On this basis, we further design two FairML strategies via sample-bias-informed minimal data editing. They can mitigate both group and individual unfairness at the cost of minimal or zero predictive utility loss. Extensive experiments and analyses on multiple real-world datasets demonstrate the effectiveness of our methods in explaining and mitigating unfairness. Code is available at https://github.com/ZhiningLiu1998/AIM.

cs.LG↗