Searcharxiv⌕ Search

arXiv subjects

Kaijie Zhu

Publications and source records attributed to Kaijie Zhu.

At least 19 recordsLinked to original sources
arXiv2026

Can LLMs Discover Scientific Laws in Real and Parallel Worlds?

Scientific equation discovery has long been central to scientific progress, proceeding through iterative cycles of hypothesis generation, observational testing, and refinement under scientific constraints. As LLM capabilities advance and their role in AI for Science expands, it remains an open problem whether they can genuinely discover scientific laws and how this ability should be evaluated. Existing evaluations, however, often either simplify discovery through synthetic settings or reuse published targets that may already be familiar to LLMs. We therefore introduce SCILAWS-BENCH, a benchmark for scientific law discovery built from published research and real scientific data. It comprises 118 problems drawn from 381 scientific papers, covering 291 candidate laws and roughly 8M real data points across six scientific disciplines. Each problem is instantiated in two complementary settings: (1) SCILAWS-REAL asks models to propose laws from fixed real observations and evaluates held-out predictive fit and scientific validity derived from the source literature, and (2) SCILAWS-PARALLEL asks models to actively query residual-calibrated worlds and recover synthesized hidden laws derived from published forms. This two-setting task design preserves each problem's scientific context while separately evaluating fixed-record law discovery and active recovery of a newly synthesized hidden law. We find that predictive fit can diverge from scientific validity, memorization shapes whether models reproduce or move beyond published formulas, and our best-of-N study reveals a selection bottleneck. Our work provides a paper-grounded benchmark and new empirical perspectives for evaluating AI for scientific discovery. Project page: https://yiyihum.github.io/SciLaws-Bench

cs.AI↗
arXiv2026

Benchmarking Trustworthiness of SLMs: Pre-trained vs. Compressed

Small Language Models (SLMs) have emerged as a more efficient alternative to traditional Large Language Models (LLMs), offering promising potential in resource-constrained scenarios. Existing approaches to building SLMs typically follow two paths: training compact models from scratch, or compressing larger pre-trained models using methods such as pruning, quantization, or distillation. As language models become increasingly integrated into real-world applications, ensuring their trustworthiness has become a critical concern. However, how to build trustworthy SLMs remains an underexplored question. In this work, we present a comprehensive evaluation of SLM trustworthiness across multiple dimensions, including fairness, robustness, privacy, and ethics. We first examine the effects of pruning and quantization, and find that quantization is significantly more effective in preserving trustworthiness compared to pruning. More importantly, we demonstrate that compressing a reliable large model via quantization can produce SLMs with superior trustworthiness and adaptability compared to using small models trained from scratch. Furthermore, knowledge distillation from trustworthy teacher models can further enhance the reliability of SLMs. We hope our findings provide practical guidance and a foundation for future research into the development and deployment of trustworthy small language models.

cs.CL↗
arXiv2026

SimSD: Simple Speculative Decoding in Diffusion Language Models

Diffusion large language models (dLLMs) have recently emerged as a promising alternative to autoregressive (AR) LLMs, offering faster inference through parallel or blockwise decoding. However, their masked language modeling formulation remains incompatible with standard token-level speculative decoding, one of the most effective acceleration techniques for AR models. In AR decoding, the causal mask preserves temporally valid token-level contexts, enabling a target model to verify multiple drafted tokens in a single forward pass. In contrast, dLLMs rely on mask tokens and bidirectional attention, causing the effective context to change across denoising steps and preventing direct token-level speculative verification. To bridge this gap, we propose a simple but effective speculative decoding algorithm for diffusion language models, named SimSD, which mainly adopts a plug-and-play masking strategy that equips dLLMs with temporally valid token-level contexts for speculative decoding. Our method explicitly introduces reference tokens from draft-model predictions and designs an attention mask that regulates their interaction with current-step tokens, allowing dLLMs to compute valid logits for drafted tokens in a single forward pass. This restores the key verification ability provided by causal masking in AR models while preserving the parallel decoding advantages of dLLMs. The proposed method is training-free and can be flexibly integrated with other acceleration techniques such as KV cache and blockwise decoding. Experiments on SDAR-family dLLMs across four benchmarks show that our method achieves up to 7.46x higher decoding throughput while maintaining and even improving average generation quality.

cs.CL↗
arXiv2026

rePIRL: Learn PRM with Inverse RL for LLM Reasoning

Process rewards have been widely used in deep reinforcement learning to improve training efficiency, reduce variance, and prevent reward hacking. In LLM reasoning, existing works also explore various solutions for learning effective process reward models (PRM) with or without the help of an expert policy. However, existing methods either rely on strong assumptions about the expert policies (e.g., requiring their reward functions) or suffer intrinsic limitations (e.g., entropy collapse), resulting in weak PRMs or limited generalizability. In this paper, we introduce rePIRL, an inverse RL-inspired framework that learns effective PRMs with minimal assumptions about expert policies. Specifically, we design a dual learning process that updates the policy and the PRM interchangeably. Our learning algorithm has customized techniques to address the challenges of scaling traditional inverse RL to LLMs. We theoretically show that our proposed learning framework can unify both online and offline PRM learning methods, justifying that rePIRL can learn PRMs with minimal assumptions. Empirical evaluations on standardized math and coding reasoning datasets demonstrate the effectiveness of rePIRL over existing methods. We further show the application of our trained PRM in test-time training, test-time scaling, and providing an early signal for training hard problems. Finally, we validate our training recipe and key design choices via a detailed ablation study.

cs.LG↗
arXiv2026

On the Trustworthiness of Generative Foundation Models: Guideline, Assessment, and Perspective

Generative Foundation Models (GenFMs) have emerged as transformative tools. However, their widespread adoption raises critical concerns regarding trustworthiness across dimensions. This paper presents a comprehensive framework to address these challenges through three key contributions. First, we systematically review global AI governance laws and policies from governments and regulatory bodies, as well as industry practices and standards. Based on this analysis, we propose a set of guiding principles for GenFMs, developed through extensive multidisciplinary collaboration that integrates technical, ethical, legal, and societal perspectives. Second, we introduce TrustGen, the first dynamic benchmarking platform designed to evaluate trustworthiness across multiple dimensions and model types, including text-to-image, large language, and vision-language models. TrustGen leverages modular components--metadata curation, test case generation, and contextual variation--to enable adaptive and iterative assessments, overcoming the limitations of static evaluation methods. Using TrustGen, we reveal significant progress in trustworthiness while identifying persistent challenges. Finally, we provide an in-depth discussion of the challenges and future directions for trustworthy GenFMs, which reveals the complex, evolving nature of trustworthiness, highlighting the nuanced trade-offs between utility and trustworthiness, and consideration for various downstream applications, identifying persistent challenges and providing a strategic roadmap for future research. This work establishes a holistic framework for advancing trustworthiness in GenAI, paving the way for safer and more responsible integration of GenFMs into critical applications. To facilitate advancement in the community, we release the toolkit for dynamic evaluation.

cs.CY↗
arXiv2026

AgentReview: Exploring Peer Review Dynamics with LLM Agents

Peer review is fundamental to the integrity and advancement of scientific publication. Traditional methods of peer review analyses often rely on exploration and statistics of existing peer review data, which do not adequately address the multivariate nature of the process, account for the latent variables, and are further constrained by privacy concerns due to the sensitive nature of the data. We introduce AgentReview, the first large language model (LLM) based peer review simulation framework, which effectively disentangles the impacts of multiple latent factors and addresses the privacy issue. Our study reveals significant insights, including a notable 37.1% variation in paper decisions due to reviewers' biases, supported by sociological theories such as the social influence theory, altruism fatigue, and authority bias. We believe that this study could offer valuable insights to improve the design of peer review mechanisms. Our code is available at https://github.com/Ahren09/AgentReview.

cs.CL↗
arXiv2026

Understanding and Mitigating Bias Inheritance in LLM-based Data Augmentation on Downstream Tasks

Generating synthetic datasets via large language models (LLMs) has emerged as a promising approach to improve LLM performance. However, LLMs inherently reflect biases in their training data, leading to a critical challenge: when models are trained on synthetic data, they may propagate and amplify the inherent biases that can significantly impact fairness and robustness on downstream tasks-a phenomenon we term bias inheritance. This work presents the first systematic investigation in understanding, analyzing, and mitigating bias inheritance. We fine-tune LLMs with a combined dataset of real and LLM-augmented data with varied bias ratio as the proportion of augmented data. Through systematic experiments across 10 classification and generation tasks, we analyze how 6 different types of biases manifest. Our results indicate that bias inheritance harms downstream task performance in bias directly-related classification and generation tasks. Then, our analysis identifies three key misalignment factors: misalignment of values, group data, and data distributions. Based on these insights, we propose three mitigation strategies: token-based, mask-based, and loss-based approaches, which can work differently on various tasks and bias, indicating the substantial challenges to mitigate bias inheritance. We hope this work can provide insights to the research of LLM data augmentation.

cs.LG↗
arXiv2026

OpenSage: Self-programming Agent Generation Engine

Agent development kits (ADKs) provide effective platforms and tooling for constructing agents, and their designs are critical to the constructed agents' performance, especially the functionality for agent topology, tools, and memory. However, current ADKs either lack sufficient functional support or rely on humans to manually design these components, limiting agents' generalizability and overall performance. We propose OpenSage, the first ADK that enables LLMs to automatically create agents with self-generated topology and toolsets while providing comprehensive and structured memory support. OpenSage offers effective functionality for agents to create and manage their own sub-agents and toolkits. It also features a hierarchical, graph-based memory system for efficient management and a specialized toolkit tailored to software engineering tasks. Extensive experiments across three state-of-the-art benchmarks with various backbone models demonstrate the advantages of OpenSage over existing ADKs. We also conduct rigorous ablation studies to demonstrate the effectiveness of our design for each component. We believe OpenSage can pave the way for the next generation of agent development, shifting the focus from human-centered to AI-centered paradigms.

cs.AI↗
arXiv2026

TermiGen: High-Fidelity Environment and Robust Trajectory Synthesis for Terminal Agents

Executing complex terminal tasks remains a significant challenge for open-weight LLMs, constrained by two fundamental limitations. First, high-fidelity, executable training environments are scarce: environments synthesized from real-world repositories are not diverse and scalable, while trajectories synthesized by LLMs suffer from hallucinations. Second, standard instruction tuning uses expert trajectories that rarely exhibit simple mistakes common to smaller models. This creates a distributional mismatch, leaving student models ill-equipped to recover from their own runtime failures. To bridge these gaps, we introduce TermiGen, an end-to-end pipeline for synthesizing verifiable environments and resilient expert trajectories. Termi-Gen first generates functionally valid tasks and Docker containers via an iterative multi-agent refinement loop. Subsequently, we employ a Generator-Critic protocol that actively injects errors during trajectory collection, synthesizing data rich in error-correction cycles. Fine-tuned on this TermiGen-generated dataset, our TermiGen-Qwen2.5-Coder-32B achieves a 31.3% pass rate on TerminalBench. This establishes a new open-weights state-of-the-art, outperforming existing baselines and notably surpassing capable proprietary models such as o4-mini. Dataset is avaiable at https://github.com/ucsb-mlsec/terminal-bench-env.

cs.AI↗
arXiv2026

DevOps-Gym: Benchmarking AI Agents in Software DevOps Cycle

Even though demonstrating extraordinary capabilities in code generation and software issue resolving, AI agents' capabilities in the full software DevOps cycle are still unknown. Different from pure code generation, handling the DevOps cycle in real-world software, including developing, deploying, and managing, requires analyzing large-scale projects, understanding dynamic program behaviors, leveraging domain-specific tools, and making sequential decisions. However, existing benchmarks focus on isolated problems and lack environments and tool interfaces for DevOps. We introduce DevOps-Gym, the first end-to-end benchmark for evaluating AI agents across core DevOps workflows: build and configuration, monitoring, issue resolving, and test generation. DevOps-Gym includes 700+ real-world tasks collected from 30+ projects in Java and Go. We develop a semi-automated data collection mechanism with rigorous and non-trivial expert efforts in ensuring the task coverage and quality. Our evaluation of state-of-the-art models and agents reveals fundamental limitations: they struggle with issue resolving and test generation in Java and Go, and remain unable to handle new tasks such as monitoring and build and configuration. These results highlight the need for essential research in automating the full DevOps cycle with AI agents.

cs.SE↗
arXiv2026

Personalized Safety in LLMs: A Benchmark and A Planning-Based Agent Approach

Large language models (LLMs) typically generate identical or similar responses for all users given the same prompt, posing serious safety risks in high-stakes applications where user vulnerabilities differ widely. Existing safety evaluations primarily rely on context-independent metrics - such as factuality, bias, or toxicity - overlooking the fact that the same response may carry divergent risks depending on the user's background or condition. We introduce personalized safety to fill this gap and present PENGUIN - a benchmark comprising 14,000 scenarios across seven sensitive domains with both context-rich and context-free variants. Evaluating six leading LLMs, we demonstrate that personalized user information significantly improves safety scores by 43.2%, confirming the effectiveness of personalization in safety alignment. However, not all context attributes contribute equally to safety enhancement. To address this, we develop RAISE - a training-free, two-stage agent framework that strategically acquires user-specific background. RAISE improves safety scores by up to 31.6% over six vanilla LLMs, while maintaining a low interaction cost of just 2.7 user queries on average. Our findings highlight the importance of selective information gathering in safety-critical domains and offer a practical solution for personalizing LLM responses without model retraining. This work establishes a foundation for safety research that adapts to individual user contexts rather than assuming a universal harm standard.

cs.CY↗
arXiv2025

Co-PatcheR: Collaborative Software Patching with Component(s)-specific Small Reasoning Models

Motivated by the success of general-purpose large language models (LLMs) in software patching, recent works started to train specialized patching models. Most works trained one model to handle the end-to-end patching pipeline (including issue localization, patch generation, and patch validation). However, it is hard for a small model to handle all tasks, as different sub-tasks have different workflows and require different expertise. As such, by using a 70 billion model, SOTA methods can only reach up to 41% resolved rate on SWE-bench-Verified. Motivated by the collaborative nature, we propose Co-PatcheR, the first collaborative patching system with small and specialized reasoning models for individual components. Our key technique novelties are the specific task designs and training recipes. First, we train a model for localization and patch generation. Our localization pinpoints the suspicious lines through a two-step procedure, and our generation combines patch generation and critique. We then propose a hybrid patch validation that includes two models for crafting issue-reproducing test cases with and without assertions and judging patch correctness, followed by a majority vote-based patch selection. Through extensive evaluation, we show that Co-PatcheR achieves 46% resolved rate on SWE-bench-Verified with only 3 x 14B models. This makes Co-PatcheR the best patcher with specialized models, requiring the least training resources and the smallest models. We conduct a comprehensive ablation study to validate our recipes, as well as our choice of training data number, model size, and testing-phase scaling strategy.

cs.AI↗
arXiv2025

NPHardEval4V: Dynamic Evaluation of Large Vision-Language Models with Effects of Vision

Large Vision-Language Models (LVLMs) have demonstrated impressive capabilities in multimodal understanding, yet their reasoning abilities remain underexplored. Existing benchmarks tend to focus on perception or text-based comprehension, offering limited insight into how well these models perform on structured, logic-driven tasks that require both visual and linguistic reasoning. To address this gap, we introduce NPHardEval4V, a multimodal benchmark suite grounded in four classical NP-hard problems: Knapsack, Set Cover, Traveling Salesperson, and Vertex Cover. Each task is presented through a combination of structured visual layouts and textual prompts, designed to assess the ability of LVLMs to perform combinatorial reasoning under visual-linguistic constraints. We evaluate a set of advanced open-source and closed-source vision-language models under a unified prompting and problem representation framework. This enables fair comparison across models and task types, while isolating key variables affecting performance. Our results show that while these models perform reasonably well on perception-based inputs, they struggle with global optimization, abstraction, and constraint satisfaction. No single model demonstrates consistent reasoning capability across all problem types, and common failure patterns reveal fundamental limitations in current architectures. By leveraging the structure and complexity of NP-hard problems, NPHardEval4V provides a scalable, interpretable, and challenging testbed for diagnosing reasoning behaviors in LVLMs. We hope this benchmark can support the community in building more robust, inference-capable multimodal systems. The benchmark dataset and code are available at https://github.com/lizhouf/NPHardEval4.

cs.CL↗
arXiv2025

PromptArmor: Simple yet Effective Prompt Injection Defenses

Despite their potential, recent research has demonstrated that LLM agents are vulnerable to prompt injection attacks, where malicious prompts are injected into the agent's input, causing it to perform an attacker-specified task rather than the intended task provided by the user. In this paper, we present PromptArmor, a simple yet effective defense against prompt injection attacks. Specifically, PromptArmor prompts an off-the-shelf LLM to detect and remove potential injected prompts from the input before the agent processes it. Our results show that PromptArmor can accurately identify and remove injected prompts. For example, using GPT-4o, GPT-4.1, or o4-mini, PromptArmor achieves both a false positive rate and a false negative rate below 1% on the AgentDojo benchmark. Moreover, after removing injected prompts with PromptArmor, the attack success rate drops to below 1%. We also demonstrate PromptArmor's effectiveness against adaptive attacks and explore different strategies for prompting an LLM. We recommend that PromptArmor be adopted as a standard baseline for evaluating new defenses against prompt injection attacks.

cs.CR↗
arXiv2025

SOPBench: Evaluating Language Agents at Following Standard Operating Procedures and Constraints

As language agents increasingly automate critical tasks, their ability to follow domain-specific standard operating procedures (SOPs), policies, and constraints when taking actions and making tool calls becomes essential yet remains underexplored. To address this gap, we develop an automated evaluation pipeline SOPBench with: (1) executable environments containing 167 tools/functions across seven customer service domains with service-specific SOPs and rule-based verifiers, (2) an automated test generation framework producing over 900 verified test cases, and (3) an automated evaluation framework to rigorously assess agent adherence from multiple dimensions. Our approach transforms each service-specific SOP code program into a directed graph of executable functions and requires agents to call these functions based on natural language SOP descriptions. The original code serves as oracle rule-based verifiers to assess compliance, reducing reliance on manual annotations and LLM-based evaluations. We evaluate 18 leading models, and results show the task is challenging even for top-tier models (like GPT-4o, Claude-3.7-Sonnet), with variances across domains. Reasoning models like o4-mini-high show superiority while other powerful models perform less effectively (pass rates of 30%-50%), and small models (7B, 8B) perform significantly worse. Additionally, language agents can be easily jailbroken to overlook SOPs and constraints. Code, data, and over 24k agent trajectories are released at https://github.com/Leezekun/SOPBench.

cs.CL↗
arXiv2025

MELON: Provable Defense Against Indirect Prompt Injection Attacks in AI Agents

Recent research has explored that LLM agents are vulnerable to indirect prompt injection (IPI) attacks, where malicious tasks embedded in tool-retrieved information can redirect the agent to take unauthorized actions. Existing defenses against IPI have significant limitations: either require essential model training resources, lack effectiveness against sophisticated attacks, or harm the normal utilities. We present MELON (Masked re-Execution and TooL comparisON), a novel IPI defense. Our approach builds on the observation that under a successful attack, the agent's next action becomes less dependent on user tasks and more on malicious tasks. Following this, we design MELON to detect attacks by re-executing the agent's trajectory with a masked user prompt modified through a masking function. We identify an attack if the actions generated in the original and masked executions are similar. We also include three key designs to reduce the potential false positives and false negatives. Extensive evaluation on the IPI benchmark AgentDojo demonstrates that MELON outperforms SOTA defenses in both attack prevention and utility preservation. Moreover, we show that combining MELON with a SOTA prompt augmentation defense (denoted as MELON-Aug) further improves its performance. We also conduct a detailed ablation study to validate our key designs. Code is available at https://github.com/kaijiezhu11/MELON.

cs.CR↗
arXiv2025

Achilles Heel of Distributed Multi-Agent Systems

Multi-agent system (MAS) has demonstrated exceptional capabilities in addressing complex challenges, largely due to the integration of multiple large language models (LLMs). However, the heterogeneity of LLMs, the scalability of quantities of LLMs, and local computational constraints pose significant challenges to hosting these models locally. To address these issues, we propose a new framework termed Distributed Multi-Agent System (DMAS). In DMAS, heterogeneous third-party agents function as service providers managed remotely by a central MAS server and each agent offers its services through API interfaces. However, the distributed nature of DMAS introduces several concerns about trustworthiness. In this paper, we study the Achilles heel of distributed multi-agent systems, identifying four critical trustworthiness challenges: free riding, susceptibility to malicious attacks, communication inefficiencies, and system instability. Extensive experiments across seven frameworks and four datasets reveal significant vulnerabilities of the DMAS. These attack strategies can lead to a performance degradation of up to 80% and attain a 100% success rate in executing free riding and malicious attacks. We envision our work will serve as a useful red-teaming tool for evaluating future multi-agent systems and spark further research on trustworthiness challenges in distributed multi-agent systems.

cs.MA↗
arXiv2024

PromptBench: A Unified Library for Evaluation of Large Language Models

The evaluation of large language models (LLMs) is crucial to assess their performance and mitigate potential security risks. In this paper, we introduce PromptBench, a unified library to evaluate LLMs. It consists of several key components that are easily used and extended by researchers: prompt construction, prompt engineering, dataset and model loading, adversarial prompt attack, dynamic evaluation protocols, and analysis tools. PromptBench is designed to be an open, general, and flexible codebase for research purposes that can facilitate original study in creating new benchmarks, deploying downstream applications, and designing new evaluation protocols. The code is available at: https://github.com/microsoft/promptbench and will be continuously supported.

cs.AI↗