arXiv subjects

Yifan Zeng

Publications and source records attributed to Yifan Zeng.

At least 19 recordsLinked to original sources

arXiv2026

Who&When Pro: Can LLMs Really Attribute Failures in AI Agents?

Automated failure attribution uses LLMs to identify where and why agentic systems fail. As agents become more capable, their failures become subtler, making automated attribution increasingly important. We introduce Who&When Pro, a large-scale benchmark for automated failure attribution in agentic systems. Using a strictly controlled pipeline that injects a failure only after exactly replaying a successful prefix, we construct 12,326 failed trajectories with golden labels across 3 modalities and 26 benchmarks covering various scenarios. Beyond benchmarking, we conduct extensive experiments and analyses, revealing systematic patterns in how models attribute failures across modalities, protocols, and model families, and providing empirical guidance for future automated failure attribution systems.

cs.AI↗

arXiv2026

When Does Multi-Agent RL Improve LLM Workflows? Workflow, Scale, and Policy-Sharing Tradeoffs

Multi-agent LLM workflows route inference through specialized roles to lift end-task accuracy, but jointly training those roles with reinforcement learning is unstable in ways that are poorly understood. We study when end-to-end RL training of multi-agent LLM workflows improves over their base models, comparing Shared-Policy training, where all roles update one policy, with Isolated-Policy training, where each role has its own parameters. Our experimental matrix spans Eval-Opt, Voting, and Orch-Workers workflows, math and code tasks, and three model scales (0.6B, 1.7B, 4B). We find that multi-agent RL usually improves over base models, but gains depend jointly on workflow, task, and scale, not on policy sharing alone. Isolated-Policy tends to reach higher peak accuracy yet more often falls off a terminal accuracy cliff, while Shared-Policy training does not eliminate failure; it redistributes failure into qualitatively different patterns. We then explain the strongest of these patterns through role-level gradient dynamics induced by workflow topology and policy routing: under Isolated-Policy, parallel same-role agents on shared prompts amplify per-role gradients and drive terminal degradation in Voting and Orch-Workers workflows; under Shared-Policy, asymmetric per-step gradient mass causes the shared policy to be captured by the dominant role, producing different failure signatures by task and workflow. Together, the empirical map and its underlying mechanisms show that policy sharing routes training pressure through different channels rather than offering uniform stability, making it a design choice with workflow- and task-conditional tradeoffs.

cs.AI↗

arXiv2025

WebGraphEval: Multi-Turn Trajectory Evaluation for Web Agents using Graph Representation

Current evaluation of web agents largely reduces to binary success metrics or conformity to a single reference trajectory, ignoring the structural diversity present in benchmark datasets. We present WebGraphEval, a framework that abstracts trajectories from multiple agents into a unified, weighted action graph. This representation is directly compatible with benchmarks such as WebArena, leveraging leaderboard runs and newly collected trajectories without modifying environments. The framework canonically encodes actions, merges recurring behaviors, and applies structural analyses including reward propagation and success-weighted edge statistics. Evaluations across thousands of trajectories from six web agents show that the graph abstraction captures cross-model regularities, highlights redundancy and inefficiency, and identifies critical decision points overlooked by outcome-based metrics. By framing web interaction as graph-structured data, WebGraphEval establishes a general methodology for multi-path, cross-agent, and efficiency-aware evaluation of web agents.

cs.AI↗

arXiv2025

The Ranking Blind Spot: Decision Hijacking in LLM-based Text Ranking

Large Language Models (LLMs) have demonstrated strong performance in information retrieval tasks like passage ranking. Our research examines how instruction-following capabilities in LLMs interact with multi-document comparison tasks, identifying what we term the "Ranking Blind Spot", a characteristic of LLM decision processes during comparative evaluation. We analyze how this ranking blind spot affects LLM evaluation systems through two approaches: Decision Objective Hijacking, which alters the evaluation goal in pairwise ranking systems, and Decision Criteria Hijacking, which modifies relevance standards across ranking schemes. These approaches demonstrate how content providers could potentially influence LLM-based ranking systems to affect document positioning. These attacks aim to force the LLM ranker to prefer a specific passage and rank it at the top. Malicious content providers can exploit this weakness, which helps them gain additional exposure by attacking the ranker. In our experiment, We empirically show that the proposed attacks are effective in various LLMs and can be generalized to multiple ranking schemes. We apply these attack to realistic examples to show their effectiveness. We also found stronger LLMs are more vulnerable to these attacks. Our code is available at: https://github.com/blindspotorg/RankingBlindSpot

cs.IR↗

arXiv2025

IteraOptiRacing: A Unified Planning-Control Framework for Real-time Autonomous Racing for Iterative Optimal Performance

This paper presents a unified planning-control strategy for competing with other racing cars called IteraOptiRacing in autonomous racing environments. This unified strategy is proposed based on Iterative Linear Quadratic Regulator for Iterative Tasks (i2LQR), which can improve lap time performance in the presence of surrounding racing obstacles. By iteratively using the ego car's historical data, both obstacle avoidance for multiple moving cars and time cost optimization are considered in this unified strategy, resulting in collision-free and time-optimal generated trajectories. The algorithm's constant low computation burden and suitability for parallel computing enable real-time operation in competitive racing scenarios. To validate its performance, simulations in a high-fidelity simulator are conducted with multiple randomly generated dynamic agents on the track. Results show that the proposed strategy outperforms existing methods across all randomly generated autonomous racing scenarios, enabling enhanced maneuvering for the ego racing car.

cs.RO↗

arXiv2025

$\text{C}^{2}\text{BNVAE}$: Dual-Conditional Deep Generation of Network Traffic Data for Network Intrusion Detection System Balancing

Network Intrusion Detection Systems (NIDS) face challenges due to class imbalance, affecting their ability to detect novel and rare attacks. This paper proposes a Dual-Conditional Batch Normalization Variational Autoencoder ($\text{C}^{2}\text{BNVAE}$) for generating balanced and labeled network traffic data. $\text{C}^{2}\text{BNVAE}$ improves the model's adaptability to different data categories and generates realistic category-specific data by incorporating Conditional Batch Normalization (CBN) into the Conditional Variational Autoencoder (CVAE). Experiments on the NSL-KDD dataset show the potential of $\text{C}^{2}\text{BNVAE}$ in addressing imbalance and improving NIDS performance with lower computational overhead compared to some baselines.

cs.CR↗

arXiv2025

Structuralist Approach to AI Literary Criticism: Leveraging Greimas Semiotic Square for Large Language Models

Large Language Models (LLMs) excel in understanding and generating text but struggle with providing professional literary criticism for works with profound thoughts and complex narratives. This paper proposes GLASS (Greimas Literary Analysis via Semiotic Square), a structured analytical framework based on Greimas Semiotic Square (GSS), to enhance LLMs' ability to conduct in-depth literary analysis. GLASS facilitates the rapid dissection of narrative structures and deep meanings in narrative works. We propose the first dataset for GSS-based literary criticism, featuring detailed analyses of 48 works. Then we propose quantitative metrics for GSS-based literary criticism using the LLM-as-a-judge paradigm. Our framework's results, compared with expert criticism across multiple works and LLMs, show high performance. Finally, we applied GLASS to 39 classic works, producing original and high-quality analyses that address existing research gaps. This research provides an AI-based tool for literary research and education, offering insights into the cognitive mechanisms underlying literary engagement.

cs.CL↗

arXiv2025

SimpleDoc: Multi-Modal Document Understanding with Dual-Cue Page Retrieval and Iterative Refinement

Document Visual Question Answering (DocVQA) is a practical yet challenging task, which is to ask questions based on documents while referring to multiple pages and different modalities of information, e.g, images and tables. To handle multi-modality, recent methods follow a similar Retrieval Augmented Generation (RAG) pipeline, but utilize Visual Language Models (VLMs) based embedding model to embed and retrieve relevant pages as images, and generate answers with VLMs that can accept an image as input. In this paper, we introduce SimpleDoc, a lightweight yet powerful retrieval - augmented framework for DocVQA. It boosts evidence page gathering by first retrieving candidates through embedding similarity and then filtering and re-ranking these candidates based on page summaries. A single VLM-based reasoner agent repeatedly invokes this dual-cue retriever, iteratively pulling fresh pages into a working memory until the question is confidently answered. SimpleDoc outperforms previous baselines by 3.2% on average on 4 DocVQA datasets with much fewer pages retrieved. Our code is available at https://github.com/ag2ai/SimpleDoc.

cs.CV↗

arXiv2025

QuantMCP: Grounding Large Language Models in Verifiable Financial Reality

Large Language Models (LLMs) hold immense promise for revolutionizing financial analysis and decision-making, yet their direct application is often hampered by issues of data hallucination and lack of access to real-time, verifiable financial information. This paper introduces QuantMCP, a novel framework designed to rigorously ground LLMs in financial reality. By leveraging the Model Context Protocol (MCP) for standardized and secure tool invocation, QuantMCP enables LLMs to accurately interface with a diverse array of Python-accessible financial data APIs (e.g., Wind, yfinance). Users can interact via natural language to precisely retrieve up-to-date financial data, thereby overcoming LLM's inherent limitations in factual data recall. More critically, once furnished with this verified, structured data, the LLM's analytical capabilities are unlocked, empowering it to perform sophisticated data interpretation, generate insights, and ultimately support more informed financial decision-making processes. QuantMCP provides a robust, extensible, and secure bridge between conversational AI and the complex world of financial data, aiming to enhance both the reliability and the analytical depth of LLM applications in finance.

cs.CE↗

arXiv2025

CSAGC-IDS: A Dual-Module Deep Learning Network Intrusion Detection Model for Complex and Imbalanced Data

As computer networks proliferate, the gravity of network intrusions has escalated, emphasizing the criticality of network intrusion detection systems for safeguarding security. While deep learning models have exhibited promising results in intrusion detection, they face challenges in managing high-dimensional, complex traffic patterns and imbalanced data categories. This paper presents CSAGC-IDS, a network intrusion detection model based on deep learning techniques. CSAGC-IDS integrates SC-CGAN, a self-attention-enhanced convolutional conditional generative adversarial network that generates high-quality data to mitigate class imbalance. Furthermore, CSAGC-IDS integrates CSCA-CNN, a convolutional neural network enhanced through cost sensitive learning and channel attention mechanism, to extract features from complex traffic data for precise detection. Experiments conducted on the NSL-KDD dataset. CSAGC-IDS achieves an accuracy of 84.55% and an F1-score of 84.52% in five-class classification task, and an accuracy of 91.09% and an F1 score of 92.04% in binary classification task.Furthermore, this paper provides an interpretability analysis of the proposed model, using SHAP and LIME to explain the decision-making mechanisms of the model.

cs.CR↗

arXiv2025

A Contrastive Federated Semi-Supervised Learning Intrusion Detection Framework for Internet of Robotic Things

In intelligent industry, autonomous driving and other environments, the Internet of Things (IoT) highly integrated with robotic to form the Internet of Robotic Things (IoRT). However, network intrusion to IoRT can lead to data leakage, service interruption in IoRT and even physical damage by controlling robots or vehicles. This paper proposes a Contrastive Federated Semi-Supervised Learning Network Intrusion Detection framework (CFedSSL-NID) for IoRT intrusion detection and defense, to address the practical scenario of IoRT where robots don't possess labeled data locally and the requirement for data privacy preserving. CFedSSL-NID integrates randomly weak and strong augmentation, latent contrastive learning, and EMA update to integrate supervised signals, thereby enhancing performance and robustness on robots' local unlabeled data. Extensive experiments demonstrate that CFedSSL-NID outperforms existing federated semi-supervised and fully supervised methods on benchmark dataset and has lower resource requirements.

cs.CR↗

arXiv2025

Quantifying Risk Propensities of Large Language Models: Ethical Focus and Bias Detection through Role-Play

As Large Language Models (LLMs) become more prevalent, concerns about their safety, ethics, and potential biases have risen. Systematically evaluating LLMs' risk decision-making tendencies and attitudes, particularly in the ethical domain, has become crucial. This study innovatively applies the Domain-Specific Risk-Taking (DOSPERT) scale from cognitive science to LLMs and proposes a novel Ethical Decision-Making Risk Attitude Scale (EDRAS) to assess LLMs' ethical risk attitudes in depth. We further propose a novel approach integrating risk scales and role-playing to quantitatively evaluate systematic biases in LLMs. Through systematic evaluation and analysis of multiple mainstream LLMs, we assessed the "risk personalities" of LLMs across multiple domains, with a particular focus on the ethical domain, and revealed and quantified LLMs' systematic biases towards different groups. This research helps understand LLMs' risk decision-making and ensure their safe and reliable application. Our approach provides a tool for identifying and mitigating biases, contributing to fairer and more trustworthy AI systems. The code and data are available.

cs.CY↗

arXiv2025

Divide, Optimize, Merge: Fine-Grained LLM Agent Optimization at Scale

LLM-based optimization has shown remarkable potential in enhancing agentic systems. However, the conventional approach of prompting LLM optimizer with the whole training trajectories on training dataset in a single pass becomes untenable as datasets grow, leading to context window overflow and degraded pattern recognition. To address these challenges, we propose Fine-Grained Optimization (FGO), a scalable framework that divides large optimization tasks into manageable subsets, performs targeted optimizations, and systematically combines optimized components through progressive merging. Evaluation across ALFWorld, LogisticsQA, and GAIA benchmarks demonstrate that FGO outperforms existing approaches by 1.6-8.6% while reducing average prompt token consumption by 56.3%. Our framework provides a practical solution for scaling up LLM-based optimization of increasingly sophisticated agent systems. Further analysis demonstrates that FGO achieves the most consistent performance gain in all training dataset sizes, showcasing its scalability and efficiency.

cs.CL↗

arXiv2025

A Common Pitfall of Margin-based Language Model Alignment: Gradient Entanglement

Reinforcement Learning from Human Feedback (RLHF) has become the predominant approach for language model (LM) alignment. At its core, RLHF uses a margin-based loss for preference optimization, specifying ideal LM behavior only by the difference between preferred and dispreferred responses. In this paper, we identify a common pitfall of margin-based methods -- the under-specification of ideal LM behavior on preferred and dispreferred responses individually, which leads to two unintended consequences as the margin increases: (1) The probability of dispreferred (e.g., unsafe) responses may increase, resulting in potential safety alignment failures. (2) The probability of preferred responses may decrease, even when those responses are ideal. We demystify the reasons behind these problematic behaviors: margin-based losses couple the change in the preferred probability to the gradient of the dispreferred one, and vice versa, often preventing the preferred probability from increasing while the dispreferred one decreases, and thus causing a synchronized increase or decrease in both probabilities. We term this effect, inherent in margin-based objectives, gradient entanglement. Formally, we derive conditions for general margin-based alignment objectives under which gradient entanglement becomes concerning: the inner product of the gradients of preferred and dispreferred log-probabilities is large relative to the individual gradient norms. We theoretically investigate why such inner products can be large when aligning language models and empirically validate our findings. Empirical implications of our framework extend to explaining important differences in the training dynamics of various preference optimization algorithms, and suggesting potential algorithm designs to mitigate the under-specification issue of margin-based methods and thereby improving language model alignment.

cs.LG↗

arXiv2024

Memory-Augmented Agent Training for Business Document Understanding

Traditional enterprises face significant challenges in processing business documents, where tasks like extracting transport references from invoices remain largely manual despite their crucial role in logistics operations. While Large Language Models offer potential automation, their direct application to specialized business domains often yields unsatisfactory results. We introduce Matrix (Memory-Augmented agent Training through Reasoning and Iterative eXploration), a novel paradigm that enables LLM agents to progressively build domain expertise through experience-driven memory refinement and iterative learning. To validate this approach, we collaborate with one of the world's largest logistics companies to create a dataset of Universal Business Language format invoice documents, focusing on the task of transport reference extraction. Experiments demonstrate that Matrix outperforms prompting a single LLM by 30.3%, vanilla LLM agent by 35.2%. We further analyze the metrics of the optimized systems and observe that the agent system requires less API calls, fewer costs and can analyze longer documents on average. Our methods establish a new approach to transform general-purpose LLMs into specialized business tools through systematic memory enhancement in document processing tasks.

cs.CL↗

arXiv2024

LLM-RankFusion: Mitigating Intrinsic Inconsistency in LLM-based Ranking

Ranking passages by prompting a large language model (LLM) can achieve promising performance in modern information retrieval (IR) systems. A common approach to sort the ranking list is by prompting LLMs for a pairwise or setwise comparison which often relies on sorting algorithms. However, sorting-based methods require consistent comparisons to correctly sort the passages, which we show that LLMs often violate. We identify two kinds of intrinsic inconsistency in LLM-based pairwise comparisons: order inconsistency which leads to conflicting results when switching the passage order, and transitive inconsistency which leads to non-transitive triads among all preference pairs. Our study of these inconsistencies is relevant for understanding and improving the stability of any ranking scheme based on relative preferences. In this paper, we propose LLM-RankFusion, an LLM-based ranking framework that mitigates these inconsistencies and produces a robust ranking list. LLM-RankFusion mitigates order inconsistency using in-context learning (ICL) to demonstrate order-agnostic comparisons and calibration to estimate the underlying preference probability between two passages. We then address transitive inconsistency by aggregating the ranking results from multiple rankers. In our experiments, we empirically show that LLM-RankFusion can significantly reduce inconsistent comparison results, improving the ranking quality by making the final ranking list more robust. Our code is available at \href{https://github.com/XHMY/LLM-RankFusion}{https://github.com/XHMY/LLM-RankFusion}

cs.IR↗

arXiv2024

HistoLens: An LLM-Powered Framework for Multi-Layered Analysis of Historical Texts -- A Case Application of Yantie Lun

This paper proposes HistoLens, a multi-layered analysis framework for historical texts based on Large Language Models (LLMs). Using the important Western Han dynasty text "Yantie Lun" as a case study, we demonstrate the framework's potential applications in historical research and education. HistoLens integrates NLP technology (especially LLMs), including named entity recognition, knowledge graph construction, and geographic information visualization. The paper showcases how HistoLens explores Western Han culture in "Yantie Lun" through multi-dimensional, visual, and quantitative methods, focusing particularly on the influence of Confucian and Legalist thoughts on political, economic, military, and ethnic. We also demonstrate how HistoLens constructs a machine teaching scenario using LLMs for explainable analysis, based on a dataset of Confucian and Legalist ideas extracted with LLM assistance. This approach offers novel and diverse perspectives for studying historical texts like "Yantie Lun" and provides new auxiliary tools for history education. The framework aims to equip historians and learners with LLM-assisted tools to facilitate in-depth, multi-layered analysis of historical texts and foster innovation in historical education.

cs.CL↗

arXiv2024

AutoDefense: Multi-Agent LLM Defense against Jailbreak Attacks

Despite extensive pre-training in moral alignment to prevent generating harmful information, large language models (LLMs) remain vulnerable to jailbreak attacks. In this paper, we propose AutoDefense, a multi-agent defense framework that filters harmful responses from LLMs. With the response-filtering mechanism, our framework is robust against different jailbreak attack prompts, and can be used to defend different victim models. AutoDefense assigns different roles to LLM agents and employs them to complete the defense task collaboratively. The division in tasks enhances the overall instruction-following of LLMs and enables the integration of other defense components as tools. With AutoDefense, small open-source LMs can serve as agents and defend larger models against jailbreak attacks. Our experiments show that AutoDefense can effectively defense against different jailbreak attacks, while maintaining the performance at normal user request. For example, we reduce the attack success rate on GPT-3.5 from 55.74% to 7.95% using LLaMA-2-13b with a 3-agent system. Our code and data are publicly available at https://github.com/XHMY/AutoDefense.

cs.LG↗