arXiv subjects

Shufan Jiang

Publications and source records attributed to Shufan Jiang.

11 recordsLinked to original sources

arXiv2026

SWE-Bench Pro Verified: A Reliable Benchmark for Software Engineering Agents

SWE-Bench Pro has emerged as a standard benchmark for evaluating software engineering agents on challenging repository-level tasks. However, our analysis work show that its evaluation is undermined by two sources of unreliability: \textbf{reward hacking}, enabled by leakage of gold solutions or hidden evaluation information, and \textbf{task quality issues}, including misleading problem statements and improperly scoped tests. These issues can inflate benchmark performance and obscure agents' true coding ability. We present \textbf{SWE-Bench Pro Verified}, a verified version of SWE-Bench Pro that addresses both problems. Our approach combines \textbf{anti-hacking} safeguards that eliminate major leakage channels without disrupting normal agent functionality, with \textbf{task refinement} that minimally corrects inconsistencies within flawed instances. Evaluations on SWE-Bench Pro Verified reveal that some models perform substantially worse than previously reported, suggesting that existing results on SWE-Bench Pro may overestimate real software engineering capability. SWE-Bench Pro Verified offers a more trustworthy benchmark for assessing software engineering agents.

cs.AI↗

arXiv2026

AgentCompass: A Unified Evaluation Infrastructure for Agent Capabilities

As Large Language Models (LLMs) evolve into autonomous agents, the need for unified evaluation infrastructure becomes critical. However, current evaluation pipelines remain highly fragmented and tightly coupled, hindering reproducibility and causing redundant engineering. To address this, we introduce AgentCompass, an open-source, lightweight, and extensible infrastructure for evaluating LLM-based agents. AgentCompass organizes the evaluation process around three independent components, namely Benchmark, Harness, and Environment, thereby enabling flexible configurations without requiring the reimplementation of complex execution logic. Furthermore, it features a fault-tolerant asynchronous runtime and comprehensive trajectory analysis tools to transparently diagnose nuanced failure modes like reward-hacking. Natively supporting over 20 benchmarks across five capability dimensions, AgentCompass provides the community with a scalable and reproducible infrastructure for advancing agent research.

cs.AI↗

arXiv2026

GBQA: A Game Benchmark for Evaluating LLMs as Quality Assurance Engineers

The autonomous discovery of bugs remains a significant challenge in modern software development. Compared to code generation, the complexity of dynamic runtime environments makes bug discovery considerably harder for large language models (LLMs). In this paper, we take game development as a representative domain and introduce the Game Benchmark for Quality Assurance (GBQA), a benchmark containing 30 games and 124 human-verified bugs across three difficulty levels, to evaluate whether LLMs can autonomously detect software bugs. The benchmark is constructed using a multi-agent system that develops games and injects bugs in a scalable manner, with human experts in the loop to ensure correctness. Moreover, we provide a baseline interactive agent equipped with a multi-round ReAct loop and a memory mechanism, enabling long-horizon exploration of game environments for bug detection across different LLMs. Extensive experiments on frontier LLMs demonstrate that autonomous bug discovery remains highly challenging: the best-performing model, Claude-4.6-Opus in thinking mode, identifies only 48.39% of the verified bugs. We believe GBQA provides an adequate testbed and evaluation criterion, and that further progress on it will help close the gap in autonomous software engineering.

cs.SE↗

arXiv2026

NERdME: a Named Entity Recognition Dataset for Indexing Research Artifacts in Code Repositories

Existing scholarly information extraction (SIE) datasets focus on scientific papers and overlook implementation-level details in code repositories. README files describe datasets, source code, and other implementation-level artifacts, however, their free-form Markdown offers little semantic structure, making automatic information extraction difficult. To address this gap, NERdME is introduced: 200 manually annotated README files with over 10,000 labeled spans and 10 entity types. Baseline results using large language models and fine-tuned transformers show clear differences between paperlevel and implementation-level entities, indicating the value of extending SIE benchmarks with entity types available in README files. A downstream entity-linking experiment was conducted to demonstrate that entities derived from READMEs can support artifact discovery and metadata integration.

cs.CL↗

arXiv2025

FURINA: A Fully Customizable Role-Playing Benchmark via Scalable Multi-Agent Collaboration Pipeline

As large language models (LLMs) advance in role-playing (RP) tasks, existing benchmarks quickly become obsolete due to their narrow scope, outdated interaction paradigms, and limited adaptability across diverse application scenarios. To address this gap, we introduce FURINA-Builder, a novel multi-agent collaboration pipeline that automatically constructs fully customizable RP benchmarks at any scale. It enables evaluation of arbitrary characters across diverse scenarios and prompt formats, as the first benchmark builder in RP area for adaptable assessment. FURINA-Builder simulates dialogues between a test character and other characters drawn from a well-constructed character-scene pool, while an LLM judge selects fine-grained evaluation dimensions and adjusts the test character's responses into final test utterances. Using this pipeline, we build FURINA-Bench, a new comprehensive role-playing benchmark featuring both established and synthesized test characters, each assessed with dimension-specific evaluation criteria. Human evaluation and preliminary separability analysis justify our pipeline and benchmark design. We conduct extensive evaluations of cutting-edge LLMs and find that o3 and DeepSeek-R1 achieve the best performance on English and Chinese RP tasks, respectively. Across all models, established characters consistently outperform synthesized ones, with reasoning capabilities further amplifying this disparity. Interestingly, we observe that model scale does not monotonically reduce hallucinations. More critically, for reasoning LLMs, we uncover a novel trade-off: reasoning improves RP performance but simultaneously increases RP hallucinations. This trade-off extends to a broader Pareto frontier between RP performance and reliability for all LLMs. These findings demonstrate the effectiveness of FURINA-Builder and the challenge posed by FURINA-Bench.

cs.CL↗

arXiv2025

NFDI4DS Shared Tasks for Scholarly Document Processing

Shared tasks are powerful tools for advancing research through community-based standardised evaluation. As such, they play a key role in promoting findable, accessible, interoperable, and reusable (FAIR), as well as transparent and reproducible research practices. This paper presents an updated overview of twelve shared tasks developed and hosted under the German National Research Data Infrastructure for Data Science and Artificial Intelligence (NFDI4DS) consortium, covering a diverse set of challenges in scholarly document processing. Hosted at leading venues, the tasks foster methodological innovations and contribute open-access datasets, models, and tools for the broader research community, which are integrated into the consortium's research data infrastructure.

cs.CL↗

arXiv2025

HAMLET: A Hierarchical and Adaptive Multi-Agent Framework for Live Embodied Theatrics

Creating an immersive and interactive theatrical experience is a long-term goal in the field of interactive narrative. The emergence of large language models (LLMs) provides a new path to achieve this goal. However, existing drama generation methods often produce LLMs that lack initiative and cannot interact with the physical scene, while typically requiring detailed input that diminishes the immersion of live performance. To address these challenges, we propose HAMLET, a hierarchical adaptive multi-agent framework focused on drama creation and real-time online performance. Given a simple topic, the framework initially generates a narrative blueprint to guide the subsequent improvisational performance. During online performance, each actor is equipped with an adaptive reasoning module that enables decision-making based on their personas, memories, goals during complex group chat scenarios. Beyond dialogue, actor agents engage in embodied interactions by changing the state of scene props through actions such as opening a letter or picking up a weapon, which are broadcast to update the global environmental context. To objectively assess the quality of live embodied theatrics, we establish a comprehensive evaluation method and introduce HAMLETJudge, a specialized critic model for automated evaluation. Experimental results demonstrate that HAMLET excels in creating expressive, coherent, and physically interactive theatrical experiences in an autonomous manner.

cs.AI↗

arXiv2024

LINKs: Large Language Model Integrated Management for 6G Empowered Digital Twin NetworKs

In the rapidly evolving landscape of digital twins (DT) and 6G networks, the integration of large language models (LLMs) presents a novel approach to network management. This paper explores the application of LLMs in managing 6G-empowered DT networks, with a focus on optimizing data retrieval and communication efficiency in smart city scenarios. The proposed framework leverages LLMs for intelligent DT problem analysis and radio resource management (RRM) in fully autonomous way without any manual intervention. Our proposed framework -- LINKs, builds up a lazy loading strategy which can minimize transmission delay by selectively retrieving the relevant data. Based on the data retrieval plan, LLMs transform the retrieval task into an numerical optimization problem and utilizing solvers to build an optimal RRM, ensuring efficient communication across the network. Simulation results demonstrate the performance improvements in data planning and network management, highlighting the potential of LLMs to enhance the integration of DT and 6G technologies.

cs.NI↗

arXiv2023

Extracting Definienda in Mathematical Scholarly Articles with Transformers

We consider automatically identifying the defined term within a mathematical definition from the text of an academic article. Inspired by the development of transformer-based natural language processing applications, we pose the problem as (a) a token-level classification task using fine-tuned pre-trained transformers; and (b) a question-answering task using a generalist large language model (GPT). We also propose a rule-based approach to build a labeled dataset from the LATEX source of papers. Experimental results show that it is possible to reach high levels of precision and recall using either recent (and expensive) GPT 4 or simpler pre-trained models fine-tuned on our task.

cs.AI↗

arXiv2023

Named Entity Recognition for Monitoring Plant Health Threats in Tweets: a ChouBERT Approach

An important application scenario of precision agriculture is detecting and measuring crop health threats using sensors and data analysis techniques. However, the textual data are still under-explored among the existing solutions due to the lack of labelled data and fine-grained semantic resources. Recent research suggests that the increasing connectivity of farmers and the emergence of online farming communities make social media like Twitter a participatory platform for detecting unfamiliar plant health events if we can extract essential information from unstructured textual data. ChouBERT is a French pre-trained language model that can identify Tweets concerning observations of plant health issues with generalizability on unseen natural hazards. This paper tackles the lack of labelled data by further studying ChouBERT's know-how on token-level annotation tasks over small labeled sets.

cs.CL↗

arXiv2021

Fine-tuning BERT-based models for Plant Health Bulletin Classification

In the era of digitization, different actors in agriculture produce numerous data. Such data contains already latent historical knowledge in the domain. This knowledge enables us to precisely study natural hazards within global or local aspects, and then improve the risk prevention tasks and augment the yield, which helps to tackle the challenge of growing population and changing alimentary habits. In particular, French Plants Health Bulletins (BSV, for its name in French Bulletin de Sant{\'e} du V{\'e}g{\'e}tal) give information about the development stages of phytosanitary risks in agricultural production. However, they are written in natural language, thus, machines and human cannot exploit them as efficiently as it could be. Natural language processing (NLP) technologies aim to automatically process and analyze large amounts of natural language data. Since the 2010s, with the increases in computational power and parallelization, representation learning and deep learning methods became widespread in NLP. Recent advancements Bidirectional Encoder Representations from Transformers (BERT) inspire us to rethink of knowledge representation and natural language understanding in plant health management domain. The goal in this work is to propose a BERT-based approach to automatically classify the BSV to make their data easily indexable. We sampled 200 BSV to finetune the pretrained BERT language models and classify them as pest or/and disease and we show preliminary results.

cs.CL↗