Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | https://arxiv.org/abs/2601.06818 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866914245739806720 |
|---|---|
| author | Liu, Xuannan Yang, Xiao Li, Zekun Li, Peipei He, Ran |
| author_facet | Liu, Xuannan Yang, Xiao Li, Zekun Li, Peipei He, Ran |
| contents | As LLM-based agents operate over sequential multi-step reasoning, hallucinations arising at intermediate steps risk propagating along the trajectory, thus degrading overall reliability. Unlike hallucination detection in single-turn responses, diagnosing hallucinations in multi-step workflows requires identifying which step causes the initial divergence. To fill this gap, we propose a new research task, automated hallucination attribution of LLM-based agents, aiming to identify the step responsible for the hallucination and explain why. To support this task, we introduce AgentHallu, a comprehensive benchmark with: (1) 693 high-quality trajectories spanning 7 agent frameworks and 5 domains, (2) a hallucination taxonomy organized into 5 categories (Planning, Retrieval, Reasoning, Human-Interaction, and Tool-Use) and 14 sub-categories, and (3) multi-level annotations curated by humans, covering binary labels, hallucination-responsible steps, and causal explanations. We evaluate 13 leading models, and results show the task is challenging even for top-tier models (like GPT-5, Gemini-2.5-Pro). The best-performing model achieves only 41.1\% step localization accuracy, where tool-use hallucinations are the most challenging at just 11.6\%. We believe AgentHallu will catalyze future research into developing robust, transparent, and reliable agentic systems. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_06818 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | AgentHallu: Benchmarking Automated Hallucination Attribution of LLM-based Agents Liu, Xuannan Yang, Xiao Li, Zekun Li, Peipei He, Ran Computation and Language As LLM-based agents operate over sequential multi-step reasoning, hallucinations arising at intermediate steps risk propagating along the trajectory, thus degrading overall reliability. Unlike hallucination detection in single-turn responses, diagnosing hallucinations in multi-step workflows requires identifying which step causes the initial divergence. To fill this gap, we propose a new research task, automated hallucination attribution of LLM-based agents, aiming to identify the step responsible for the hallucination and explain why. To support this task, we introduce AgentHallu, a comprehensive benchmark with: (1) 693 high-quality trajectories spanning 7 agent frameworks and 5 domains, (2) a hallucination taxonomy organized into 5 categories (Planning, Retrieval, Reasoning, Human-Interaction, and Tool-Use) and 14 sub-categories, and (3) multi-level annotations curated by humans, covering binary labels, hallucination-responsible steps, and causal explanations. We evaluate 13 leading models, and results show the task is challenging even for top-tier models (like GPT-5, Gemini-2.5-Pro). The best-performing model achieves only 41.1\% step localization accuracy, where tool-use hallucinations are the most challenging at just 11.6\%. We believe AgentHallu will catalyze future research into developing robust, transparent, and reliable agentic systems. |
| title | AgentHallu: Benchmarking Automated Hallucination Attribution of LLM-based Agents |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2601.06818 |