Problem
Research questions and friction points this paper is trying to address.
AI-written papers
evaluation framework
hallucination
presentation quality
reliability
Paper Reconstruction Evaluation: Evaluating Presentation and Hallucination in AI-written Papers
🤖 AI Summary
Existing approaches lack a systematic evaluation of the quality and risks associated with AI-generated academic papers. This work proposes PaperRecon, a novel framework that decouples assessment into two orthogonal dimensions: presentation quality and hallucination. Presentation quality is evaluated using a rule-based scoring rubric, while hallucinations are detected through an agent-based mechanism that traces claims back to their source in the original paper. To facilitate empirical analysis, we introduce PaperWrite-Bench, a benchmark comprising 51 top-tier conference papers. Our evaluation reveals a notable trade-off: ClaudeCode achieves higher presentation quality but exhibits over 10 hallucinations per paper on average, whereas Codex generates fewer hallucinations at the cost of lower presentation quality.
📝 Abstract
This paper introduces the first systematic evaluation framework for quantifying the quality and risks of papers written by modern coding agents. While AI-driven paper writing has become a growing concern, rigorous evaluation of the quality and potential risks of AI-written papers remains limited, and a unified understanding of their reliability is still lacking. We introduce Paper Reconstruction Evaluation (PaperRecon), an evaluation framework in which an overview (overview.md) is created from an existing paper, after which an agent generates a full paper based on the overview and minimal additional resources, and the result is subsequently compared against the original paper. PaperRecon disentangles the evaluation of the AI-written papers into two orthogonal dimensions, Presentation and Hallucination, where Presentation is evaluated using a rubric and Hallucination is assessed via agentic evaluation grounded in the original paper source. For evaluation, we introduce PaperWrite-Bench, a benchmark of 51 papers from top-tier venues across diverse domains published after 2025. Our experiments reveal a clear trade-off: while both ClaudeCode and Codex improve with model advances, ClaudeCode achieves higher presentation quality at the cost of more than 10 hallucinations per paper on average, whereas Codex produces fewer hallucinations but lower presentation quality. This work takes a first step toward establishing evaluation frameworks for AI-driven paper writing and improving the understanding of its risks within the research community.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Paper Reconstruction Evaluation
AI-written papers
Hallucination assessment
Presentation quality
PaperWrite-Bench
🔎 Similar Papers
No similar papers found.
