Paper Pdf Two Pass Review

What This Skill Does

The Paper Pdf Two-Pass Review skill is a specialized agentic workflow designed for academic research, specifically tailored for CVPR-style papers. Unlike standard summarization tools, this skill implements a rigorous two-pass process. During the first pass, the agent systematically extracts verifiable facts, mapping every insight to specific page numbers, sections, figures, or tables within the source PDF. During the second pass, it generates a comprehensive, PhD-level critical review written in Chinese, strictly synthesized from the data collected in the first step. This structure effectively minimizes hallucinations and ensures that all academic discourse remains anchored to the provided text.

Installation

To integrate this skill into your OpenClaw environment, execute the following command in your terminal: clawhub install openclaw/skills/skills/gpiox/paper-pdf-two-pass-review

Use Cases

• Reproducibility Assessment: Evaluate if a research paper provides sufficient algorithmic detail and implementation parameters to recreate results.
• Deep Academic Critique: Obtain a detailed, logical evaluation of a paper's contributions, methodology, and experimental results compared against its stated goals.
• PhD-level Literature Review: Use this tool to process large volumes of technical documents into structured notes that highlight core innovation and architectural decisions.
• Thesis Research: Quickly determine if a specific paper aligns with your ongoing research questions by verifying core claims against the provided evidence tags.

Example Prompts

1. "I've uploaded the CVPR 2024 paper 'Adaptive Vision Transformers'. Please run a two-pass review and focus specifically on the ablation study section."
2. "Please perform a deep critique of the attached PDF. I need a 2500-word analysis covering the experimental limitations and potential scalability issues."
3. "Run a two-pass analysis on this paper. Ensure every claim regarding the model's accuracy improvement has a corresponding citation tag like [Evidence: p.X, Sec.Y]."

Tips & Limitations

• Data Grounding: The model is strictly constrained to the uploaded PDF. It will not incorporate external citations or training knowledge, making it ideal for objective analysis.
• Accuracy: The skill is designed for technical precision. If information is missing from the paper, the output will explicitly label it as 'Insufficient evidence'.
• Performance: This is a resource-intensive process due to the dual-pass validation, so allow the agent sufficient time to process long documents (e.g., papers exceeding 15 pages).