P

paper-improvement-agent

Free

Upload a research paper PDF, get a peer review grounded in Semantic Scholar and OpenAlex, and edit it with natural-language commands.

Visit Website
ai-agentanthropiccitationsclaudecsl-jsonaiai assistantautomation

Overview

Paper Improvement Agent is a tool that uploads a research paper PDF, provides a peer review grounded in academic search (Semantic Scholar + OpenAlex), and allows improvement of the paper with natural-language editing commands without losing citations. It was built for the AnswerThis founding engineer assessment. The tool requires an ANTHROPIC_API_KEY for peer review and editing, while upload and citation parsing work without it. A SEMANTIC_SCHOLAR_API_KEY is optional for higher rate limits, and OpenAlex needs no key.

The system parses a research PDF into a structured document through a six-stage pipeline (extract → structure → locate references → segment → parse/resolve → link markers). Every reference is verified against OpenAlex/Semantic Scholar into CSL-JSON with a clickable source link, and every in-text marker is bound to its entry. Peer review streams reviewer-style findings, including relevant work the paper doesn't cite (year-filtered) and claim–citation checks judging each citing sentence against the cited work's real abstract. Editing by instruction runs a tool-use agent that proposes typed operations, shown as word-level diffs for approval. A deterministic validator rejects any edit that would lose a citation, cite a nonexistent entry, or add an unverified source.

The tool exports the paper as compilable LaTeX (verified with tectonic) with \cite keys and a citeproc/CSL-rendered bibliography, plus a BibTeX file and a structured Markdown version. It was tested end-to-end on arXiv 1706.03762 ('Attention Is All You Need'): 40/41 references verified with links, 102→71 markers after footnote/math de-noising, and the reviewer independently caught the paper's known byte-pair-encoding mis-citation. Known limitations include rejection of scanned PDFs (no text layer) and no OCR support. Future plans include deploying to Vercel with a storage swap, production path with Postgres and object storage, and claim checking against full texts.

Pricing

Free

Detailed plans have not been confirmed in our catalog. Check the official website for current limits and billing terms.

Visit Website

Prices and limits may change. Confirm the currency, billing period, seat minimum and usage caps on the official website.

Use Cases

  • Researchers use Paper Improvement Agent to upload a research paper PDF and receive a peer review grounded in academic search.
  • Authors use it to improve their paper with natural-language editing commands without losing citations.
  • Teams use it to verify references against OpenAlex and Semantic Scholar and bind in-text markers to entries.
  • Reviewers use it to check claims against cited works' abstracts.
  • Users use it to edit papers by instruction, such as making the introduction more concise or adding supporting citations.
  • Researchers use it to export papers as compilable LaTeX, BibTeX, and Markdown.

Features

Accepts uploaded research paper PDFs

Produces peer reviews of papers

Grounds reviews in scholarly sources

Edits papers through natural-language commands

Preserves citations during editing

Avoids invented source references

Reviews

No reviews yet. Be the first to share your experience!