---
name: citation-neighborhood-walker
description: "Given a seed paper, build a 2-hop citation graph and rank neighbors by relevance to a given thesis — a literature survey that's actually grounded."
title: Citation Neighborhood Walker
category: search-retrieval
difficulty: advanced
author: admin
icon: 🕸️
input: structured-data
output: structured-json
phase: enhance
domain: research
tags: citation-graph,literature-survey,semantic-search,graph-traversal,llm-ranking,openalexapi,bibliographic-metadata,relevance-scoring,research-discovery,knowledge-graph,paper-recommendation,doi-resolution
best_for:
  - Literature survey scoping from a single anchor paper
  - Domain-specific paper discovery without keyword drift
  - Thesis validation via citation neighborhood mapping
  - Research gap identification in 2-hop citation clusters
---

## Description

Input: a seed DOI and a thesis statement. Output: a ranked list of papers in the 2-hop citation neighborhood (papers the seed cites, plus papers those cite), scored by LLM-judged relevance to the thesis. Designed to replace the dice-roll of keyword search when you already have one good anchor paper.

## Why it works

A well-chosen seed paper is a better search oracle than keywords because its bibliography encodes what the author thought was relevant. Walking the citation graph inherits that expertise. Limiting to 2 hops keeps the result set tractable; thesis-scoring keeps it on-topic.

## How it works

1. Fetch the seed's bibliography via OpenAlex / Semantic Scholar / Crossref. 2. For each cited paper, fetch its bibliography (2nd hop). 3. Deduplicate the combined set and pull abstracts. 4. For each unique paper, prompt the LLM with (thesis, paper abstract) and a 1-5 relevance score rubric. 5. Return the top N papers sorted by relevance with a one-line rationale each, and a graph file (DOT or JSON) of the citation neighborhood for visualization.