Summarize Paper logo SummarizePaper.com
Premium

Who works on it?

Describe a research topic: we find the people and institutions who sign the matching papers of the last years, in journals and on preprint servers, how they sign them, and the approaches these papers take.

How to map researchers and approaches? Illustrated guide

Looking for an internship or PhD? Explore research placement leads.

A real example: what you get

Example in Astrophysics Computer science and AI Biology and medicine

15 institutions, 30 people and 118 papers on “RAG for Reducing LLM Hallucinations”.

  • The landscape of the topic
  • 4 approaches, each named and described
  • 30 people who sign first, last or as corresponding author
  • 13 countries
  • Exports, a collection of the papers, a monthly follow-up
The landscape

The retrieval-quality approach focuses on improving evidence selection and grounding within RAG, whereas the architecture approach examines the broader design, efficiency, and knowledge-integration choices of RAG systems. The combined-validation approach treats retrieval as one component of a wider hallucination-detection and mitigation toolkit, overlapping with RAG-specific work on confidence and verification. The task-specific approach brings these concerns to particular outputs, domains, …

An approach

These papers review and develop RAG-specific hallucination mitigation through retrieval comparisons, retrieval optimization, confidence-aware retrieval, self-consistency verification, and handling evidence from multiple sources.

Retrieval Quality and Grounding · 62 papers

The first results

  1. Tsinghua University CN · 3 papers
  2. Amazon (United States) US · 4 papers

13 more results hidden

See a complete example, explained step by step →

Create a free account: your first search with this tool is free, its first results shown in full. Continue with Google Create a free account Log in

How it works

  1. We write several searches from your topic and run them on OpenAlex, which knows the papers of journals and preprint servers with the institutions of their authors. You see the searches and can change them.
  2. An AI reads the titles and abstracts and keeps the papers that address the topic. A preprint and its published version count once.
  3. For each person: the papers they sign first, last or as corresponding author, apart from alphabetical author lists and the long lists of large collaborations; one person per ORCID. Each institution counts for its share of each paper.
  4. The papers are grouped by approach (method, data, instrument or theory), so you see who does what.
  5. With your ORCID iD, you and your co-authors of the last five years are left out when you look for referees, collaborators or competing teams.