DEV Community

orville wang
orville wang

Posted on

How I Built a Tool to Organize AI Conference Papers by Topic (and Why It Matters)

I maintain paperlist.ai, an open tool for browsing AI conference papers organized by research topic.

How many ICLR papers are about RAG? What about agents or multimodal models? If you have ever tried to answer these questions by browsing a conference website, you know the pain. Thousands of papers, no topic grouping, and a PDF list that takes hours to skim.

I built paperlist.ai to fix this.

The Problem

Every year, AI conferences like ICLR, NeurIPS, and ICML publish thousands of accepted papers. Most researchers and developers only care about a tiny subset. But finding those relevant papers means:

  • Scrolling through alphabetical PDF lists on the conference website
  • Manually opening 50+ papers just to check which ones are relevant
  • Missing important work because it was buried in an unrelated section

The information is there. It is just not organized.

How paperlist.ai Works

paperlist.ai groups papers by research topic:

  • RAG: retrieval-augmented generation
  • Agents / Multi-Agent: autonomous AI agents
  • Multimodal: vision + language models
  • RL / Alignment: reinforcement learning and AI safety
  • Efficient LM: model compression and inference optimization

Each topic page lists relevant papers with their titles, authors, and links. No more opening 50 PDFs just to find the 3 that matter to your work.

The Tech Stack

The site is built with Next.js and uses OpenNomos for tracking community contributions like article shares and SEO submissions. The paper data comes from conference proceedings, organized through a mix of automated topic assignment and manual curation.

I wrote about this in more detail in a separate post, but the key insight is: organizing information is often more valuable than generating new content. An LLM can summarize a paper. But telling you which papers to read — that is a curation problem, not a generation problem.

What I Learned

  • Curation scales better than content generation: The bottleneck for researchers is not "I need more papers" but "I need to find relevant papers fast."
  • Community feedback drives topic selection: The most popular topics on paperlist.ai came from user requests, not my guesses.
  • Open tools win: Keeping it free and open means more researchers use it and more people contribute topic suggestions.

Try It

If you are catching up on ICLR 2026 papers, check out paperlist.ai. RAG-related papers are already organized, with more topics coming.

What tools do you use to stay on top of AI research? Would love to hear in the comments.

Top comments (0)