Show HN: Visualizing 8k+ LLM papers with t-SNE (awesome-LLM-papers.github.io) (awesome-llm-papers.github.io)

A community-built interactive map (awesome-LLM-papers.github.io) plots over 8,000 LLM papers as dots in a t-SNE projection so similar papers sit near one another. Each point is derived from the paper’s content (title/abstract/metadata) and can be filtered with a search bar by title, authors, abstract or tags, enabling quick exploration of clusters, neighbors, and thematic pockets across the sprawling LLM literature. The site also invites contributors to correct metadata or add new papers, turning it into a living, crowd-curated literature index. For researchers and practitioners, this is a fast, visual way to survey the field, spot emergent topics, find related work, and identify underexplored niches without reading dozens of unrelated papers. Technically, the project scales text embeddings + t-SNE to thousands of documents, surfacing local structure and dense clusters; users should be aware t-SNE preserves local relationships but can distort global distances and is sensitive to hyperparameters and randomness. Complementing this map with alternative embeddings, UMAP, or clustering metrics can improve robustness. Overall, the tool is a pragmatic, low-friction aid for literature discovery, syntheses, and meta-analysis in the rapidly evolving LLM ecosystem.