# Awesome Loop Models

Awesome Loop Models is a curated, interactively browsable list of papers and long-form technical posts about loop models, architectures that reuse a shared learned layer or block within a single forward pass.

Awesome Loop Models catalogs papers and selected technical blog posts on loop models, defined strictly as architectures where, within a single forward pass, a shared learned internal layer, block, module or operator is reused. The main catalog deliberately excludes broader iteration patterns that do not directly connect to that definition, such as agent loops, repeated full-model calls, external solver rounds, energy-based models or plain sequence-time recurrence, though a small number of curator-approved exceptions are kept and clearly labeled as adjacent work.

The project is organized around three flat paper categories in its public browsing layer, and an interactive browser lets readers search, filter and explore entries with links to arXiv, code and OpenReview, including a Release Intelligence view of publication trends and per-release dossiers covering authors, summaries, tags and metrics. Because the browser loads a papers.json file, previewing it locally requires running a small Python HTTP server rather than opening the page directly from disk.

Website: https://huskydoge.github.io/Awesome-Loop-Models/index.html