Organized 900k research papers on AI in a queryable format (old.reddit.com)

A new project has organized roughly 900,000 AI-related research papers into a single, queryable corpus designed for fast discovery and programmatic use. The resource aggregates metadata (titles, authors, venues, years), extracted or linked full text where available, and search interfaces that support both keyword and semantic queries—making it practical to run targeted literature searches, build citation maps, or feed documents into retrieval-augmented systems. The dataset is intended to lower the bar for systematic literature reviews, trend analyses, and tooling that depends on reliable access to large swaths of AI research. For the AI/ML community this matters because scale and structure unlock new workflows: you can quickly surface prior work, generate comprehensive benchmarks of approaches over time, train or evaluate retrieval models and embeddings on a domain-specific corpus, and automate parts of meta-research (e.g., citation trend detection, reproducibility audits). Important technical considerations include deduplication and normalization of records, quality of PDF-to-text extraction, handling of embargoed or paywalled content, licensing constraints, and the need for well-documented APIs and schemas. If maintained and responsibly licensed, a 900k-paper queryable corpus becomes a powerful infrastructural asset for research, tooling, and reproducible science across AI.