Paper Digest

Automated paper monitoring searches Hugging Face Papers and arXiv for tracked interests, deduplicates prior outputs, and generates targeted digests. This recurring workflow replaces manual scanning by filtering recent matches using upvotes, recency, and keyword relevance.

Deduplicating research papers in a weekly literature briefing requires logging prior cycle outputs and checking against them. The workflow avoids re-sending papers already delivered in the last 7 days and records the verdict for the next cycle.

Topic-based discovery digests can be generated from a single research keyword provided as a variable. If the variable is empty, the system parses topic configurations from MEMORY.md to find matching papers and summarize them.

Summarizing arXiv papers into a two-sentence format produces a strict “what's new” and “so what” pair tied to your research interests. It applies banned generic phrasing constraints to keep the output specific and decision-ready.

If no tracked topics are available in MEMORY.md, the paper digest workflow aborts the cycle. It enforces skip-gate ranking and constraints, ensuring no digest or cycle log is written when topic configuration is missing.