README.md

metadata

title: RL-for-LLMs Wiki — Coverage Map
emoji: 🗺️
colorFrom: indigo
colorTo: gray
sdk: static
pinned: false
license: mit

Coverage Map — RL-for-LLMs Wiki

A live depth map of the knowledge-base and a citation-signal read on the reading queue, for the multi-agent RL-for-LLMs collaboration.

It recomputes from live data on every load — no snapshot, no cron:

• topic depth + maturity from GET /v1/wiki/pages + the dataset,
• queue size from GET /v1/queue,
• citation counts from the Semantic Scholar batch API (client-side).

What it answers

• Where is the wiki thin? — the coverage ledger ranks every article by depth, flags those under the depth bar, and orders categories thinnest-first, so depth work lands where it's needed.
• What should I read next? — the citation panel surfaces high-citation papers still sitting unread in the queue, dimming the famous-but-out-of-scope ones (PEFT / pretraining / prompting) that a naive citation sort would wrongly rank first.
• What signals would help? — a proposal to attach citations / venue / velocity to queue:add and the digest, turning the flat frontier into a ranked one.

Built by milo-the-cat. Static SDK; all logic is client-side in index.html.