Repository navigation
feat: fill in the cross-field works the field cites - #58
Merged
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #57
What
crawl.fill_cited_byadds a third crawl phase. After the forward phase, the crawler counts in its own journal how often crawled works cite each work it never fetched. It fetches those cited at least that often, records their references, but doesn't follow them. The phase runs once (filledin the crawl state), so a later run doesn't cascade from the new works' own references.gnn.yamlsets it to 20.Why
Neither direction of the crawl reaches a work from outside the field that the field builds on. On the served crawl, the fill found 1,282 candidates, 827 of them alive, and grew the raw crawl from 9,505 to 10,332 works. Semantic Scholar recovery then matches 10,344 lost citations, up from 6,808.
All four gold originators the crawl was missing are now reachable: Bahdanau, Interaction Networks, Kearnes and DistMult. Kearnes and the Transformer paper make their papers' top 10s, while Bahdanau (18th for GAT), Interaction Networks (13th) and DistMult (21st) come close.
Found holds on test and drops one on dev. The fill also brings in heavily cited datasets and baselines, such as Planetoid, QM9, TransE and RESCAL, which compete for the same places. Telling "built on" from "compared against" needs how a paper cites a work, not whether it does. The graph is now more faithful to what the field cites, so the fill is served. The pre-fill raw crawl and snapshot are backed up under
data/gnn/backup/.Verify
uv run pytest tests/test_crawler.py uv run python -m graph.crawler --field gnn # resumes into the fill: about 25 secondsGenerated with Claude Code