Dataset Viewer
Auto-converted to Parquet Duplicate
schema_version
string
snapshot_date
timestamp[s]
incremental_watermark
timestamp[s]
dump_file
string
dump_sha256
string
bytes
int64
format
string
row_counts
dict
tables
list
1
2026-08-09T00:00:00
2026-08-09T00:00:00
base/arxiv_graph_2026-08-09.dump.xz
1c65c9347dba421c288db597beb05b4f8d79f005e1b611f94f8671c2cf6b7b31
9,691,532,520
pg_dump-custom-uncompressed+xz
{ "papers": 2571654, "citation_edges": 25562450, "paper_authors": 4147089, "raw_citations": 99191237, "arxiv_meta": 6 }
[ "papers", "citation_edges", "paper_authors", "raw_citations", "arxiv_meta" ]

AGGS arXiv citation graph

One dated custom-format pg_dump of the arXiv citation graph used by OpenCode’s arxiv tool.

Project page: dlab-cmu/agent-guided-graph-search Wikipedia snapshots: dlab-cmu/aggs-wikipedia

Layout

manifest.json
base/arxiv_graph_YYYY-MM-DD.dump.xz

opencode research install --arxiv defaults to this manifest.json. Download uses hf download (Xet). Restore decompresses, then pg_restore --jobs, and writes incremental_delta_last_applied_date.

Not in this dataset: arXiv e-prints / PDFs.

Current base

Path Size
base/arxiv_graph_2026-08-09.dump.xz 9.69 GB
manifest.json watermark 2026-08-09

Download check: 9,691,532,520 bytes in 85s (~114 MB/s); SHA-256 matched.

License

Graph metadata follows arXiv metadata terms. This is not a redistribution of arXiv source tars.

Downloads last month
19

Collection including dlab-cmu/aggs-arxiv