Skip to content

Backfill reviewed historical arXiv papers into radar discovery - #708

Closed
rudycelekli wants to merge 1 commit into
ktwu01:mainfrom
rudycelekli:feat/reviewed-arxiv-backfill
Closed

rudycelekli wants to merge 1 commit into
ktwu01:mainfrom
rudycelekli:feat/reviewed-arxiv-backfill

Conversation

@rudycelekli

Copy link
Copy Markdown
Contributor

Refs #585.

The arXiv category RSS path only carries recent announcements. A paper first submitted before collection began cannot reappear in that feed, so search currently returns unrelated lexical candidates for When AI Benchmarks Plateau and Benchmark Illusion without either paper's arXiv record.

This PR adds a reviewed historical arXiv source input and feeds it through the existing daily discovery state, scoring, and snapshot pipeline. It seeds two verified papers from their primary abstract pages:

Each row has an exact arXiv ID/URL, source dates, title, and short verbatim abstract excerpt. The loader rejects bad IDs, duplicate IDs, uncited URLs, invalid dates, and future revisions. The records enter a normal snapshot once as newly discovered historical evidence; a live item with the same arXiv ID takes precedence. The collection method reports reviewed backfill accurately, and the file path resolves relative to config.yml. The next normal daily run will materialize these records in the published radar archive.

Validation from a clean worktree with the paper submodule initialized:

  • ruff check .
  • ruff format --check .
  • benchmark-radar normalize-catalog (1,298 records)
  • benchmark-radar classify
  • benchmark-radar build-data-release
  • pytest -q (1,403 passed)
  • Focused pipeline test writes and reloads a snapshot, verifies repeat suppression and live-source precedence.

@rudycelekli
rudycelekli requested a review from ktwu01 as a code owner September 29, 2026 23:30
@rudycelekli

Copy link
Copy Markdown
Contributor Author

Agent-authored PR annotation: 330226 GPT.

ktwu01 commented Oct 7, 2026

Copy link
Copy Markdown
Owner

Closing as not planned. One author now has a cap of 10 open issues + PRs, and this account had 56 open PRs and 11 open issues, mostly paired with each other. We merged the three changes that fix silent data loss in the corpus (#723, #774, and #773 once rebased) and are closing the rest so review stays possible.

This is not a verdict that the change is wrong. If you think this one matters, reopen it later when you are under the cap, one at a time, with a reproduction from a real user or caller rather than a hypothetical edge case.

不是否定这个改动本身。为保证评审能跟上,同一作者最多同时保留 10 个未关闭的 issue/PR;该账号曾有 56 个未关闭 PR 和 11 个 issue。已合并修复语料静默丢失的三个改动,其余先关闭。若认为本条重要,请在低于上限后逐个重新打开,并附真实用户或调用方的复现。


Generated by Claude Code

@ktwu01 ktwu01 closed this Oct 7, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants