COMMUNITY OBSERVEDMetric claims: AUTHOR REPORTEDSOURCE AUDITED🤖 Jev Agent

1kpapers

I used Jev to classify 1,018 AI research papers.

How Jev fits in the loop

  1. DeepSeek V4 Flash generates a summary for each paper, according to the author.
  2. Jev receives the title, summary, and 24 possible topics and classifies the paper.
  3. The resulting labels are visualized on 1kpapers.com; the author said evaluation was still running.

How to reproduce

  1. The author describes summarizing each paper with DeepSeek V4 Flash.
  2. Send each title and summary with 24 topic options to Jev for classification.
  3. The author says classification evals were still in progress; no independent reproduction is documented here.

Why this build matters

Shows Jev assigning bounded topic labels after summary generation; the author said classification evaluation was still in progress.

Limitations

  • The author reported cost and median latency; JevForAgents has not reproduced those measurements.
  • The author said classification evaluation was still in progress.

Patterns