Skip to content
FARTBench Core R01
  • Leaderboard
  • Results
  • Review
  • Scenes
  • Methodology
  • Data
  • About
  • Archive

Legacy / experimental Everything in the archive is historical. This page belongs to the Core v0.1 pilot design, with its single composite FARTScore, FartElo and single-score leaderboard. It is kept in the archive as a historical record and is not the current method — see the Core R01 leaderboard and methodology.

FARTBench history

Archive

Earlier designs and the R00 pilot, kept exactly as a record. None of it is the current method; the Core R01 leaderboard and methodology are.

  • Legacy / experimental R00 pilot case study The first blind round (FB-C02 only): its reveal, results and what it taught the Core design.
  • Legacy / experimental Legacy leaderboard The Core v0.1 commissioning matrix and single-score leaderboard design. No official scores.
  • Legacy / experimental FARTScore The pilot-era weighted composite score, kept for reference. Core R01 does not use it.
  • Legacy / experimental FartElo Arena A static prototype of pairwise voting. Never live.
  • Legacy / experimental Pilot tests The eight Core v0.1 test specifications.
  • Legacy / experimental Pilot failure taxonomy The pilot-era failure codes.
  • Legacy / experimental Pilot model registry Commissioned configurations under the pilot protocol.
  • Legacy / experimental Component gallery The site foundation’s illustrative components.

FARTBench

Flatulatory Audiovisual Realism & Temporal Benchmark

Independent benchmark research

Core R01

  • Leaderboard
  • Results
  • Review blind clips
  • The four scenes

How it works

  • Methodology
  • Models & provenance
  • Human review
  • Round status

Release

  • Data & downloads
  • Technical report
  • Citation
  • About

Archive

  • Archive home
  • R00 pilot
  • Leaderboard (legacy)
  • FARTScore (legacy)
  • FartElo Arena (legacy)
  • Pilot tests
  • Pilot failure taxonomy

Core R01 · human instrument fartbench-core-r01-human-v3 · Overall Ranking V1 · Technical Report FB-TR-2026-001

Core R01 results are revealed and frozen: blind and reproducible, with Overall Ranking V1 adopted after the reveal. Pilot-era surfaces live in the archive. The subject matter is intentionally ridiculous; the evaluation protocol is not.