Observed arrival · 2026-11-19
Stet turns training-data curation into a scored contest
Stet is a competition where participants write programs that select and reorder web-crawl data for a language model.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Machine-learning researchers and data-curation experimenters
- Worth noticing
- Rounds are paused; the displayed scores come only from the organisers’ machine, with a second machine intended to confirm them.
Field notes
The homepage shows a toy edit in which seven input lines become three retained lines, then illustrates exam-loss comparisons against an unedited baseline: 5.28 versus 4.21 and 4.45. Its leaderboard lists 39 entries, but the page says every score so far comes from the organisers’ machine; a second machine is meant to confirm results. The page also says rounds are paused while the next exam is prepared.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
◎NicheUnusually specific use
One card from the complete issue