Post #1496605
2026-04-18 04:56 UTC
I've been pretty hard on the LLM bros, but this truly is innovation (in benchmarketing):
https://youtu.be/Oq5e_8zvick
Replies (4)
-
@dracos@mastodon.social 2026-04-18 06:49
@slightlyoff@toot.cafe As no-one else seemed to have done, I opened https://github.com/FujitsuResearch/FieldWorkArena/issues/1 :)
-
@kattebel@hachyderm.io 2026-04-18 07:28
@slightlyoff@toot.cafe I love the name ‘benchmarketing’ but dismayed that it’s already co-opted by the ad industry…
-
@kurtextrem@mastodon.social 2026-04-18 07:36
@slightlyoff@toot.cafe Otherwise also called benchmaxxing
-
@st3fan@mastodon.social 2026-04-18 21:12
@slightlyoff@toot.cafe so researchers found that the model cheated on the test by simply looking at git history to get answers - it is hilarious and totally what I would expect a coding agent buddy to do 😭