[RecSys-HR 2026 WorkRB challenge] - Graded-relevance benchmarks for skill extraction

Hi there,

I'm excited to share the launch of the RecSys-HR 2026 Challenge, powered by
WorkRB. 🚀


For this challenge, we have re-annotated the most widely used open-source
skill-extraction and normalisation benchmarks (House, Tech, TechWolf,
SkillSkape, SkillNorm) under a new graded-relevance scheme. Every (text,
skill) pair is now labeled with one of five interpretable levels, ranging
from explicitly demonstrated to wrong domain entirely. The full set of new
annotations is released to the community through WorkRB. 🐝


Graded relevance directly addresses the core problems of current binary
benchmarks: they treat near-misses the same as pure nonsense, and we cannot
realistically expect the set of true positives to be exhaustive in a target
space as large as ESCO. With this challenge, we therefore want to measure
real semantic understanding.

 Why participate?

   -

   Win big: compete for a share of the *€4500* *cash prize pool*.
   -

   Get stage time: everyone has a chance to be invited to present their
   work live at RecSys-HR in Minneapolis.
   -

   Drive innovation: help reshape how the world evaluates work-related AI.

Ready to go?

   -

   Full story, metric, and timeline:
   https://workrb.techwolf.ai/challenges/recsys-hr-2026.html
   -

   Register on CodaBench: https://www.codabench.org/competitions/16900/
   -

   Jump straight into the challenge with a pre-configured, working dev
   environment that already incorporates WorkRB's capabilities:
   https://github.com/techwolf-ai/workrb-recsys-hr-challenge-2026


We would love to see your team on the leaderboard, and we are equally happy
if you simply forward this to colleagues or students who might be
interested.

Best regards,


Warre Veys
Warre Veys
NLP PhD Researcher
[image: LinkedIn] add me on LinkedIn
<https://www.linkedin.com/in/warre-veys/>
[image: Phone] 0032468225144
<https://www.techwolf.ai/>

    ----------------------------------------------------------------------------------------
    To unsubscribe from CHI-ANNOUNCEMENTS send an email to:
     mailto:[email protected]

    To manage your SIGCHI Mailing lists or read our polices see:
     https://sigchi.org/operations/listserv/
    ----------------------------------------------------------------------------------------