The authors treat harness optimization as a two-level process.
Verified self-improvement makes agent harness optimization more reliable
VERSE improves an optimizer’s own procedures while testing proposed changes to software-engineering agents through execution.
Big Tech
Zekai Wang · Yingqiang Ge · Zekun Wang · Hai Wang · Yuhui Xu · Joshua Frandsen · +3 more
MIT · Amazon
Research Digest··2 min read
Wang et al.
Why this paper
From Amazon and MIT
In one line
VERSE improves harness optimizers by letting them test edits and evolve their own verification tools.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ✓Reports numbers on named benchmarks (2 benchmarks)
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§