evosearch

Run the CaP Fix Loop + Evolutionary Search experiment on a user-supplied list of low-performing benchmark tasks: iterative K=8 candidate search on development seeds 51–65, validation selection, and final held-out evaluation on seeds 1–50.

Try your first task

Check readiness for evolutionary search

Start a fresh conversation.

Use the evosearch skill to review readiness for one low-performing benchmark task I provide. Check benchmark setup, SAM3 authentication, perception services, and any existing fix_code.py and baseline held-out rate. Explain how candidate_A and the development and held-out partitions apply, then list missing inputs. Stop before dispatching workers or running evaluations.

Required inputs

One low-performing task identified by benchmark, suite, and task_id.
An ASPIRE checkout and the target benchmark's setup details.
The task's existing fix_code.py and baseline held-out rate, when available.

Suggested template by Skill Leaderboard, grounded in the linked source. Task evidence ↗

Inside SKILL.md

53 lines

Section content has not been verified for this skill yet.

Full source remains withheld under the current source-display policy. Read the author’s source ↗