thetalab/vector-edit-gym

Coding

Multimodal

106 hand-authored SVG editing tasks across four difficulty tiers (easy / medium / hard / very_hard). Each task is a corrupted SVG plus a natural-language fix instruction. Strict structural and preservation metrics catch stylistic drift.

← Back to Registry

Run this task

CLI:

inspect eval inspect_harbor/thetalab_vector_edit_gym --model openai/gpt-5

Python:

from inspect_ai import eval
from inspect_harbor import thetalab_vector_edit_gym

eval(thetalab_vector_edit_gym(), model="openai/gpt-5")

Dataset information

Harbor registry	thetalab/vector-edit-gym
Inspect task	`thetalab_vector_edit_gym`
Latest digest	sha256:b2ce5a813991c55c7b610c77a52a5c54f656aeb2b7239558a0a846a9be62c625
Samples	106

See Task Parameters for the parameter set shared across all Harbor tasks.