refactor(benchmark): unify LHTB and SWE-Marathon execution - #4671

Merged

huangruiteng merged 10 commits into

mainfrom

codex/benchmark-runtime-unification

Sep 18, 2026

Commits on Sep 17, 2026

feat(runtime): support explicit external isolation and retain Goal timeout evidence
huangruiteng
committed
9879f1e
refactor(benchmark): share Harbor Codex execution across benchmark families
huangruiteng
committed
2e876d2
docs(benchmark): document shared execution and retire WEN runner guidance
huangruiteng
committed
0147726
fix(benchmark): align shared bootstrap and preflight with current CLI
huangruiteng
committed
3c9c20a
fix(benchmark): pin source provenance and cover Turn recovery
huangruiteng
committed
d668688
docs(benchmark): remove retired wake timeout setting
huangruiteng
committed
ae4bc7a
fix(runtime): preserve scoped Turn approval and clean up cancelled workers
huangruiteng
committed
e6c023c
docs(benchmark): record shared runtime engineering checkpoint
huangruiteng
committed
5fd33d0
fix(benchmark): archive the verified immutable source revision
huangruiteng
committed
587b48c

Commits on Sep 18, 2026

feat(benchmark): separate seeded task entry from LoopX planning (#4676)
huangruiteng
authored
5728591