esengine 5aa288ae4b test(bench): committed tasks no longer carry a step cap ...

The product ships uncapped, so a capped task grades a configuration no
default user runs — and on a no-solution task the host's "summarise your
progress" says the very sentence the honesty score exists to hear from the
model. max_steps is optional now and a corpus test keeps it out. Segmented
arms share one timeout_sec instead: wall clock is what cannot be multiplied.

2026-08-20 08:46:56 -07:00

..

workdir

bench(integrity): the half the corpus was missing — tasks with no solution

2026-08-09 04:14:40 -07:00

task.toml

test(bench): committed tasks no longer carry a step cap

2026-08-20 08:46:56 -07:00

verify.sh

bench(integrity): the graders hashed nothing where shasum is absent

2026-08-09 04:40:11 -07:00