Add sophistry_bench_sprint_env: single-agent advocacy reward-hacking environment#787

Open

acharyaanusha wants to merge 13 commits into

huggingface:mainfrom

acharyaanusha:feature/sophistry_bench_sprint_env

Commits on Jun 11, 2026

feat(sophistry_bench_sprint_env): scaffold OpenEnv package + vendor sprint wheel

acharyaanusha
and
claude
committed
feat(sophistry_bench_sprint_env): add AdvocacyAction/AdvocacyObservation models

acharyaanusha
and
claude
committed
feat(sophistry_bench_sprint_env): typed HTTP client

acharyaanusha
and
claude
committed
feat(sophistry_bench_sprint_env): environment construction + reset()

acharyaanusha
and
claude
committed
feat(sophistry_bench_sprint_env): step() scoring with canonical reward parity

acharyaanusha
and
claude
committed
feat(sophistry_bench_sprint_env): FastAPI app + Dockerfile

acharyaanusha
and
claude
committed
docs(sophistry_bench_sprint_env): README + build/usage

acharyaanusha
and
claude
committed
fix(sophistry_bench_sprint_env): surface reward components over HTTP

acharyaanusha
and
claude
committed
fix(sophistry_bench_sprint_env): error path survives wire + serialization regression test

acharyaanusha
and
claude
committed
docs+test(sophistry_bench_sprint_env): correct image tag; error wire-survival test

acharyaanusha
and
claude
committed
feat(sophistry_bench_sprint_env): depend on published sophistry-bench-sprint 0.1.5 (drop vendored wheel)

acharyaanusha
and
claude
committed
fix(sophistry_bench_sprint_env): address review + align with merged-env conventions

acharyaanusha
and
claude
committed

Commits on Jun 12, 2026

fix(sophistry_bench_sprint_env): harden reward weighting + clarify wire/test contracts

acharyaanusha
and
claude
committed