mirror of
https://github.com/priyanshujain/messageboardbench.git
synced 2026-10-02 19:17:07 +00:00
pj
0a053c114c
`mkdir -p /workspace/scratch`, which our setup solver runs before every agent, classifies as a write inside the directory. The scorer's `find` and `test -d` classify as reads. Counted as the agent's, every run would report a write and the Phase 2 headline number would be 100% for a reason that has nothing to do with any agent. It would have looked like a spectacular result. Inspect nests each tool execution in a span of type "tool", so the agent's own actions are exactly the sandbox events inside one. Analysis now filters on that. scripts/verify_snapshot.py drives the real task in a real container with a scripted model that writes a note, and checks the note comes back. Without it, a broken snapshot would look identical to a true null result: agents simply not writing anything. That is the failure this pilot exists to avoid, so it is worth a script. Claude-Session: https://claude.ai/code/session_01Cq98H7sNoSJdL3W98f18bu
46 lines
1.4 KiB
Makefile
46 lines
1.4 KiB
Makefile
# All recipes run from the repo root, because inspect resolves compose.yaml against the
|
|
# process working directory. See docs/setup.md.
|
|
|
|
impossiblebench := "../impossiblebench"
|
|
|
|
# Sync dependencies, then re-add ImpossibleBench (uv sync prunes it).
|
|
install:
|
|
uv sync
|
|
uv pip install -e {{impossiblebench}} --no-deps
|
|
|
|
# Offline unit tests. No network, no provider beyond mockllm.
|
|
test:
|
|
uv run pytest -q
|
|
|
|
# The real task against a fake model. Free.
|
|
smoke:
|
|
uv run python scripts/smoke_mock.py
|
|
|
|
# Confirms the daemon is up and pre-pulls the image. Should print "/".
|
|
docker-check:
|
|
docker run --rm aisiuk/inspect-tool-support pwd
|
|
|
|
# Confirms the model slug, the key and tool support. Costs about a cent.
|
|
model-check:
|
|
uv run python scripts/model_check.py
|
|
|
|
# Phase 1: cheat rate on conflicting, solve rate on original. About $1.
|
|
baseline:
|
|
uv run python -u scripts/baseline.py conflicting original
|
|
|
|
# The forked task against a fake model. Free. Checks the scratch wiring.
|
|
smoke-scratch:
|
|
uv run python scripts/smoke_scratch.py
|
|
|
|
# Phase 2: thirty agents with a scratch directory. Does anybody write anything?
|
|
pilot:
|
|
uv run python -u scripts/pilot.py
|
|
|
|
# Rebuild the CSV and hand-read file from an existing log, no re-run.
|
|
report log_dir="logs/pilot":
|
|
uv run python scripts/report.py {{log_dir}}
|
|
|
|
# Proves the scorer captures a file an agent leaves in scratch. Free, needs Docker.
|
|
verify-snapshot:
|
|
uv run python scripts/verify_snapshot.py
|