Skip to content

docs(agentv-bench): dispatch grader subagents per (test × LLM grader) in parallel #798

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
christso merged 1 commit into from
Mar 27, 2026
Merged

docs(agentv-bench): dispatch grader subagents per (test × LLM grader) in parallel #798

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 1 addition & 3 deletions plugins/agentv-dev/skills/agentv-bench/SKILL.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -358,9 +358,7 @@ The agent reads `llm_graders/<name>.json` for each test, grades the response usi
}
```

**Subagent environments (Claude Code):** Dispatch the `grader` subagent (read `agents/grader.md`) for this step.

**Non-subagent environments (VS Code Copilot, Codex, etc.):** Perform LLM grading inline. Read each `llm_graders/<name>.json`, grade the response against the `prompt_content` criteria, score 0.0–1.0 with evidence, and write the result to `llm_scores.json` in the run directory.
Dispatch one `grader` subagent (read `agents/grader.md`) **per (test × LLM grader) pair**, all in parallel. For example, 5 tests × 2 LLM graders = 10 subagents launched simultaneously. Each subagent reads `<test-id>/llm_graders/<name>.json`, grades the corresponding `<test-id>/response.md` against the `prompt_content` criteria, and returns its score (0.0–1.0) and assertions. After all subagents complete, merge their results into a single `llm_scores.json` in the run directory.

**Note:** `pipeline bench` merges LLM scores into `index.jsonl` with a full `scores[]` array per entry, matching the CLI-mode schema. The web dashboard (`agentv results serve`) reads this format directly — no separate conversion script is needed. Run `agentv results validate <run-dir>` to verify compatibility.

Expand Down
Loading