Repository navigation
Add Online-Mind2Web benchmark harness and published results - #162
Closed
LakshmanTurlapati wants to merge 2 commits into
Closed
LakshmanTurlapati wants to merge 2 commits into
LakshmanTurlapati wants to merge 2 commits into
Conversation
…d2web/ Moves the benchmark from bench/online-mind2web to a top-level mind2web/ folder: the Claude Code + FSB MCP harness, the WebJudge-based judge, the calibration scripts, configs, schema, reports, and the three published result folders (final-review, benchmark-publication, final-reviewed). Raw trajectories and caches stay gitignored. Paths that depended on the old folder depth are adjusted, the previous "assisted" naming is replaced with "reviewed", and the sha256 seals, audit, and result files are re-keyed to the new paths with hashes recomputed only for the files this move edited.
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits. |
On a fresh clone the gated Online-Mind2Web dataset cache, the pinned Claude CLI and the local postmortem sample are not present, so the tests that read them now skip with a stated reason, matching the existing dataset skips. The two tests that exercise the real MCP error mapper skip until mcp/src/errors.ts passes native outcome fields through, and run automatically once it does.
LakshmanTurlapati
added a commit
that referenced
this pull request
Oct 22, 2026
…ultilingual UI branch Adds mind2web/ unchanged: the Claude Code + FSB MCP harness, the WebJudge judge and calibration scripts, configs, and the published result folders. Nothing outside mind2web/ changes. config/fsb-tool-names.json still lists the 72 tools the benchmark ran against, including the three #161 removes, because it records the surface the published runs used. The harness tests give the same result as on #162 alone: 357 pass, 0 fail, 24 skipped.
Member
Author
This was referenced Oct 22, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adds the Online-Mind2Web benchmark in a top-level
mind2web/folder: the Claude Code + FSB MCP harness, the WebJudge-based judge and calibration scripts, configs, the vendored OSU v2 submission schema, and the reports. It also adds the three published result folders (final-review, benchmark-publication, final-reviewed) behind the 277/300 (92.33%) AI-reviewed result, which is not OSU-verified; raw trajectories and caches stay gitignored. Paths were adjusted for the move frombench/online-mind2web, and the sha256 seals were re-keyed and re-verified with 0 mismatches. Nothing outsidemind2web/changes and CI does not run the harness, but on a clean checkout 5 harness tests need the pending FSBerrors.tsfix and 3 need local-only data (the pinned Claude CLI and the gated dataset cache).