Repository navigation
docs(symphony): eighteen more groups are ready, their whole recorded sets replayed with no difference, and batch 4's sets are in - #2883
Conversation
…sets replayed with no difference, and batch 4's sets are in The whole recorded set of eighteen groups replayed on main's harness and parser with no difference beyond bellwether #17's separator bytes: glm-5.3-flash, qwen-agentworld-35b-a3b, qwen-drive-1.0-4b, qwen3.5-2b, qwen3.5-9b, qwen3.5-27b, qwen3.5-35b-a3b, qwen3.6-27b, qwen3.6-35b-a3b, qwen3.8-2.4t-a95b, qwen3.8-27b, qwen3.8-flash-next, qwen2.5-7b-instruct-1m, qwen3-vl-8b-instruct, qwen3-vl-8b-thinking, qwen3-vl-30b-a3b-instruct, qwen3-vl-235b-a22b-thinking and webworld-32b, so their rows say ready. bellwether's batch 4 (#85) gives seventeen groups with a table their benchmark-scale sets, so their rows move from awaiting fixtures to replaying the set, with the set sizes from bellwether's README at 32a6f2b6. The table now reads 25 ready, 27 replaying, 0 awaiting fixtures, 26 pending. Signed-off-by: Simo Lin <25425177+slin1237@users.noreply.github.com> Co-authored-by: Chang Su <8605658+CatherineSue@users.noreply.github.com>
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configuration
📒 Files selected for processing (2)
Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 3 remain after this review. 📝 SummarySummary by CodeRabbit
WalkthroughThe test inventory and README now report updated recorded set counts and checkpoint statuses for model groups. The README summary reports 25 ready, 27 replaying, none awaiting fixtures, and 26 pending. ChangesModel checkpoint inventory
Priority: ⬇️ Low Estimated code review effort: 2 (Simple) | ~10 minutes Change: Other Suggested reviewers: Merge Risk: ⚪ Minimal · up to The inventory and README are consistent, and no concrete defect is established. The reported replay results are not independently verified by the checked-in tests, but no actionable merge blocker is identified. 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Comment |
|
👋 The PR description doesn't fully follow
Please update the PR description so reviewers have the context they need. |
There was a problem hiding this comment.
Data-only change, checked against its sources:
- Set sizes: all 17 new
setvalues match the parse-case counts in bellwether's README at 32a6f2b6 (batch 4, #85).step3was recorded with 0 parse cases, so it stays pending. The existing groups' rows did not change upstream, so the sets already listed are still current. - Status moves: the 18 groups moved to
Readyare exactly the 18 in the PR's replay table. In each row, bitwise + separators-only + listed adds up to the recorded set size. Separators-only is the harness's accepted bellwether #17 class. - Tally: 7+18=25 ready, 28−18+17=27 replaying, 0 awaiting, 26 pending, 78 total.
- README: all 78 generated rows match
GROUPSfor slug, set and status. Theevery_status_has_what_it_claimschecks hold.
🔴 0 · 🟡 0 · 🟣 0
hello-alexmcc
left a comment
There was a problem hiding this comment.
Rule-17 review of #2883 at cef4604: approve. I checked the README's numbers against their sources.
- Tally: 25 ready, 27 replaying, 0 awaiting fixtures and 26 pending make 78, as 7, 28, 17 and 26 did before.
- Eighteen to "ready": they are exactly the description's eighteen replays. In each, bitwise plus separator-only plus listed cases add up to the group's cases, with 0 differences. Their case counts in the table are unchanged.
- Seventeen to "replaying the set": they are exactly batch 4's groups (#85, bellwether main 32a6f2b6) except
step3, which has no table and stays pending. Each one's case count equals the parse cases in bellwether's README at that commit: 17 of 17.
Gates. I ran them at cef4604 on main 21d5e44, with Rust 1.98.0, its own target dir and -j 2; each exited 0:
cargo +nightly fmt -p smg-symphony --checkcargo clippy -p smg-symphony --all-targets --all-features -- -D warningscargo test -p smg-symphony: 307 lib, 10 fixture, 7 contract and 5 README testsRUSTDOCFLAGS="-D warnings" cargo doc -p smg-symphony --no-deps
Rule 17:
- Fixture ids: none changed.
- Protected paths: the crate's README and its README test.
- Trailers: the author's
Signed-off-byand aCo-authored-by, and no AI trailer. - Threads: none.
I didn't re-run the eighteen whole-set replays; the figures above are the description's, checked for consistency.
What
The models table in
crates/symphony/README.md, regenerated fromtests/readme.rs(SYMPHONY_WRITE_README=1 cargo test -p smg-symphony --test readme), insidecrates/symphony/only:reasoning-parsercrate #17's separator bytes (the template's newline the reference keeps in its text), every chunking the harness has.Replays
"Separators only" is bellwether #17's class: the reference keeps the template's newline in its text, the parser does not, and the two agree once it is trimmed; "listed" are the two bellwether #16 probes a table reads as its own markers. Logs: the queue's
whole_queue.logand onewhole-<slug>.logper group.Gates
By exit code (
.claude/symphony/gates.sh): fmt, clippy-D warnings, test (307 lib, 10 fixture, 7 contract, 5 README), doc, width, codespell, all 0.Not in this PR
Merge
Squash under rule 25 (rule 18 does not apply: data only). Subject: the PR title plus
(#<n>). Body: