Repository navigation
fix(tasks): a reply chain counts +2, not +6, in the duplicate check - #1497
Merged
Merged
Conversation
On 213 labelled pairs from real chats, pairs shown as possible duplicates because one task's message replies to the other's were wrong 11 times in 13. At +2 the shown pairs' precision rose on both labelled sets (52% -> 57%, 43% -> 47%) for one lost true pair each. Other weights unchanged. The review-list tests relied on the old +6 to flag a bare "any news?" reply; their reminders now name the invoice and amount, as a real reminder would. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
leemour
added a commit
to WireCatLabs/cli-docs
that referenced
this pull request
Oct 18, 2026
…cond labelled set Weights table and the real-chats paragraph, in en/ru/es. Merge after the cli-core release that carries WireCatLabs/cli-core#1497. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
leemour
added a commit
to WireCatLabs/cli-docs
that referenced
this pull request
Oct 18, 2026
## Task Document the reply-chain weight change from WireCatLabs/cli-core#1497. ## Why Among the pairs shown as "maybe" or "likely" in 213 labelled real-chat pairs, the ones in one reply chain were wrong 11 times in 13. The weight goes from +6 to +2. Source: [cli-testing, "Follow-up: a second labelled set"](https://github.com/WireCatLabs/cli-testing/tree/main/performance/tasks/duplicates#follow-up-a-second-labelled-set-2026-10-17). ## What In `content/docs/tasks-architecture{,.ru,.es}.mdx`: - the weights table row "Reply chain: yes / no" now reads +2 / 0; - the "On real chats" paragraph says what the second set showed. ## Merge Draft. Merge after the cli-core release that carries #1497. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
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
1 participant
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.
Task
The owner chose to lower the reply-chain weight in the task duplicate check after the second labelled set of real-chat pairs.
Why
cli-testing performance/tasks/duplicates, "Follow-up: a second labelled set" labelled 213 pairs from real chats. Among the pairs shown as "maybe" or "likely", the ones with a reply chain were wrong 11 times in 13. At +2, shown-pair precision rose on both sets: 52% → 57% on the old set and 43% → 47% on the new one, losing one true pair on each.
What
TASK_SIMILARITY_WEIGHTS.reply.yesgoes from +6 to +2, with its reason updated.docs/dev/tasks-plan.mdgets the new weight in its weights table.Testing
pnpm typecheck: passes.pnpm lint: no errors (the 69 warnings were already there).pnpm test: 5155 passed, 4 skipped.pnpm docs:check: ok.pnpm standard:docs: no changes.Draft docs PR (merge after the release): WireCatLabs/cli-docs#305.
🤖 Generated with Claude Code