Repository navigation
Release 1.1.2 - #14
Merged
Merged
Conversation
- Title: "FA-Bench: A Benchmark for Phone- and Word-Level Timestamp Accuracy in Forced Alignment and ASR on Clean and Noisy Speech". - The word table's caption names Olign 0.9, the version that produced its numbers. - The abstract and the introduction open with a new first sentence. The abstract keeps its word count. - The text gives Qwen3-FA's F1 at 20 ms as 0.40, as its table does. Same five pages: the introduction ends on page 1 and the conclusion on page 4. docs/paper.pdf is the rebuilt paper.
The Olign rows in records/202608 and records/202609 came from Olign 0.9, and the tables, the provenance block, the adapter's docs and the paper caption called it Olign 1.0. Each now says Olign 0.9. No number changes. - update_public_tables.py displays every Olign row as "Olign 0.9", and the records are regenerated from it. - gen_provenance.py and _provenance.json give the version as v0.9. - gen_paper_tables.py writes Olign 0.9 into both captions.
ElevenLabs' /v1/forced-alignment endpoint is handed the reference transcript with the audio and returns the time of each word, so it is a Track 1 aligner on the word tier, beside its own Track 2 row `elevenlabs`. - The cloud adapter gains `sends_transcript`. When set, align() sends the reference transcript, refuses an item without one, and puts the text in the cache key. The Track 2 rows leave it off, so their keys and cached responses are unchanged. - The adapter finds its cache from where its recipe lives, so a recipe under evals/aligners/ caches there. - The recipe ships with `enabled: false`, since every request bills the account. No numbers are published for it. - cloud_check.py takes --text to check a forced aligner, and cloud_cost.py prices the row and counts its cache.
- fabench.__version__ is 1.1.2. - CHANGELOG: Unreleased becomes 1.1.2. ElevenLabs forced alignment is added, the cloud adapter can send the transcript, Olign is named 0.9 and docs/paper.pdf is the revised paper. No change moves a published number, so records/202609 stands as scored for 1.1.0.
Code reviewNo issues found. Checked for bugs and CLAUDE.md compliance. |
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.
v1.1.2, a maintenance release. No change moves a published number, so
records/202609/stands as scored for v1.1.0.docs/paper.pdfis the revised paper. The title is now "FA-Bench: A Benchmark for Phone- and Word-Level Timestamp Accuracy in Forced Alignment and ASR on Clean and Noisy Speech", the word table names Olign 0.9, and the abstract and the introduction open with a new first sentence. The tables are unchanged.elevenlabs_fa. It ships withenabled: falseand no published numbers, because every request bills the account.fabench.__version__is 1.1.2, and the CHANGELOG has the 1.1.2 section.Lint, test and selftest pass on a clean export of the branch.