ettin-17m-sponsor-v2

Ettin 17M sponsor detector v2 (archived)

Full retraining experiment on the audited dataset revision. It did not replace v1 and is kept for comparison only.

• Status: archived experiment
• Base encoder: jhu-clsp/ettin-encoder-17m @ 987607455c61e7a5bbc85f7758e0512ea6d0ae4c
• Architecture: ModernBERT-family Ettin 17M (ModernBertForTokenClassification)
• Labels: O, B-SPONSOR, I-SPONSOR, L-SPONSOR, U-SPONSOR
• Repo: CuriousDragon/ettin-17m-sponsor-v2

Use with Transformers

from transformers import AutoModelForTokenClassification, AutoTokenizer

repo = "CuriousDragon/ettin-17m-sponsor-v2"
tokenizer = AutoTokenizer.from_pretrained(repo)
model = AutoModelForTokenClassification.from_pretrained(repo)

For full YouTube transcripts, tokenize into overlapping 768-token windows (128-token overlap), decode BILOU spans, map characters back onto caption cues, and merge gaps of 24 normalized characters or 1500 ms. See the Flow ml/sponsor_detection inference pipeline.

Token metrics (held-out test, 33,578 windows)

Intended use

On-device or local inference over English video transcripts to propose paid sponsor ranges. In Flow, SponsorBlock remains authoritative when it has segments; these weights are a fallback / shadow detector.

Limitations

• English transcripts only.
• Target is external paid sponsorships. Self-promotion, memberships, merchandise, and interaction reminders are treated as negatives.
• SponsorBlock labels are noisy and incomplete. Evaluations against them are weak labels, not ground truth.
• Full-video reconstruction interpolates timestamps inside caption cues.
• Window-level scores do not by themselves skip video; Flow stitches overlapping 768-token windows with a 128-token overlap.

Use v1 or v3-replay instead of this checkpoint for new work.

License

These weights are released under CC BY-NC-SA 4.0.

The base encoder jhu-clsp/ettin-encoder-17m is MIT. Fine-tuning used SponsorBlock labels (CC BY-NC-SA 4.0) and the pinned Xenova/sponsorblock-768 window corpus. Non-commercial use and share-alike apply to this derivative.

This project is not affiliated with SponsorBlock.