Datasets:
sponsor-segment-control-v1
10,800 original typed rows; 400 complete synthetic families. Original generated content and labels are CC0-1.0. Source code keeps the repository's MIT license.
Prepared and independently audited only. Not used for training or model evaluation. This addition changes neither the frozen release-v2 mixture nor the data used by the published 2B/9B checkpoints.
Categorize original timestamped transcript segments with one Choice per segment: sponsor, self_promo, intro, outro, recap, content or other. Paid third-party promotion needs affirmative funding evidence; creator-owned promotion is separate. A brand, discount code or promo marker alone does not establish sponsorship. Complete video families and payment-evidence counterfactuals share a split; OOD reserves complete sentence wording. The data are finite synthetic transcripts, not scraped videos or an audio/visual benchmark.
| Split | Typed rows |
|---|---|
| train | 6,345 |
| calibration | 756 |
| validation | 540 |
| test | 999 |
| ood | 2,160 |
The community interface reference supplies the task shape. The task instances, labels and instructions are independently authored. No original third-party transcript, whole documentation page or model response is bundled.
Use only state, question, kind and options as model inputs. Targets are reference labels; metadata and auxiliary cases may contain privileged information. The Parquet columns and JSON decoding rules match the root dataset card. Decompressing each raw split recovers its original JSONL bytes exactly. Whole-family splits include correlated counterfactuals; row counts must not be described as independently collected real-world cases.
From the source snapshot's root, regenerate in a new directory with:
python3 -m jev.case_sponsor_segments --output-dir data/sponsor-segment-control-v1 --groups 400 --ood-groups 80 --seed 42
No trained-model accuracy, useful compaction, natural-video classification, or general API failure-detection performance is claimed by this release.