XChava/arn-hai-tos-mdeberta-v1.1-thai

Arn-Hai ToS Clause Classifier v1.1 Thai

Fine-tuned from XChava/arn-hai-tos-mdeberta-v1 using Thai ToS / Privacy Policy data from:

https://github.com/Kantinan20/TOS-Thaidatasets.git

Task

Input: one ToS / Privacy Policy clause Output: joint label:

Category__RiskLevel

Examples:

• —Third Party Sharing__MEDIUM
• —Cookie Tracking__MEDIUM
• —Data Retention__MEDIUM
• —Arbitration__HIGH

Notes

• —This is not legal advice.
• —v1.1 focuses on Thai adaptation.
• —Use with rule-based fallback and human review.
• —If many rows are weak_rule, labels may contain noise.

Training

• —Base model: XChava/arn-hai-tos-mdeberta-v1
• —Thai rows: 1574
• —Max length: 256
• —Epochs: 2
• —Learning rate: 1e-05
• —FP16: False