Research map / Federated learning and privacy / Privacy attacks and secure learning
CAPP-130: A Corpus of Chinese Application Privacy Policy Summarization and Interpretation
Pengyun Zhu, Long Wen, Jinfei Liu, Feng Xue, Jian Lou, Zhibo Wang, Kui Ren
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use RoBERTa instead of Bert as encoder for topic screening | TCSI-pp information extraction · CAPP-130 topic-multi | Micro-F1: 0.802 -> 0.819 | Improved |
| use RoBERTa instead of SBERT as encoder for topic screening | TCSI-pp information extraction · CAPP-130 topic-multi | Micro-F1: 0.781 -> 0.819 | Improved |
| use RoBERTa instead of PERT for important-binary screening | TCSI-pp information extraction · CAPP-130 important-binary | Macro-F1: 0.897 -> 0.899 | Improved |
| use RoBERTa instead of PERT for risk-binary screening | TCSI-pp information extraction · CAPP-130 risk-binary | Micro-F1: 0.922 -> 0.920 | No clear effect |
| use RoBERTa instead of Bert for risk-binary screening | TCSI-pp information extraction · CAPP-130 risk-binary | Macro-F1: 0.719 -> 0.711 | No clear effect |
| use mT5 instead of Bert2gpt for sentence rewriting | TCSI-pp-zh rewriting model · CAPP-130 rewriting | Carburacy: 0.764 -> 0.833 | Improved |
| use mT5 instead of ERNIE2gpt for sentence rewriting | TCSI-pp-zh rewriting model · CAPP-130 rewriting | ROUGE-1: 0.623 -> 0.753 | Improved |
| use mT5 instead of Bert2gpt for sentence rewriting | TCSI-pp-zh rewriting model · CAPP-130 rewriting | ROUGE-2: 0.574 -> 0.609 | Improved |
| use TCSI-pp-zh instead of GPT-4 summaries | privacy policy summarization · Readability survey | share of votes: 42.13% -> 46.23% | Improved |
| use TCSI-pp-zh instead of Llama2-Chinese summaries | privacy policy summarization · Readability survey | share of votes: 6.39% -> 46.23% | Improved |
| data usage topic classified with RoBERTa/mBERT | fine-grained screening · CAPP-130 topics | F1 | Hurt |
Closest papers by meaning
- Position: Contextual Integrity is Inadequately Applied to Language Models (ICML 2025 · similarity 0.82)
- Can LLMs Keep a Secret? Testing Privacy Implications of Language Models via Contextual Integrity Theory (ICLR 2024 · similarity 0.81)
- Empowering Users in Digital Privacy Management through Interactive LLM-Based Agents (ICLR 2025 · similarity 0.81)
- Beyond Memorization: Violating Privacy via Inference with Large Language Models (ICLR 2024 · similarity 0.81)
- Synthesizing Privacy-Preserving Text Data via Finetuning *without* Finetuning Billion-Scale LLMs (ICML 2025 · similarity 0.81)
- SynBench: A Benchmark for Differentially Private Text Generation (NeurIPS 2026 · similarity 0.81)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
