Catch @ZihaoZhao1 at today’s poster session (10:30–12:00) where he'll be presenting SynthTextEval! Stop by if you're interested in synthetic text for high-stakes domains. Zihao also has another EMNLP paper on private text generation, for people interested in this space! @jhuclsp

Thank you to @anjalie_f for advising. Hands-on with DP-SGD? Start with our another paper and open-source package ( https://t.co/XZK5nG7d93 https://t.co/rnZXQsM2cX)

🔗 𝗣𝗮𝗽𝗲𝗿 & 𝗰𝗼𝗱𝗲 Paper is accepted to EMNLP 2025 Main arXiv: https://t.co/IgX1Jh7DZf Code: https://t.co/RzbFVzyWtA #SyntheticData #Privacy #NLP #LLM #Deidentification #HealthcareAI #LLM

4/5 📈 𝗨𝘁𝗶𝗹𝗶𝘁𝘆 On TAB, prefix-tuning+masking gives best utility (Perplexity ≈ 10.2, MAUVE ≈ 0.83), beating ICL and DP-SGD. Similar trends on MIMIC-III.

3/5🔒 𝗣𝗿𝗶𝘃𝗮𝗰𝘆 ICL+blocking: ~0.00% privacy leakage (avg in our runs). Prefix-tuning+masking yields the lowest ROUGE vs training data (e.g., ROUGE-L ≈ 0.098), indicating less copying.

2/5 🔧 𝗛𝗼𝘄 𝗶𝘁 𝘄𝗼𝗿𝗸𝘀 • Build control codes from detected private entities (PERSON, ORG, LOC, etc.). • Generate with either ICL (and block those identifiers at decode time) or prefix-tuning with a privacy mask + KL/contrastive losses.

🚀Text anonymization is hard; DP often hurts utility. We use entity-aware control codes + either ICL(with bad-token blocking) or prefix-tuning w/ masking to get strong privacy–utility tradeoffs on legal & clinical data, outperforming DP-SGD in practice. https://t.co/Kt0PIoYsq3

We introduce WaltzRL🎶, a multi-agent RL framework that treats LLM safety as a positive-sum game between conversation & feedback agents. It strikes an elegant balance between helpfulness & harmlessness, boosting safety & reduces overrefusals without degrading capabilities!

Forget modeling every belief and goal! What if we represented people as following simple scripts instead (i.e "cross the crosswalk")? Our new paper shows AI which models others’ minds as Python code 💻 can quickly and accurately predict human behavior! https://t.co/1t2fsW7jyL🧵