Reposted by Jia ▫️Mingqian Zheng @mingqian-zheng.bsky.social · 20/10/2025How and when should LLM guardrails be deployed to balance safety and user experience? Our #EMNLP2025 paper reveals that crafting thoughtful refusals rather than detecting intent is the key to human-centered AI safety. 📄 arxiv.org/abs/2506.00195 🧵[1/9] 193
Reposted by Jia ▫️Mingqian Zheng @mingqian-zheng.bsky.social · 20/10/2025[9/9] Big THANKS to my amazing collaborators @jiajiah.bsky.social @pigeonzow.bsky.social Motahhare Eslami, Jena Hwang @faebrahman.bsky.social, @carolynrose.bsky.social @maartensap.bsky.social from @ltiatcmu.bsky.social Pareto.ai @sfu.ca @ai2.bsky.social ♥️ 📂 github.com/EEElisa/LLM-Guardrailsgithub.comGitHub - EEElisa/LLM-GuardrailsContribute to EEElisa/LLM-Guardrails development by creating an account on GitHub. 022
Jia ▫️ @jiajiah.bsky.social · 02/06/2025Some life updates here: got a car, managed to escape grad school, working on expert data curation and future of work these days, had a new cat, moved back to SF, finally feels alive now. 050
Reposted by Jia ▫️Patrick @ninjapleasedj.bsky.social · 19/05/2023Ok, guys I cleared out the fridge and I… OH NO 1923442
Jia ▫️ @jiajiah.bsky.social · 05/05/2023From what I’ve see, most proposed LLM based tool is really just as a proof of concept. If all the “moderation tool” do is to prompt the GPT to produce a formatted JSON file, there’s no way to tune and no barriers of entry. 022
Jia ▫️ @jiajiah.bsky.social · 04/05/2023Twitter right now: Apollo gets crazy after getting shot in the heart. 🙈 020
Jia ▫️ @jiajiah.bsky.social · 04/05/2023My daily reminder of Crazy people posts, but most people are neither crazy nor post. 010
Reposted by Jia ▫️Jay 🦋 @jay.bsky.team · 03/05/2023it’s important for y’all to realize that this app is wild right now because all of you are posters. you’re probably not meant to meet each other in the wild without a cushioning biome of lurkers around 52642128
Jia ▫️ @jiajiah.bsky.social · 04/05/2023[6/]In the past week all my spare bandwidth has been devoted to think about impact of generative ai on social networks. Will we believe the same truth? Are we entering an era or chaos before the new coordination mechanism emerge? Are we getting further away from coordination or closer to it? 000
Jia ▫️ @jiajiah.bsky.social · 04/05/2023[5/] will that hold true now that we have new information retrieval and new narrative generative tool? 100
Jia ▫️ @jiajiah.bsky.social · 04/05/2023[4/] Shared narrative is a coordination mechanism. If people consume the same media, believe in the same news, people see roughly the same truth. That has been historically helped by the fact that some are better at creating narrative than other, and some narrative disseminate further than others. 100
Jia ▫️ @jiajiah.bsky.social · 04/05/2023[3/] Reality is already fractured. As we observe from state laws getting more extreme, opposing side of political spectrum are barely sharing reality, and now shattering of public town-square like Twitter. 110
Jia ▫️ @jiajiah.bsky.social · 04/05/2023[2/] Generative AI seem to be able to make reasonably combing content, if paired with skilled human promoter and editor, the power of making highly convincing narrative will so democratized, that subtle differences of each person’s reality will now become apparent. 110
Jia ▫️ @jiajiah.bsky.social · 04/05/2023[1/] My immediate world has been crazy lately, centralized and decentralized social media, generative AI, political ideology getting extreme. Making sense of those madness makes me feel vulnerable and disoriented. 120
Jia ▫️ @jiajiah.bsky.social · 30/04/2023Decentralized smart fridge, decentralized social media are pretty similar 🤔 020
Jia ▫️ @jiajiah.bsky.social · 27/04/2023Actually kind of love it. The community feels vibrant and every seems excited to build ❤️ 000
Reposted by Jia ▫️Rose 🌹 @rose.bsky.team · 27/04/2023Bluesky SF Meetup #2 happening next Wed (5/3) at the SF Commons. @emily.bsky.team and I will be there to hangout, answer q's, handout invites, etc. Pls RSVP 💙🦋: lu.ma/54jtbp4n 2711223