Junjie Hu
@JunjieHu12
Followers
2K
Following
3K
Media
4
Statuses
392
Assistant Professor @UWMadison PhD in #NLProc & #ML & #AI @LTIatCMU @SCSatCMU @CarnegieMellon. Interned @GoogleAI @MSFTResearch.
Madison, WI
Joined November 2014
RT @gneubig: Multimodal LLMs lack cultural knowledge, especially for less well-represented cultures and languages. To fix this, we created….
0
13
0
With Dawei Zhou (Virginia Tech), we’ll advance algorithms, theory & systems that enable foundation models to reason over rich, multimodal, and ever-evolving open-world knowledge. Excited for the journey ahead!.🔗
2
0
0
I'm thrilled to share that our project on large open-world foundation models with evolving world knowledge has been funded by the NSF IIS Core program—my first NSF lead PI grant! 🚀.#NSF #AI #NLP #MachineLearning . 🔗
3
0
21
RT @gneubig: Summary of GPT-OSS architectural innovations:. 1. sliding window attention (ref: .2. mixture of expert….
arxiv.org
Deploying Large Language Models (LLMs) in streaming applications such as multi-round dialogue, where long interactions are expected, is urgently needed but poses two major challenges. Firstly,...
0
361
0
Excited to share our recent work on empowering foundation models with LoT for automatic theorem proving!!.
🚀 Introducing MA-LoT Theorem Framework: An open-source multi-agent framework utilizing the Long Chain-of-Thought to boost automated theorem-proving🎉. ✅ Achieving 61.07% accuracy rate under pass@32 on MiniF2F-Test outperforming Goedel-Prover, Lean_STP and DeepSeek-Prover-V1.5
0
0
6
Excited to share our latest work on pluralistic alignment of LLMs! Individuals across communities share diverse or even conflicting preferences! Yet current alignments on mixed human preferences make LLMs biased toward major preferences—our GDPO fixes it!.
🚀 Excited to announce our paper "No Preference Left Behind: Group Distributional Preference Optimization (GDPO)" has been accepted to #ICLR2025! 🎉 We investigate: How can a probabilistic LLM align with distributional preferences within a group? 🧵[1/n]
2
5
75
RT @RUppaal: @iclr_conf paper alert! The de facto way to align a model through tuning-based methods like DPO is powerful, yet expensive and….
0
10
0
Our paper has been accepted to #ICLR2025 Check out our latest version for a lightweight, noise-robust way for preference tuning! congrats to the team!.
Super excited to share our work on improving the safety of LLMs. Our work provides a lightweight, noise-robust way to reduce model toxicity. lots of fun working w/ my amazing collaborators @RUppaal @ApratimDey2,Yiting He,@Yiqiao_Zhong #NLP #ML.
0
1
15
RT @kwangmoo_yi: I generally don't like being controversial, but I am sad that this release is completely misleading. It is NOT generative….
0
89
0
RT @Thom_Wolf: Exponentially growing number of open-source AI models over the course of the past 30 months
0
14
0
Curious to know how AI improves scientific peer reviews on AI papers? Please check out our latest work at #CSCW2024 it’s a great pleasure to work with @LuSun_Selena . Chat with her at the venue!.
Can AI improve the peer review process? I am excited to present MetaWriter: Exploring the Potential and Perils of AI Writing Support in Scientific Peer Review at #cscw2024 @ACM_CSCW .Today🗓️Session 3a 2:30pm-4:00pm. Please join us for the discussion. Collaboration with
1
1
5
We are excited to share our latest work on benchmarking current LLM-based machine translation and traditional NMT on culture-specific concepts. Chat with us at 4-5:30pm on the poster session #EMNLP2024! Joint work w/ @Binnie8545 @SeleenaJiang @Diyi_Yang.
🚀 Excited to present our paper "Benchmarking Machine Translation with Cultural Awareness" at #EMNLP2024! . We build CAMT, a novel parallel corpus enriched with culture-specific item annotations, and evaluate how well NMT and LLM-MT systems handle cultural entities.
1
4
18
Come and chat with us at EMNLP! Excited to share our recent paper on role-playing LLM agents!.
Excited to present my paper on role-playing LLM agents at #EMNLP2024! 🎉. Paper Title: “Beyond Demographics: Aligning Role-playing LLM-based Agents Using Human Belief Networks”. Come say hi and let's chat about some exciting research! 🤖🧠✨. TL;DR: How can we make LLM agents
0
0
11
RT @SeanChuang4: Excited to present my paper on role-playing LLM agents at #EMNLP2024! 🎉. Paper Title: “Beyond Demographics: Aligning Role-….
0
15
0
RT @hanzhao_ml: 🚨🚨 We are hiring! RT appreciated!. Prof. Rui Song ( and I will recruit post-doc scientists through….
amazon.science
The program offers recent PhD graduates an opportunity to advance research while working alongside experienced scientists with backgrounds in industry and academia.
0
101
0
Feel free to chat with me and Sean in the poster section! I'm at #NAACL2024, hosting the "ML for NLP" presentation section at 9-10:30 AM, and organizing the BoF section for "ML for NLP" at 3:30-5 PM on Jun 18 (tomorrow).
I am at #NAACL2024 this week. Can we use LLM agents to simulate human-like society? . TL;DR: They are not human-like by default, unless you explicitly induce human-like cognitive bias!. Come chat with me! .⌚Wed (6/19) 11AM - 12:30PM.
0
0
13
Super excited to share our work on improving the safety of LLMs. Our work provides a lightweight, noise-robust way to reduce model toxicity. lots of fun working w/ my amazing collaborators @RUppaal @ApratimDey2,Yiting He,@Yiqiao_Zhong #NLP #ML.
Excited to share our latest research on improving the safety of LLMs! We've developed DeTox, a tuning-free and noise robust alignment method that significantly reduces model toxicity without the need for large-scale preference data. 🚀 1/n
0
0
18
Very excited to get the data science award supported by @amfam with @Kaiping_Chen Look forward to developping new techniques to enhance multimodal foundation models!.
✨I am so excited to get the @datascience_uw award supported by the @amfam again, w/t @JunjieHu12! We will build Vision-Language Model for advancing risk assessment and communication in #AI! See our project and other amazing projects here👇:.
2
0
28
Are LLM agents really capable of role-playing different personas in "social interactions"? Short-turn conversations: maybe to some extent. Long-turn conversations: not yet! Their opinions converge, showing huge room for improvement! Check out our NAACL paper led by @SeanChuang4.
(1/9) LLM agents are increasingly used to simulate human-like societies. But how human-like are their “social interactions,” actually? For example, can we really use them to simulate human-like debates on issues like climate change? Our recent work was accepted at NAACL 2024!
0
0
9
RT @ChaoweiX: 🚨Making Backdoor for Good!! .We’re thrilled to share our new paper BackdoorAlign, where the idea of “Backdoor Attack” is appl….
0
28
0