Jiahuan Pei

The registration page for #INLG2025 is now live! Join us in Vietnam at the Oct 29 - Nov 2 for the best conference on #NaturalLanguageGeneration 2025.inlgmeeting.org/registration.h… Curious to see what will be presented? Check out this list of accepted papers! 2025.inlgmeeting.org/accepted-paper…

#LargeReasoningModel Can prompt difficulty be predicted online to accelerate RL finetuning of Reasoning Models? YES! with MoPPS: ✅Predict success rates without costly evaluation 🎯Select informative prompts online ⚡Faster training 🚀Better performance 📄arxiv.org/abs/2507.04632

@AlbertW24045555 @overleaf Sync to GitHub or switch back to local compiler for latex :)

Overleaf is desk rejecting NeurIPS2025 submissions, not limited to professional subscribers🤣. @overleaf

@svakulenk0 Congrats 🎉

Honoured to receive prestigious WWTF personal grant of 1.5M Euro for establishing a new research group in beautiful Vienna. Our group will investigate and design novel approaches for modelling external knowledge sources for large language models. wwtf.at/funding/progra…

Unfortunately the current application link is not valid due to the change of management policy. But will be valid soon. Thank you for your kind understanding and I am happy to chat with the research proposal.

Hi everyone, We are hiring a new PhD candidate at VU Amsterdam. Are you interested in making AI Systems more trustworthy? Join our team and work with research related to “Trustworthy Conversational AI Agents for Mental Health”! workingat.vu.nl/vacancies/phd-…

Not Limited to #DeepSeek: Two years after finishing my PhD at VAE's birthplace, AMLab, I'm thrilled to share a VAE-inspired work to robustify meta-learning, domain randomization, and large models' PEFT😀. arxiv.org/pdf/2501.11039📜 Code available: github.com/thu-rllab/MPTS 🥳

🚨🚨🚨Tutorial Recording Alert!🚨🚨🚨 We just released the recording of our tutorial titled: Unbiased Learning to Rank: On Recent Advances in the Foundations and Applications youtu.be/fcqCkjgOnJ8 Originally presented at the #SIGIR2023, #FIRE2023 and #WSDM2024 conferences.

【20 July 2024】 MLLM Talk: 🍀 OmniH2O: Universal and Dexterous Human-to-Humanoid Whole-Body Teleoperation and Learning 📷Tairan He @TairanHe99 | Carnegie Mellon University tairanhe.com 📷More mllm-ai.com #LLM #Robot

🚀 Check out our latest interview with Konstantia Zarkogianni from Maastricht University on the #VOXReality Project! Read more! 👇 voxreality.eu/voxrealitys-pa… #XR #Innovation #AI #EthicsInAI #MaastrichtUniversity

How to enable a higher rank variation while preserving the computational advantage when using low-rank adapters for parameter-efficient fine-tuning? You can consider Mini-Ensemble LoRAs at #ACL2024. Congrats, @jayren3 @ChengshunS @zhaochun_ren @mdr arxiv.org/pdf/2402.17263…

🚀 Ready to revolutionise XR? Join the #VOXReality Open Call! 💡 Funding 200K EUR! Apply now: voxreality.eu/open-call/ #VOXReality #XR

Attention #XR enthusiasts 📢 Join our 2nd webinar Q&A for insights on the #VOXReality Open Call! 🗓 Date: May 21st, 2024 ⏰ Time: 11:00 - 12:00 CEST 💻 Online: Zoom Link sent to registered participants 📝 Register: f6s.com/voxreality-ope…

We are ready for our second edition of Spring School on Social XR at @CWInl 30 students will join this afternoon for 5days with lectures from 20 international speakers expert on #SocialXR

Can behavioral codes help clients and psychotherapists? See our recent paper accepted by @LrecColing. Congrats to Xin Sun, Jan de Wit, Mohammad Aliannejadi, Emiel Krahmer, Jos T.P. Dobber, and Jos A. Bosch! @ResearchGate: researchgate.net/publication/37…

I'm on the academic job market! I work on identifying and mitigating risks from LLM deployment. All my materials are on cloudygoose.github.io . Dear friends please help retweet, thanks!

🚀 New Paper Alert! #AIChallenge: How do we prevent "hallucination snowballing" in Large Language Models (LLMs)? And, how can we use verification results to enhance trustworthiness in AI-generated text? These are critical issues in #LLMs. 🧠 Introducing EVER (Real-time Verification and Rectification), which step-wise rectifies hallucinations during the generation process, not just at the end. Also, it flags potential unverifiable concepts with a user warning to increase the trustworthiness of the generated content. 🔍 EVER works in both non-retrieval (i.e., self-answering validation queries) and retrieval (i.e., utilizing external evidence) modes, Plus, it's a great complement to RAG. Led by @haoqik322 Paper: arxiv.org/abs/2311.09114 Code (coming soon): github.com/mk322/ever_hal…

🚀 Explore the power of predicted intents in open-domain dialogue systems with Enhancing Open-Domain Dialogue Answer Selection through Intent-Calibrated Self-Training by Jiahuan Pei from @cwi_dis Uncover the future of AI-driven conversations! 🤖 bit.ly/48wn7tp

Greetings from Amsterdam 👋, where the #VOXReality team gathered for the much-anticipated Third General Assembly hosted by CWI! 🌍✨Together, we’re shaping the future of Extended Reality! 🌌🔮

🚀 Exciting partnership alert! Meet the @cwi_dis, our XR innovation experts shaping the future of #VOXReality - They blend cutting-edge tech with human-focused design, ensuring immersive, intuitive experiences. Thrilled to redefine #XR together! 🌟🔧