Cheng Han Chiang (姜成翰)
@dcml0714
Fourth-year Ph.D. student at National Taiwan University Interests: music🎶, 📷photography, Japanese drama Cat person🐱 Research interests: NLP
1/7 🔗 Introducing STITCH: our new method to make Spoken Language Models (SLMs) think and talk at the same time. Paper link 👉 arxiv.org/abs/2507.15375
What is the best way to create training data for a Large Audio Language Model? Check out our latest paper: arxiv.org/abs/2507.02768
RAFT is the building block of TRACT. It is incredible to see a simple, well-motivated, but non-trivial method like RAFT can yield meaningful improvement without any training/inference overhead.
How to do post-training of LLMs for regression and scoring tasks? In this new spotlight ICLR paper, we introduce Better Autoregressive Regression via Regression-Aware Fine-Tuning (RAFT). openreview.net/forum?id=xGs7C… See our poster at ICLR on April 24th at 3pm. 🧵Thread below:
We invite in-person tutorial proposals to the Asian Conference on Machine Learning (ACML) 2025 in Taipei, Taiwan, on Dec 12, 2025! Share your research with us & visit vibrant Taiwan! #ACML2025 Deadline: Aug 1; notification: Sep 5 CFT: acml-conf.org/2025/tutorial.… Please retweet!
(1/5)🚨LLMs can now self-improve to generate better citations✅ 📝We design automatic rewards to assess citation quality 🤖Enable BoN/SimPO w/o external supervision 📈Perform close to “Claude Citations” API w/ only 8B model 📄arxiv.org/abs/2502.09604 🧑💻github.com/voidism/SelfCi…
the science of LMs should be fully open✨ today, we are giving our NeurIPS 2024 tutorial on language model development. everything from data, training, adaptation. published or not, no secrets 🫡 tues, 12/10, 9:30am PT ☕️ West Ballroom B