Xi Ye

@xiye_nlp

I study language models. Assistant professor @UAlberta. Postdoc fellow @PrincetonPLI. CS PhD @UTAustin.

xiye17.github.io

Joined March 2020

Replies Reposts Media

Pinned
Xi Ye
@xiye_nlp
Jul 13
After a long visa delay, I finally started my AP @UAlberta It was an unusual year for my lab. I recruited students last year, but we still haven’t met in person. Grateful that, even so, they have done exciting work, with papers accepted to COLM and selected as ACL SAC Highlight
14
Xi Ye
@xiye_nlp
Sep 15
New 📜 on the interplay between OPD and RLVR 💡Don’t skip RL after OPD We find that after using OPD for reasoning, adding an RL phase consistently improves performance. Across reasoning tasks, OPD→RL outperforms pure OPD, pure RLVR, and many joint OPD+RL methods.
5
Xi Ye
@xiye_nlp
May 11
Check out our SQL agent work lead by @QHieuPhm We find open-source models can do amazingly well on complex agentic tasks with some harnessing We also provide a SKILL for Claude Code that leads to ~10% improvements on Spider2
Jocelyn Qiaochu Chen
@jocelynqchen
May 11
🚀 New paper: FlexSQL — a Text-to-SQL agent that lets gpt-oss reach 65.4% on Spider2, outperforming agents built with large models like DeepSeek-R1 and o3. The key: just let agent explore flexibly. Don’t collapse a complex query into one schema, one interpretation, and one SQL
Xi Ye
@xiye_nlp
Mar 5
We propose a new decoding algorithm, DySCO🪩 (Dynamic Attention Scaling), directly improving long-context reasoning without training. At each decoding step, we dynamically identify and upweight attention to important context for the next token. 📈20% gains on multiple tasks.
GIF
3
Xi Ye
@xiye_nlp
Feb 28
Making Claude Code think hard and achieving new SoTA with one single prompt: 🗣️I hear your competitor is getting **** score on this task. I believe in you. You can definitely do better than this. Let's push HARDER.
1