| Nov 03, 2026 | π Our new paper, βFINDER: Studying and Repairing What Test-Time Learning Changes in Deep-Research Agents,β has been accepted to WSDM 2027! The arXiv preprint is coming soon. See you in HK! |
| Oct 26, 2026 | π Our survey Learning from Use: Test-Time Learning in Large Language Models and Agents was accepted by Nature Reviews Computing! |
| Sep 29, 2026 | π New preprint: OSWorld-Science: A Benchmark of Computer Use Agents for Learning and Using Scientific Software is now available on arXiv! Please check out the project page for more details! |
| Aug 20, 2026 | π New survey: Learning from Use: Test-Time Learning in Large Language Models and Agents is now available! |
| Apr 30, 2026 | π One paper accepted to ICML 2026! |
| Apr 06, 2026 | π One paper accepted to ACL 2026! |
| Mar 02, 2026 | π One paper accepted to ICLR 2026 Workshop! |
| Oct 17, 2025 | π New preprint: TokDrift: When LLM Speaks in Subwords but Code Speaks in Grammar is now available on arXiv! 1οΈβ£ LLMsβ subword tokenizers donβt align well with programming language grammar: tiny whitespace or renaming tweaks -> different tokenization -> flipped outputs. 2οΈβ£ Our framework TokDrift systematically tests 9 code LLMs on 3 tasks, showing their sensitivity to tokenization changes: up to 60% outputs change under a single semantic-preserving rewrite. 3οΈβ£ If your win margin is ~1 pp, beware: spacing & naming can swing results. |
| Feb 15, 2025 | Homepage Acknowledgements |
| Feb 15, 2025 | Hello my personal website! Letβs make a brithday for itπ. Glad to see you there but it was still under construction. Hopefully it will be done soon. |