| Oct 26, 2026 | I was invited by Prof. Jared Tanner at the Mathematical Institute, University of Oxford to give a talk on “Understanding Transformers in LLMs, Token Distinguishability, Learnability, and Safety Alignment”. |
| Sep 25, 2026 | One paper accepted at NeurIPS 2026! ZetaEvolve: Learning to Search through History-Conditioned Potential Value explores self-evolving agents and recursive self-improvement. |
| May 01, 2026 | 3 ICML papers accepted, with one Spotlight! 😉 Two of them are about the fundamental understanding of the Transformer architecture and its training dynamics. The other is about private training of SGD with adaptive clipping strategies. |
| Jan 29, 2026 | 4 papers accepted by ICLR 2026 😉! Two are for highly skillful AI models for simulation deformable objects, one for mixture of depth-recurrrent architecture design for improving LLM resasoning, and the other for analyzing training dynamics of semi-supervised learning. |