"I can fit my mother to a line on a log-log plot" -Herbert Kroemer, Nobel Laureate There's a lot of truth to this. log-log plots can be very misleading 1/5

*spred: Solving L1 Penalty with SGD* by Liu Ziyin Zihao Wang Having worked in NN sparsity, this is really nice! A simple reparameterization (2x params) turns L1 reg into L2 reg and makes it easier to optimize. arxiv.org/abs/2210.01212

IJCAIconf It's my real honor to give a talk IJCAIconf FL workshop about privacy problems in LLMs, based on work mostly done by my student Haoran Li Haoran LI in our group KnowComp HKUST. Slides are available at home.cse.ust.hk/~yqsong/papers… #IJCAI23 #IJCAI2023 #LLMs

Gathering at IJCAI 2023 and enjoying a delicious buffet dinner #IJCAI2023

#新头像

The most concise poster in this NeurIPS #NeurIPS2023

Check out our new WWW 2024 Tutorial on New Frontiers of Knowledge Graph Reasoning: Recent Advances and Future Trends where we have discussed many interesting new things happening in the field of knowledge graphs #webconf #knowledgegraph #KG The Web Conference Zihao Wang Jiaxin Bai

🚀 Thrilled to announce our paper "Rethinking the Bounds of LLM Reasoning: Are Multi-Agent Discussions the Key?" has been accepted at #ACL2024! 🎉 We explored whether multi-agent discussions truly outperform single agents in LLM reasoning, an area largely unexamined.