COLM Efficient Reasoning After party with Dinner
COLM ticket or invitation required. Stay hungry and join us for a night of food, drinks. 10 min walk from the conference.
Cohost with Hyperbolic. Lottery provided by Phala.com.
Our Friends
Songlin Yang — Technical Lead, Pretraining Science at Thinking Machines Lab; MIT PhD. First author of Gated Delta Networks.
Beidi Chen — Assistant Professor at CMU, Head of InfiniAI Lab, and Amazon Scholar. Creator of FlexGen, H2O, and Deja Vu.
Jiawei Zhao — Research Scientist at Meta Superintelligence Labs, leading an Optimization & Reasoning team. Creator of GaLore and DeepConf.
Zhijian Liu — Assistant Professor at UCSD and NVIDIA Research Scientist; MIT PhD. Known for DFlash, ParoQuant, and NVILA.
Zhen Dong — NVIDIA Senior/Staff Research Scientist and incoming UCSB Assistant Professor. Known for HAWQ, SqueezeLLM, and Nemotron.
Weijia Shi — Researcher at Ai2 and University of Washington. Co-first author of s1 and first author of FlexOlmo (NeurIPS 2025 Spotlight).
Kaitao Song — Senior Research Scientist at Apple, formerly Microsoft. Co-author of HuggingGPT and MPNet.
Cheng Luo — Founder of the Efficient Reasoning workshop series, formerly Microsoft Research. Research on efficient LLM training and inference.
Yuchen Zhu — Georgia Tech ML researcher. Co-first author of DMPO (ICML 2026 Spotlight), focusing on diffusion language models and coding agents.
Zefan Cai — UW–Madison researcher, known for PyramidKV and R-KV, focusing on long-context inference and KV cache optimization.
Rui Jie Zhu — UCSC researcher who led Ouro's 7.7-trillion-token pretraining process, with contributions to RWKV and SpikeGPT.
Jiahao Yu — Assistant Professor at NYU Abu Dhabi and PI of the Secure Reasoning Lab. Known for GPTFuzzer and research in AI security.
About Hyperbolic:
Hyperbolic is the AI infrastructure platform that turns fragmented, heterogeneous bare-metal compute into a unified, productive, liquid AI cloud. We give AI companies a single platform to buy, sell, manage, and orchestrate compute at scale, while helping neoclouds bring capacity to market faster, increase utilization, and maximize yields on their infrastructure.