News › Computer science & data
Computer science & dataKAIST: CluSTER cuts LLM fine-tuning time by up to 69.6% with nearly the same quality
Professor Jae-Gil Lee's team at KAIST proposes CluSTER, which groups and selects data with similar learning effect and spreads it evenly over multiple GPUs to reduce repeated learning. On ~7B and 13B models, three coding/medical datasets and eight benchmarks, models kept nearly the same performance with only about 31% of the data while training time dropped by up to 69.6% (10 hours to about 3). Experiments were mainly on 4 GPUs; presented at EMNLP 2026.
Source: KAIST
Need a custom AI tool?
Describe your need briefly; you will get clarifying questions and a scoped proposal before any work starts.
Related news
Computer science & data · 2026-11-03
NSF selects FabAID: a $24.5 million national data fabric for AI-driven science
Computer science & data · 2026-11-02
UCL: the "REM" AI model reconstructs entire particle collisions, efficiency 61% to 92% (in simulation)
Computer systems & chips · 2026-11-04