AI & ML interests

Data-Centric ML

Recent Activity

s-kangasl  new activity about 2 months ago

Harvard-DCML/boomerang-qwen3-4.9B:Substantially lower accuracy on reasoning benchmarks such as GSM8K (1.5%) and MATH-500 (4.2%)

nihalnayak  authored a paper 2 months ago

Revisiting Generalization Across Difficulty Levels: It's Not So Easy

jgeuter  updated a model 3 months ago

Harvard-DCML/boomerang-qwen3-2.3B

View all activity

Papers

Boomerang Distillation Enables Zero-Shot Model Size Interpolation

View all Papers

Team members 6