MetaMerge: Model Merging with Meta Networks

Merging ViTs and LLMs using a pretrained (graph) neural net.

15 min read   ·   2026

REAM: Compressing Mixture-of-Experts LLMs

Merging experts in Mixture-of-Experts (MoE) LLMs to compress a 235B LLM.

35 min read   ·   2026

NiNo: Learning to Accelerate Training of Neural Networks

Explaining our ICLR 2025 paper and visualizing neuron permutation symmetry.

15 min read   ·   2025