PROJECT

PruneForge: Zero-Training LLM Pruning for Agents

PruneForge searches for a task-specific, smaller language model by removing MLP channels, attention heads and layers while checking output similarity against a small prompt set. The pruning loop uses no fine-tuning and saves an inspectable pruning recipe.

Plot of semantic similarity and model compression during a PruneForge pruning run.
PROJECTS UPDATES