cd/entity/SimNPO· home› entities› SimNPO
grep -l @simnpo /news/*.json | wc -l → 2

SimNPO

mentions 2 type Organization feed RSS

// recent coverage 2 mentions

22:11
2026-07-20
lesswrong.com
ai-safety

Does routine compression undo LLM unlearning? A short project

A two-week project by a BlueDot Project cohort participant found that standard post-training compression processes (quantization, magnitude pruning, SVD truncation) minimally reverse unlearning on the…

04:00
2026-06-17
arxiv.org
large-language-models

RepSelect: Robust LLM Unlearning via Representation Selectivity

Researchers propose RepSelect, a method for robust LLM unlearning that isolates forget-set-specific representations by collapsing top principal components of weight gradients, achieving 4-50x larger r…

// co-occurs with top 8 entities
NPO 2 RepSelect 1 Llama 3 1 Qwen 3.5 1 Gemma 4 E4B 1 DeepSeek V2 Lite 1 GradDiff 1 BlueDot Project 1
// topics top 4 topics
large language models 2 ai safety 2 machine learning 1 ai research 1