attention mechanisms 1certified token eviction 1exact deletion 1exact unlearning 1language model memory 1memory networks 1memory representation 1model editing 1online learning 1support vector machines 1support-vector memory 1
From the 2 of 2 linked papers with an AI index.
2 papers
cs.LG2026
Subtract, Transport, or Replay? Auditable Deletion from Language-Model Memory
Vishwajith Ramesh
The paper studies how to precisely delete specific records from pretrained language models by either subtracting addressable memory entries or replaying model states, demonstrating…
cs.LG2026
Forgetful Attention: An Auditable Support-Vector Memory for Selective Retention and Verified Deletion
Vishwajith Ramesh
The paper proposes Support Vector Attention, a max‑margin memory built on a one‑class SVM that can certify token removal and perform exact unlearning, and demonstrates its effectiv…