conftrace_

downstream fine-tuning

3 papers

Explore in graph

Co-occurring keywords · papers mentioning both

unlearning (1) hazardous knowledge (1) invariant risk minimization (irm) (1) large language model (1) parameter efficiency (1) compute efficiency (1) scaling law (1) riemannian geometry (1) invariance (1) asymptotic analysis (1)

Papers

On the Asymptotics of Self-Supervised Pre-training: Two-Stage M-Estimation and Representation Symmetry COLT 2026 Invariance Makes LLM Unlearning Resilient Even to Unanticipated Downstream Fine-Tuning ICML 2025 Scale Efficiently: Insights from Pretraining and Finetuning Transformers ICLR 2022