Unverified50% confidenceFactExact time
Gradient-based machine unlearning methods typically achieve unlearning by optimizing a forget loss and a retain loss, but similar loss values do not necessarily imply similar output-space behavior.
1
Sources
50%
Confidence
Long-term
Relevance
10/30/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThis study reveals a fundamental disconnect between loss and behavior in gradient-based machine unlearning and provides a unified output-space analysis framework.78% similarUnverifiedWhen unlearning objectives designed for multi-step models are applied to few-step distilled (FSD) models, the unlearning effectiveness is significantly diminished.74% similarUnverifiedThe logistic loss does not enjoy the property of polynomial losses whereby a finite-dimensional representation suffices for exact continuation of learning; exactly preserving learning capacity theoretically requires unbounded memory overhead.73% similarUnverifiedThe two core metrics for measuring machine unlearning quality are Forget Rate and Retain Rate; these two metrics are in natural tension, as excessive forgetting can lead to concept collapse.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1113840API
curl https://kongchang.com/api/v1/knowledge/claims/1113840MCP
get_claim(id=1113840)