aiqbench analysis
The paper reframes LLM unlearning around "knowledge associations," the conditional predictive distribution linking a query context to its target knowledge, and uses reinforcement learning with token-level rewards and proximity-aware retain sampling to suppress specific associations. The reported gains are relative to existing baselines on a range of models and benchmarks, but the abstract doesn't quantify them, so the practical advantage needs checking against the full results. For readers working on privacy, copyright, or model editing, the useful question is whether this granularity holds up under adversarial probing and whether retained capabilities stay intact at scale.
What Should LLMs Forget: Reinforcement Unlearning of Knowledge Associations
arXiv:2610.41616v1 Announce Type: new Abstract: Large language models (LLMs) acquire extensive factual knowledge during pre-training, which may include sensitive, private, or copyrighted information, motivating the development of unlearning methods that selectively remove specific knowledge while preserving general capabilities. However, existing approaches typically optimize entire responses without explicitly identifying the knowledge to be forgotten, resulting in coarse and uncontrolled forgetting. To address this, we revisit the forgetting target of LLM unlearning and introduce Knowledge Association (KA), which characterizes the association between a query context and its target knowledge and is instantiated by the corresponding conditional predictive distribution, as the fundamental target of forgetting. Based on KA, we propose Knowledge Association Reinforcement Unlearning (KARU), a unified reinforcement learning framework that suppresses forgotten knowledge associations while preserving retained ones. Specifically, KARU derives complementary mechanisms from the KA-based formulation: token-level reward construction for finer-grained forgetting supervision, distribution-level regularization for precise knowledge suppression, and proximity-aware retain sampling to mitigate collateral degradation. Experiments across diverse LLMs on representative benchmarks demonstrate that KARU consistently improves forgetting effectiveness over existing baselines while maintaining comparable retention performance and general utility.
Abstract:Large language models (LLMs) acquire extensive factual knowledge during pre-training, which may include sensitive, private, or copyrighted information, motivating the development of unlearning methods that selectively remove specific knowledge while preserving general capabilities. However, existing approaches typically optimize entire responses without explicitly identifying the knowledge to be forgotten, resulting in coarse and uncontrolled forgetting. To address this, we revisit the forgetting target of LLM unlearning and introduce Knowledge Association (KA), which characterizes the association between a query context and its target knowledge and is instantiated by the corresponding conditional predictive distribution, as the fundamental target of forgetting. Based on KA, we propose Knowledge Association Reinforcement Unlearning (KARU), a unified reinforcement learning framework that suppresses forgotten knowledge associations while preserving retained ones. Specifically, KARU derives complementary mechanisms from the KA-based formulation: token-level reward construction for finer-grained forgetting supervision, distribution-level regularization for precise knowledge suppression, and proximity-aware retain sampling to mitigate collateral degradation. Experiments across diverse LLMs on representative benchmarks demonstrate that KARU consistently improves forgetting effectiveness over existing baselines while maintaining comparable retention performance and general utility.
Artificial Intelligence (cs.AI)
arXiv:2610.41616 [cs.AI]
(or arXiv:2610.41616v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2610.41616 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 270 words
Related coverage
All-in-One: A Composable Attack on LLM Transformations
arXiv · Oct 27, 2026
Scaling Reinforcement Learning for Diffusion Models via Velocity Matching
arXiv · Aug 26, 2026
Learning New Facts with QLoRA: An Acquisition-Retention Frontier
arXiv · Aug 27, 2026
Agentic Knowledge-Model Co-Evolution via Consistency Regularization
arXiv · Oct 23, 2026