AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Hallucination-Guided Unlearning: Using Hallucination Traces to Reveal Overfitted Memories

Jiayu Zhang, Ziqi Zhong, Yuliang Gai, Canran Xiao

NeurIPS 2026 · Large language models › Language, safety and interpretability · approach: LLM pruning and scaling data · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes Hallucination-Guided Unlearning (HGU), which uses the model's own hallucinations as signals to identify overfitted or memorized content to forget in LLMs.

Problem

Selective unlearning in LLMs needs a forget set, but harmful or memorized content is latent and entangled and hard to specify explicitly.

Approach

Treats hallucination traces produced by the model as a signal for what to forget, identifying overfitted memories from the model's own failure modes rather than relying solely on explicit supervision.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
trace hallucinated spans to sparse MLP neurons and gate them under a KL utility constraintLLM unlearning · TOFU forget05Avg (MU, FE): 0.6433 -> 0.7221 vs best baseline RapidUnImproved
use HGU under mixed forget/retain promptsLLM unlearning · TOFU SepS forget05H-Avg: 0.3140 -> 0.4382 vs IDK+APImproved
apply HGU on WMDP auxiliary EconomicsZephyr-7B · WMDP auxiliary EconomicsForget accuracy: 16.8 -> 14.2 vs RapidUnImproved
raise keep-KL weight lambda_KL (ablation)HGU · TOFU / WMDPforgetting vs utility driftMixed
restrict intervention to traced channel set C (ablation)HGU · hallucination reduction vs utilitysignificance of hallucination reductionImproved
HGU vs GA+GD on TOFU forget10 averageHGU · TOFU forget10FE: 0.9346 -> 0.8915Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy