Image de couverture de livre

SHRED: Retain-Set-Free Unlearning via Self-Distillation with Logit Demotion

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Hu, Zizhao, Godbole, Ameya, Wei, Johnny Tian-Zheng, Rostami, Mohammad, Thomason, Jesse, Jia, Robin
Format:	Preprint
Publié:	2026
Sujets:	Machine Learning Artificial Intelligence
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Expert Personas Improve LLM Alignment but Damage Accuracy: Bootstrapping Intent-Based Persona Routing with PRISM
par: Hu, Zizhao, et autres
Publié: (2026)

Multi-modal Synthetic Data Training and Model Collapse: Insights from VLMs and Diffusion Models
par: Hu, Zizhao, et autres
Publié: (2025)

Spiking the training data to correct for test set contamination
par: Wei, Johnny Tian-Zheng, et autres
Publié: (2026)

TNG-CLIP:Training-Time Negation Data Generation for Negation Awareness of CLIP
par: Cai, Yuliang, et autres
Publié: (2025)

An Intermediate Fusion ViT Enables Efficient Text-Image Alignment in Diffusion Models
par: Hu, Zizhao, et autres
Publié: (2024)

Verify with Caution: The Pitfalls of Relying on Imperfect Factuality Metrics
par: Godbole, Ameya, et autres
Publié: (2025)

SCENE: Self-Labeled Counterfactuals for Extrapolating to Negative Examples
par: Fu, Deqing, et autres
Publié: (2023)

Interrogating LLM design under a fair learning doctrine
par: Wei, Johnny Tian-Zheng, et autres
Publié: (2025)

Reversing the Forget-Retain Objectives: An Efficient LLM Unlearning Framework from Logit Difference
par: Ji, Jiabao, et autres
Publié: (2024)

UNDIAL: Self-Distillation with Adjusted Logits for Robust Unlearning in Large Language Models
par: Dong, Yijiang River, et autres
Publié: (2024)

Language Models can Infer Action Semantics for Symbolic Planners from Environment Feedback
par: Zhu, Wang, et autres
Publié: (2024)

Learning to Deliberate: Meta-policy Collaboration for Agentic LLMs with Multi-agent Reinforcement Learning
par: Yang, Wei, et autres
Publié: (2025)

Why Do Some Inputs Break Low-Bit LLM Quantization?
par: Chang, Ting-Yun, et autres
Publié: (2025)

Lateralization MLP: A Simple Brain-inspired Architecture for Diffusion
par: Hu, Zizhao, et autres
Publié: (2024)

Harmonizing Multi-Objective LLM Unlearning via Unified Domain Representation and Bidirectional Logit Distillation
par: Zhong, Yisheng, et autres
Publié: (2026)

PDDL-Mind: Large Language Models are Capable on Belief Reasoning with Reliable State Tracking
par: Zhu, Wang Bill, et autres
Publié: (2026)

TokenSmith: Streamlining Data Editing, Search, and Inspection for Large-Scale Language Model Training and Interpretability
par: Khan, Mohammad Aflah, et autres
Publié: (2025)

Is Retain Set All You Need in Machine Unlearning? Restoring Performance of Unlearned Models with Out-Of-Distribution Images
par: Bonato, Jacopo, et autres
Publié: (2024)

Adjust for Trust: Mitigating Trust-Induced Inappropriate Reliance on AI Assistance
par: Srinivasan, Tejas, et autres
Publié: (2025)

BLUR: A Benchmark for LLM Unlearning Robust to Forget-Retain Overlap
par: Hu, Shengyuan, et autres
Publié: (2025)

Static Key Attention in Vision
par: Hu, Zizhao, et autres
Publié: (2024)

Peak-Controlled Logits Poisoning Attack in Federated Distillation
par: Tang, Yuhan, et autres
Publié: (2024)

Federated Progressive Self-Distillation with Logits Calibration for Personalized IIoT Edge Intelligence
par: Wang, Yingchao, et autres
Publié: (2024)

Demoting Security via Exploitation of Cache Demote Operation in Intel's Latest ISA Extension
par: Kim, Taehun, et autres
Publié: (2025)

M3PT: A Transformer for Multimodal, Multi-Party Social Signal Prediction with Person-aware Blockwise Attention
par: Tang, Yiming, et autres
Publié: (2025)

Logit Distillation on Manifolds: Mapping by Learning
par: Yang, Yiru, et autres
Publié: (2026)

The statistical advantage of automatic NLG metrics at the system level
par: Wei, Johnny Tian-Zheng, et autres
Publié: (2021)

Do Localization Methods Actually Localize Memorized Data in LLMs? A Tale of Two Benchmarks
par: Chang, Ting-Yun, et autres
Publié: (2023)

When Parts Are Greater Than Sums: Individual LLM Components Can Outperform Full Models
par: Chang, Ting-Yun, et autres
Publié: (2024)

Efficient End-to-End Visual Document Understanding with Rationale Distillation
par: Zhu, Wang, et autres
Publié: (2023)

CS-SHRED: Enhancing SHRED for Robust Recovery of Spatiotemporal Dynamics
par: da Silva, Romulo B., et autres
Publié: (2025)

OGLS-SD: On-Policy Self-Distillation with Outcome-Guided Logit Steering for LLM Reasoning
par: Yang, Yuxiao, et autres
Publié: (2026)

Words that make SENSE: Sensorimotor Norms in Learned Lexical Token Representations
par: Gupta, Abhinav, et autres
Publié: (2026)

Hubble: a Model Suite to Advance the Study of LLM Memorization
par: Wei, Johnny Tian-Zheng, et autres
Publié: (2025)

Maestro: Learning to Collaborate via Conditional Listwise Policy Optimization for Multi-Agent LLMs
par: Yang, Wei, et autres
Publié: (2025)

Distillation Robustifies Unlearning
par: Lee, Bruce W., et autres
Publié: (2025)

Synthetic Data for Robust AI Model Development in Regulated Enterprises
par: Godbole, Aditi
Publié: (2025)

Corrective Machine Unlearning
par: Goel, Shashwat, et autres
Publié: (2024)

Iterative Formalization and Planning in Partially Observable Environments
par: Gong, Liancheng, et autres
Publié: (2025)

Proving membership in LLM pretraining data via data watermarks
par: Wei, Johnny Tian-Zheng, et autres
Publié: (2024)