Copertina

Guardrail Baselines for Unlearning in LLMs

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Thaker, Pratiksha, Maurya, Yash, Hu, Shengyuan, Wu, Zhiwei Steven, Smith, Virginia
Natura:	Preprint
Pubblicazione:	2024
Soggetti:	Computation and Language
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

Position: LLM Unlearning Benchmarks are Weak Measures of Progress
di: Thaker, Pratiksha, et al.
Pubblicazione: (2024)

BLUR: A Benchmark for LLM Unlearning Robust to Forget-Retain Overlap
di: Hu, Shengyuan, et al.
Pubblicazione: (2025)

SAEs $\textit{Can}$ Improve Unlearning: Dynamic Sparse Autoencoder Guardrails for Precision Unlearning in LLMs
di: Muhamed, Aashiq, et al.
Pubblicazione: (2025)

Unlearning or Obfuscating? Jogging the Memory of Unlearned LLMs via Benign Relearning
di: Hu, Shengyuan, et al.
Pubblicazione: (2024)

Membership Inference Attacks for Unseen Classes
di: Thaker, Pratiksha, et al.
Pubblicazione: (2025)

On the Benefits of Public Representations for Private Transfer Learning under Distribution Shift
di: Thaker, Pratiksha, et al.
Pubblicazione: (2023)

Enhancing One-run Privacy Auditing with Quantile Regression-Based Membership Inference
di: Liu, Terrance, et al.
Pubblicazione: (2025)

PARALLELPROMPT: Extracting Parallelism from Large Language Model Queries
di: Kolawole, Steven, et al.
Pubblicazione: (2025)

Unlearned but Not Forgotten: Data Extraction after Exact Unlearning in LLM
di: Wu, Xiaoyu, et al.
Pubblicazione: (2025)

UnStar: Unlearning with Self-Taught Anti-Sample Reasoning for LLMs
di: Sinha, Yash, et al.
Pubblicazione: (2024)

No Free Lunch in LLM Watermarking: Trade-offs in Watermarking Design Choices
di: Pang, Qi, et al.
Pubblicazione: (2024)

Generate-then-Verify: Reconstructing Data from Limited Published Statistics
di: Liu, Terrance, et al.
Pubblicazione: (2025)

Bypassing Safety Guardrails in LLMs Using Humor
di: Cisneros-Velarde, Pedro
Pubblicazione: (2025)

AI Governance and Accountability: An Analysis of Anthropic's Claude
di: Priyanshu, Aman, et al.
Pubblicazione: (2024)

LLM-Based Persuasion Enables Guardrail Override in Frontier LLMs
di: Nogueira, Rodrigo, et al.
Pubblicazione: (2026)

Is Safer Better? The Impact of Guardrails on the Argumentative Strength of LLMs in Hate Speech Countering
di: Bonaldi, Helena, et al.
Pubblicazione: (2024)

Gradient-Controlled Decoding: A Safety Guardrail for LLMs with Dual-Anchor Steering
di: Chiniya, Purva, et al.
Pubblicazione: (2026)

Learning-Time Encoding Shapes Unlearning in LLMs
di: Wu, Ruihan, et al.
Pubblicazione: (2025)

Unlearning Isn't Deletion: Investigating Reversibility of Machine Unlearning in LLMs
di: Xu, Xiaoyu, et al.
Pubblicazione: (2025)

Knowledge Distillation with Structured Chain-of-Thought for Text-to-SQL
di: Thaker, Khushboo, et al.
Pubblicazione: (2025)

Leverage Unlearning to Sanitize LLMs
di: Boutet, Antoine, et al.
Pubblicazione: (2025)

Trust-Oriented Adaptive Guardrails for Large Language Models
di: Hu, Jinwei, et al.
Pubblicazione: (2024)

Eight Methods to Evaluate Robust Unlearning in LLMs
di: Lynch, Aengus, et al.
Pubblicazione: (2024)

Building Guardrails for Large Language Models
di: Dong, Yi, et al.
Pubblicazione: (2024)

Unlearning in LLMs: Methods, Evaluation, and Open Challenges
di: Lizzo, Tyler, et al.
Pubblicazione: (2026)

Semantic Agreement Enables Efficient Open-Ended LLM Cascades
di: Soiffer, Duncan, et al.
Pubblicazione: (2025)

Intrinsic Guardrails: How Semantic Geometry of Personality Interacts with Emergent Misalignment in LLMs
di: Aneja, Krishak, et al.
Pubblicazione: (2026)

Persona-Augmented Benchmarking: Evaluating LLMs Across Diverse Writing Styles
di: Truong, Kimberly Le, et al.
Pubblicazione: (2025)

SEPS: A Separability Measure for Robust Unlearning in LLMs
di: Jeung, Wonje, et al.
Pubblicazione: (2025)

Uncovering the Potential Risks in Unlearning: Danger of English-only Unlearning in Multilingual LLMs
di: Hwang, Kyomin, et al.
Pubblicazione: (2025)

OpenGuardrails: A Configurable, Unified, and Scalable Guardrails Platform for Large Language Models
di: Wang, Thomas, et al.
Pubblicazione: (2025)

RAG Makes Guardrails Unsafe? Investigating Robustness of Guardrails under RAG-style Contexts
di: She, Yining, et al.
Pubblicazione: (2025)

Position: Enough of Scaling LLMs! Lets Focus on Downscaling
di: Goel, Yash, et al.
Pubblicazione: (2025)

Benchmarking LLM Guardrails in Handling Multilingual Toxicity
di: Yang, Yahan, et al.
Pubblicazione: (2024)

Triaging Threats to Specialized Guardrails
di: Mo, Wenjie Jacky, et al.
Pubblicazione: (2026)

Unlearning What Matters: Token-Level Attribution for Precise Language Model Unlearning
di: Wu, Jiawei, et al.
Pubblicazione: (2026)

Catastrophic Failure of LLM Unlearning via Quantization
di: Zhang, Zhiwei, et al.
Pubblicazione: (2024)

Everybody Prune Now: Structured Pruning of LLMs with only Forward Passes
di: Kolawole, Steven, et al.
Pubblicazione: (2024)

Rectifying Belief Space via Unlearning to Harness LLMs' Reasoning
di: Niwa, Ayana, et al.
Pubblicazione: (2025)

Less is More: Geometric Unlearning for LLMs with Minimal Data Disclosure
di: Tan, Chenchen, et al.
Pubblicazione: (2026)