REVIEW 11 cited by
Catastrophic Failure of LLM Unlearning via Quantization
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Signed reviews
read the original abstract
Large language models (LLMs) have shown remarkable proficiency in generating text, benefiting from extensive training on vast textual corpora. However, LLMs may also acquire unwanted behaviors from the diverse and sensitive nature of their training data, which can include copyrighted and private content. Machine unlearning has been introduced as a viable solution to remove the influence of such problematic content without the need for costly and time-consuming retraining. This process aims to erase specific knowledge from LLMs while preserving as much model utility as possible. Despite the effectiveness of current unlearning methods, little attention has been given to whether existing unlearning methods for LLMs truly achieve forgetting or merely hide the knowledge, which current unlearning benchmarks fail to detect. This paper reveals that applying quantization to models that have undergone unlearning can restore the "forgotten" information. To thoroughly evaluate this phenomenon, we conduct comprehensive experiments using various quantization techniques across multiple precision levels. We find that for unlearning methods with utility constraints, the unlearned model retains an average of 21\% of the intended forgotten knowledge in full precision, which significantly increases to 83\% after 4-bit quantization. ... Our code is available at: \href{https://github.com/zzwjames/FailureLLMUnlearning}{https://github.com/zzwjames/FailureLLMUnlearning}.
Forward citations
Cited by 11 Pith papers
-
Bits and Memories: Measuring Verbatim Extraction Across LLM Quantization
Quantizing LLMs selectively forgets memorized text faster than capability, but 1B-scale 4-bit models still extract ~72% of memorized sequences, so quantization is not a privacy defense.
-
One Modality to Forget Them All: Enhancing Cross-Modal Unlearning in Vision-Language Models
Cross-modal unlearning transfer in vision-language models is asymmetric, architecture-dependent, and shallow under typographic attacks; influence-guided block selection reduces the measured gap.
-
SoftWater: Class-Aware Rate Allocation for Softmax Quantization
SoftWater, a KL-divergence-based quantizer for LLM softmax heads, allocates bit rate by class frequency and variance and beats WaterSIC at matched head rates on 59 of 60 test points.
-
Towards Evaluation for Real-World LLM Unlearning
DCUE evaluates LLM unlearning by comparing core-token confidence score distributions of the unlearned model and the original model, corrected by a validation set, using the Kolmogorov-Smirnov test.
-
Capturing the Effects of Quantization on Trojans in Code LLMs
At 4-bit inference quantization, CodeLlama-7b generates more accurate SQL and activates a planted backdoor far less often, while Llama-2-7b is hardly affected, across two training seeds.
-
A mean teacher algorithm for unlearning of language models
A mean teacher optimizer that approximates slow natural gradient descent, paired with a new negative log-unlikelihood loss, reduces memorization and privacy leakage on MUSE-News and MUSE-Books, with the strongest vari...
-
Model Tampering Attacks Enable More Rigorous Evaluations of LLM Capabilities
Model tampering attacks, especially few-shot fine-tuning, reliably re-elicit unlearned capabilities in Llama-3-8B and can bound the success of held-out input-space attacks.
-
Huff-LLM: End-to-End Lossless Compression for Efficient LLM Inference
Huff-LLM splits FP16/BF16 LLM weights into small bit groups, Huffman-compresses each group, and uses custom hardware decoders so weights stay compressed through the memory hierarchy during inference.
-
Module-Aware Parameter-Efficient Machine Unlearning on Transformers
MAPE-Unlearn uses Fisher-information-based scores and greedy search to select important heads and filters, then applies sparse unlearning updates, claiming improved efficacy-fidelity trade-offs on Transformers.
-
SoK: Machine Unlearning for Large Language Models
A new taxonomy for LLM unlearning distinguishes removal-intended from suppression-intended methods, and argues that gradient ascent methods functionally behave like suppression.
-
Revealing the Deceptiveness of Knowledge Editing: A Mechanistic Analysis of Superficial Editing
Edits made by ROME, MEMIT, PMET and other methods often leave the original fact intact inside later-layer attention heads, so a contextual prompt makes the edited model fall back to the old answer.
Discussion (0). Continue with ORCID to comment.