Challenge: Existing methods for unlearning large language models fine-tune by maximizing loss, but they are unstable . this creates instability, especially on larger datasets, leading to over-unlearning .
Approach: They propose a novel unlearning method that leverages self-distillation to adjust logits . this method ensures smooth convergence and avoids catastrophic forgetting .
Outcome: The proposed method achieves smooth convergence and avoids catastrophic forgetting even on large datasets and sequential unlearning requests.

Similar Papers

Unilogit: Robust Machine Unlearning for LLMs Using Uniform-Target Self-Distillation (2025.findings-acl)

Copied to clipboard

Challenge: Extensive experiments on public benchmarks and an in-house e-commerce dataset demonstrate Unilogit’s superior performance in balancing forget and retain objectives, outperforming state-of-the-art methods such as NPO and UnDIAL.
Approach: They propose a self-distillation method that dynamically adjusts target logits to achieve a uniform probability for the target token.
Outcome: Extensive experiments on public benchmarks and an in-house e-commerce dataset demonstrate Unilogit’s superior performance in balancing forget and retain objectives.
Dissecting Fine-Tuning Unlearning in Large Language Models (2024.emnlp-main)

Copied to clipboard

Challenge: Existing methods for fine-tuning-based unlearning are ineffective at completely erasing model-embedded knowledge, but their true effectiveness remains unclear.
Approach: They propose to use activation patching and parameter restoration experiments to examine the limitations of fine-tuning-based unlearning methods for erasing harmful, sensitive, or copyrighted information within large language models.
Outcome: The proposed methods alter the model’s knowledge retrieval process rather than genuinely erasing the problematic knowledge embedded in the model parameters.
Distilling Robustness into Natural Language Inference Models with Domain-Targeted Augmentation (2024.findings-acl)

Copied to clipboard

Challenge: Knowledge distillation optimises a smaller student model to behave similarly to a larger teacher model, retaining some performance benefits.
Approach: They propose to augment the distillation with generated unlabelled examples that match the target distribution and upsamples data points among the training set that are similar to the target.
Outcome: The proposed method outperforms previous robustness solutions on the task of natural language inference (NLI) it also improves performance on OOD domains even beyond the target domain.
FIRST: Teach A Reliable Large Language Model Through Efficient Trustworthy Distillation (2024.emnlp-main)

Copied to clipboard

Challenge: Experimental results show that a well-calibrated model is more reliable than a fine-tuned model due to “tuning-induced mis-calibration”.
Approach: They propose a method which utilizes a small portion of teacher’s knowledge to obtain a reliable language model in a cost-efficient way.
Outcome: The proposed method reduces the computational burden by utilizing teacher's knowledge to obtain a reliable language model in a cost-efficient way.
Direct Token Optimization: A Self-Contained Approach to Large Language Model Unlearning (2026.findings-acl)

Copied to clipboard

Challenge: Existing methods for large language models (LLMs) rely on external resources such as auxiliary models, retain datasets, or even commercial AI services.
Approach: They propose a self-contained unlearning approach that optimizes the token-level objectives to unlearn specific sequences without external resources.
Outcome: The proposed approach improves the forget quality up to 16.8 over the latest benchmarks while maintaining comparable model utility.
Self-Distillation Bridges Distribution Gap in Language Model Fine-Tuning (2024.acl-long)

Copied to clipboard

Challenge: Experimental results show that fine-tuning of large language models for specific tasks can be challenging . distribution shift during fine-timing can lead to performance degradation in general task capabilities .
Approach: They propose a new approach that bridges the distribution gap between task datasets and LLMs by guiding fine-tuning with a distilled dataset generated by the model itself.
Outcome: The proposed approach achieves comparable or superior performance on downstream tasks compared to the vanilla approach.
Logits-Based Finetuning (2025.emnlp-main)

Copied to clipboard

Challenge: Existing methods for developing compact and efficient large language models lack token-level dependencies and linguistic diversity.
Approach: They propose a logits-based fine-tuning framework that integrates supervised learning and knowledge distillation to build enriched training targets using teacher logits and ground truth labels.
Outcome: The proposed method outperforms existing methods on a large-scale logits dataset and a series of science-focused models.
Robustness Challenges in Model Distillation and Pruning for Natural Language Understanding (2023.eacl-main)

Copied to clipboard

Challenge: Recent studies have focused on compressing pre-trained language models (PLMs) however, few studies have examined the impact of compression on generalizability and robustness of compressed models for out-of-distribution data.
Approach: They propose to use knowledge distillation and pruning to reduce model generalization and generalization on out-of-distribution data.
Outcome: The proposed compression techniques overfit on shortcut samples and generalize poorly on hard ones.
Sparse Logit Sampling: Accelerating Knowledge Distillation in LLMs (2025.acl-long)

Copied to clipboard

Challenge: Knowledge distillation is a cost-effective technique to distill knowledge in Large Language Models, if the teacher output logits can be pre-computed and cached.
Approach: They propose an importance-sampling-based method which provides unbiased estimates, preserves the gradient in expectation, and requires storing significantly sparser logits.
Outcome: The proposed method enables faster training of student models with marginal overhead (10%) compared to cross-entropy based training, while maintaining competitive performance compared with full distillation.
From Evasion to Concealment: Stealthy Knowledge Unlearning for LLMs (2025.findings-acl)

Copied to clipboard

Challenge: Existing approaches to unlearning often treat nonsensical responses or template-based refusals as the unlearning target, making the process even more vulnerable to attacks and jailbreaks.
Approach: They propose a method that uses inverted facts to remove the need for auxiliary models or retaining data while avoiding leakage.
Outcome: Evaluated on the ToFU Knowledge Unlearning dataset using Llama2-7B-Chat and Phi-1.5, MEOW outperforms baselines in forgetting quality while preserving model utility.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations