# BalDRO: A Distributionally Robust Optimization based Framework for Large Language Model Unlearning

arXiv: https://arxiv.org/abs/2601.09172 · HTML: https://mlsanity.com/paper/2601.09172/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Placed in Machine unlearning, one of 12 directions in LLM safety and security.
- Builds on 31 of its 66 references that are in the map.
- First posted on arXiv 14 Jan 2026.

## Placed in

- LLM safety and security / [Machine unlearning](https://mlsanity.com/safety/machine-unlearning/)

## Builds on (most cited first, 30 of 31)

- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Extracting Training Data from Large Language Models](https://mlsanity.com/paper/2012.07805/) (2020)
- [TOFU: A Task of Fictitious Unlearning for LLMs](https://mlsanity.com/paper/2401.06121/) (2024)
- [Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearning](https://mlsanity.com/paper/2404.05868/) (2024)
- [Safety Alignment Should Be Made More Than Just a Few Tokens Deep](https://mlsanity.com/paper/2406.05946/) (2024)
- [Large Language Model Unlearning](https://mlsanity.com/paper/2310.10683/) (2023)
- [MUSE: Machine Unlearning Six-Way Evaluation for Language Models](https://mlsanity.com/paper/2407.06460/) (2024)
- [Simplicity Prevails: Rethinking Negative Preference Optimization for LLM Unlearning](https://mlsanity.com/paper/2410.07163/) (2024)
- [ST-WebAgentBench: A Benchmark for Evaluating Safety and Trustworthiness in Web Agents](https://mlsanity.com/paper/2410.06703/) (2024)
- [Large Language Model Unlearning via Embedding-Corrupted Prompts](https://mlsanity.com/paper/2406.07933/) (2024)
- [Advancing Pose-Guided Image Synthesis with Progressive Conditional Diffusion Models](https://mlsanity.com/paper/2310.06313/) (2023)
- [SOUL: Unlocking the Power of Second-Order Optimization for LLM Unlearning](https://mlsanity.com/paper/2404.18239/) (2024)
- [What makes unlearning hard and what to do about it](https://mlsanity.com/paper/2406.01257/) (2024)
- [LLM Unlearning via Loss Adjustment with Only Forget Data](https://mlsanity.com/paper/2410.11143/) (2024)
- [Exploring Criteria of Loss Reweighting to Enhance LLM Unlearning](https://mlsanity.com/paper/2505.11953/) (2025)
- [Finding and Editing Multi-Modal Neurons in Pre-Trained Transformers](https://mlsanity.com/paper/2311.07470/) (2023)
- [Alternate Preference Optimization for Unlearning Factual Knowledge in Large Language Models](https://mlsanity.com/paper/2409.13474/) (2024)
- [SafeEraser: Enhancing Safety in Multimodal Large Language Models through Multimodal Machine Unlearning](https://mlsanity.com/paper/2502.12520/) (2025)
- [Separable Multi-Concept Erasure from Diffusion Models](https://mlsanity.com/paper/2402.05947/) (2024)
- [We Should Identify and Mitigate Third-Party Safety Risks in MCP-Powered Agent Systems](https://mlsanity.com/paper/2506.13666/) (2025)
- [Learning from Heterogeneity: Generalizing Dynamic Facial Expression Recognition via Distributionally Robust Optimization](https://mlsanity.com/paper/2507.15765/) (2025)
- [Precise Localization of Memories: A Fine-grained Neuron-level Knowledge Editing Technique for LLMs](https://mlsanity.com/paper/2503.01090/) (2025)
- [Editing as Unlearning: Are Knowledge Editing Methods Strong Baselines for Large Language Model Unlearning?](https://mlsanity.com/paper/2505.19855/) (2025)
- [AgentSociety Challenge: Designing LLM Agents for User Modeling and Recommendation on Web Platforms](https://mlsanity.com/paper/2502.18754/) (2025)
- [Not All Data Are Unlearned Equally](https://mlsanity.com/paper/2504.05058/) (2025)
- [Maximizing Local Entropy Where It Matters: Prefix-Aware Localized LLM Unlearning](https://mlsanity.com/paper/2601.03190/) (2026)
- [AdvAnchor: Enhancing Diffusion Model Unlearning with Adversarial Anchors](https://mlsanity.com/paper/2501.00054/) (2024)
- [Temporally and Distributionally Robust Optimization for Cold-Start Recommendation](https://mlsanity.com/paper/2312.09901/) (2023)
- [Multi-Value Alignment for LLMs via Value Decorrelation and Extrapolation](https://mlsanity.com/paper/2511.17579/) (2025)
- [DUSK: Do Not Unlearn Shared Knowledge](https://mlsanity.com/paper/2505.15209/) (2025)

## Built on by (most cited first, 8 of 8)

- [PASs-MoE: Mitigating Misaligned Co-drift among Router and Experts via Pathway Activation Subspaces for Continual Learning](https://mlsanity.com/paper/2601.13020/) (2026)
- [Revisiting Robustness for LLM Safety Alignment via Selective Geometry Control](https://mlsanity.com/paper/2602.07340/) (2026)
- [Lingua-SafetyBench: A Benchmark for Safety Evaluation of Multilingual Vision-Language Models](https://mlsanity.com/paper/2601.22737/) (2026)
- [VC-Soup: Value-Consistency Guided Multi-Value Alignment for Large Language Models](https://mlsanity.com/paper/2603.18113/) (2026)
- [StressWeb: A Diagnostic Benchmark for Web Agent Robustness under Realistic Interaction Variability](https://mlsanity.com/paper/2604.16385/) (2026)
- [Mitigating Error Amplification in Fast Adversarial Training](https://mlsanity.com/paper/2604.24332/) (2026)
- [Beyond Cross-Lingual Transfer: Benchmarking Propagation Boundaries in Multilingual LLM Unlearning](https://mlsanity.com/paper/2609.05976/) (2026)
- [One Anchor for All: Unified Multilingual and Multimodal Safety Alignment for LVLMs](https://mlsanity.com/paper/2607.27917/) (2026)

## Abstract (arXiv)

As Large Language Models (LLMs) increasingly shape online content, removing targeted information from well-trained LLMs (also known as LLM unlearning) has become critical for web governance. A key challenge lies in sample-wise imbalance within the forget set: different samples exhibit widely varying unlearning difficulty, leading to asynchronous forgetting where some knowledge remains insufficiently erased while others become over-forgotten. To address this, we propose BalDRO, a novel and efficient framework for balanced LLM unlearning. BalDRO formulates unlearning as a min-sup process: an inner step identifies a worst-case data distribution that emphasizes hard-to-unlearn samples, while an outer step updates model parameters under this distribution. We instantiate BalDRO via two efficient variants: BalDRO-G, a discrete GroupDRO-based approximation focusing on high-loss subsets, and BalDRO-DV, a continuous Donsker-Varadhan dual method enabling smooth adaptive weighting within standard training pipelines. Experiments on TOFU and MUSE show that BalDRO significantly improves both forgetting quality and model utility over existing methods, and we release code for reproducibility.