MLUBench: New Benchmark for Evaluating Lifelong Unlearning in Multimodal AI Models
Researchers introduced MLUBench, a large-scale benchmark with 127 entities across 9 classes to evaluate how multimodal large language models (MLLMs) handle sequential data removal requests. The benchmark reveals that existing unlearning methods suffer from cumulative degradation, particularly because removing content from one modality (text or image) can damage the model's overall multimodal alignment. The work proposes LUMoE, a new method designed to mitigate these degradation problems and addresses an increasingly important challenge as data owners request removal of their content from trained models.
Researchers have created MLUBench, a comprehensive benchmark designed to evaluate lifelong unlearning in multimodal large language models—a critical capability as data owners increasingly request removal of specific content from trained models. The benchmark contains 127 entities across 9 classes and tests how models handle sequential unlearning requests over time, a scenario more realistic than single-instance removal. Extensive experiments using MLUBench demonstrate that existing unlearning methods experience severe, cumulative performance degradation as more unlearning requests are processed. A key finding is that multimodal models face a unique constraint absent in unimodal systems: unlearning from one modality (such as text or images) can degrade the entire model's performance due to the tight coupling between modalities. To address this challenge, the researchers propose LUMoE, a method that significantly reduces degradation compared to baseline approaches. The source code and dataset have been made publicly available.
What different sources said
- arXiv cs.AICenter
MLUBench: A Benchmark for Lifelong Unlearning Evaluation in MLLMs
Related
Topology-Aware Thermodynamics Improves DNA Probe Specificity Design
Researchers developed a new framework for designing DNA probes that accounts for the spatial organization of matched sequences, not just overall thermodynamic stability. Traditional methods rely on scalar measures like melting temperature and free energy, which miss how mismatches are distributed along the probe. The approach could improve diagnostic accuracy in applications like HPV detection and gene expression profiling.
Study Identifies Optimal Thermal Dose for Combining Focused Ultrasound with Immunotherapy in Tumors
Researchers used multimodal PET imaging to identify an optimal thermal dose range for focused ultrasound ablation that destroys tumor tissue while preserving conditions for immunotherapy delivery. The study found that excessive heating collapses blood vessels needed for antibody access, while insufficient heating fails to adequately reduce tumor burden. The findings could guide clinical design of combination treatments pairing thermal ablation with immunotherapies.
Plant MSH1 Protein Functions as Mismatch-Directed Nuclease for Organelle Genome Maintenance
Researchers have identified the precise mechanism by which the AtMSH1 protein in Arabidopsis plants recognizes and cleaves DNA mismatches and lesions, preventing mutations in organellar genomes. The protein combines a DNA mismatch recognition module with a nuclease domain that makes staggered cuts at specific positions relative to DNA damage. This discovery explains how plants maintain unusually low mutation rates in their mitochondrial and chloroplast DNA compared to other eukaryotes.