JumpLoRA introduces adaptive sparsity in LoRA blocks via JumpReLU gating for continual learning in LLMs, achieving dynamic parameter isolation to prevent task interference. The method is modular, compatible with existing LoRA-based continual learning approaches, and significantly boosts performance over IncLoRA by constraining both magnitude and direction of updates.
AEGIS addresses catastrophic forgetting when fine-tuning vision-language models for robotic control by preventing cross-modal gradient asymmetry—high-magnitude continuous action gradients overwriting the VLM's cross-entropy pre-trained manifold. Uses anchor-enforced gradient isolation to preserve VQA capabilities while injecting flow-matching action supervision, unlike stop-gradient or LoRA approaches.
MMOT introduces an Optimal Transport-based framework for online incremental learning that maintains evolving mixture model centroids instead of fixed or single adaptive centroids per class. The approach better handles multimodal data streams in continual learning scenarios where distributional shifts are severe and replay buffers have limited utility. Novel contribution is the dynamic centroid evolution mechanism grounded in OT theory.
SkillClaw enables LLM agent skills to evolve autonomously by aggregating interaction experiences across users, with an 'agentic evolver' that refines capabilities from real-world usage. Achieves +42.1% improvement by shifting from static, manually-engineered skills to continuously improving ones learned from collective deployment data.