This work proposes MAEM, a replay-free, capacity-controlled dynamic mixture-of-experts framework for class-incremental graph learning that mitigates recency bias, outperforms strong continual graph learning baselines, and keeps model capacity bounded under strict replay-free learning.
Jun-Lin Chen, Wen-Hao Wang, Jun Xu et al.· 0 citations
Graph few-shot class-incremental learning (GFSCIL) requires a model to continually recognize emerging classes from only a few labeled nodes while preserving previously acquired knowledge. Beyond the catastrophic forgetting inherited from conventional graph continual learning, GFSCIL presents two distinctive challenges:...
Jun-Lin Chen, Yu-Han Wang, Xue-Fei Wang et al.· 0 citations
This tutorial provides a systematic overview of LLM personalization, covering core paradigms such as user-aware representations, prompt-based adaptation, memory augmentation, parameter-efficient fine-tuning, and continual learning.
Rui-Jie Wang, Qing-Kai Zeng, Xue-Fei Wang et al.· Proceedings of the 32nd ACM...· 0 citations
PAC-Bayes-regularized Meta-LoRA is proposed, which uses a meta-learned LoRA initialization as both the adaptation start and prior center, while adjusting update strength according to support-set size and predictive uncertainty to limit overfitting under sparse or ambiguous evidence.
Xuefei Wang, Jun Han, Zi-Xuan Wang et al.· 0 citations
Hierarchical Response Preservation (HiRP) is proposed, a hierarchical response that keeps each historical-class probability and sums new-class probabilities, preserving historical distinctions and aggregate competition while allowing distinctions within the new class group to adapt.
Haopeng Zhang, Yu-Han Wang, Yu-Bing Su et al.· 0 citations
This work introduces the Counterfactual Quotient Model, which treats action-conditioned futures as equivalent when they differ only by a component shared across actions, and establishes the decision sufficiency, identifiability, common-mode invariance, approximation behavior, and regret properties of the resulting repr...
LLMODE is proposed, a token-efficient framework for irregular spatio-temporal forecasting with a frozen LLM backbone that shows competitive overall performance, with clearer advantages under sparse or dynamically complex irregular sampling.
Di Zhang, Jing-Yang Zhang, Zi-Qian Wang et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.