Tag
MAAT introduces a multi-phase LoRA-adapter unlearning method along with the 5WBENCH benchmark, revealing that causal 'Why' knowledge is uniquely difficult to forget due to long multi-hop answer chains and gradient dilution, achieving strong forget–retain trade-offs on Llama 3.2-3B.
Proposes DIVE, a compression adapter for embedding dimensionality reduction that uses self-limiting gradient updates and head-wise NT-Xent contrastive loss to prevent overfitting on small datasets, outperforming existing methods on BEIR benchmarks.
This paper introduces an Information Bottleneck Adapter (IB-Adapter) for Vision-Language-Action (VLA) models to improve robustness against unseen visual disturbances without requiring extra data, achieving up to 30% improvement with minimal parameter overhead.
Microsoft’s new Teams TypeScript SDK lets developers expose existing AI agents or bots as Teams apps with a three-line HTTP server adapter, enabling shared agent logic across Slack and Teams.