FocusGUI adapter cuts sequential distraction in omni-modal GUI agents
Across four frozen backbones, authors report Type Match up 33.2 points and Exact Match 3.0 times the frozen baseline.
Stories per day · 30 days
3 RSS
8 stories, newest first · one card per story
Stories tagged #adapters
Towards In-Parameter Memory Augmentation for Large Language Models
Survey maps in-parameter memory methods that plug reusable knowledge into LLM inference without growing context.
Conditional Rank Allocation for Taxonomy-Aware Medical Language Model Adaptation
ARBOR's conditional rank allocation lifts Qwen3-8B medical QA accuracy to 69.69% across four benchmarks.
Local Support Learning
Local Support Learning limits catastrophic forgetting in LLMs up to 7B without replaying prior data.
Backdoor Mitigation in Decentralized LLM Fine-Tuning
Chorus lets decentralized LLM fine-tuning nodes reject backdoored adapters, cutting neighbor attack success to at most 2.2%.
New LoRA Skills Should Read but Never Write
READ composes LoRA adapters so new skills can read old subspaces but cannot overwrite them.
A tiny software layer from lab-grown neurons promises faster, cheaper AI video
TBC plans a neuron-inspired video adapter, but the commercial model and performance claims remain unverified.
2
ServeGuard: Verifiable, Bounded-Residual Confinement of Operator-Invisible Channels Without Revealing the Certified Read Factor
ServeGuard proves open-weight adapters contain no hidden monitor-blind channel without exposing certified weights.