The Complexity Trap: Simple Observation Masking Is as Efficient as LLM Summarization for Agent Context Management Paper โข 2508.21433 โข Published Aug 29, 2025 โข 8
TAPS: Task Aware Proposal Distributions for Speculative Sampling Paper โข 2603.27027 โข Published Mar 27 โข 143
NearID: Identity Representation Learning via Near-identity Distractors Paper โข 2604.01973 โข Published 28 days ago โข 31
Mind-the-Glitch: Visual Correspondence for Detecting Inconsistencies in Subject-Driven Generation Paper โข 2509.21989 โข Published Sep 26, 2025 โข 23