Tag
This paper introduces EvtGraph, a framework for efficient multimodal temporal graph learning that compresses sequences into event-level tokens under a budget constraint, demonstrating improved performance and efficiency over Transformer and recurrent baselines on clinical and cross-domain benchmarks.
The paper proposes Mass-Aware Attention (MAA), a generalization of standard attention normalization that retains evidence accumulation information by scaling numerator and denominator at different rates under repetition, improving representation informativeness and task performance across temporal graph and other domains.
A detailed breakdown of loop engineering for AI agents: moving from manual agent supervision to autonomous loops with triggers, makers, checkers, and persistent state. Recommends Zep's Graphiti for temporal knowledge graphs and Comet's Opik for observability to build reliable unattended agent systems.
TCAR-Gen proposes a framework combining query-conditioned graph neural networks, temporal evidence fusion, and chain-of-trees reasoning for temporal graph retrieval in knowledge-grounded generation. It achieves improved recall on the Victorian Crime Diaries benchmark across multiple query types.
A thread discussing the importance of schema discipline in agent memory, introducing Zep AI's open-source Graphiti library for building temporal knowledge graphs with constrained entity and relationship types.