Paper-Conference

Persistent Memory in Multi-Agent LLM Inference: What It Costs, What It Buys, and When You Can Tell

Agent decomposition bounds peak KV cache 2.5× below single-pass and RAG baselines; the persistent memory tier such systems are usually paired with costs cache and buys no …

Hochan Son

• Sep 30, 2026 • 1 min read

Generative AI

SYNTHONY: A Stress-Aware, Intent-Conditioned Agent for Deep Tabular Generative Models Selection

Stress-aware, intent-conditioned agent that recommends the best-fit tabular generative model for a dataset instead of defaulting to one architecture.

Hochan Son

• Mar 31, 2026 • 1 min read

No results found

Paper-Conference

Persistent Memory in Multi-Agent LLM Inference: What It Costs, What It Buys, and When You Can Tell

SYNTHONY: A Stress-Aware, Intent-Conditioned Agent for Deep Tabular Generative Models Selection