Ramp Labs Introduces Multi-Agent Memory Sharing Solution, Token Consumption Reduced by Up to 65%

By: theblockbeats.news|2026/04/11 06:19:36
0
Share
copy

BlockBeats News, April 11th, AI infrastructure company Ramp Labs released research results on "Latent Briefing", achieving efficient memory sharing among multi-agent systems through direct compression of large-scale model KV cache, significantly reducing Token consumption without sacrificing accuracy.


In mainstream multi-agent architectures, the Orchestrator decomposes tasks and repeatedly calls Worker models. As the inference chain extends, Token usage exponentially inflates. The core idea of Latent Briefing is to leverage the attention mechanism to identify the truly critical parts in the context, directly discard redundant information at the representation layer, rather than relying on the slow-speed LLM summary or the unstable RAG retrieval.


In the LongBench v2 benchmark test, this method performed remarkably: Worker model Token consumption decreased by 65%, the median Token savings for medium-length documents (32k to 100k) reached 49%, the overall accuracy improved by approximately 3 percentage points compared to the baseline, and the additional time for each compression was only about 1.7 seconds, achieving a speedup of about 20 times compared to the original algorithm.


The experiment used Claude Sonnet 4 as the Orchestrator, and Qwen3-14B as the Worker model, covering various document scenarios such as academic papers, legal documents, novels, and government reports. The research also found that the optimal compression threshold varies depending on task difficulty and document length—difficult tasks are suitable for aggressive compression to filter out speculative reasoning noise, while long documents are more suitable for mild compression to retain scattered key information.

-- Price

--

You may also like

Morning Report | Samsung announces a 265.5 trillion won investment plan, focusing on semiconductor and AI computing power data centers; Vitalik publishes an article detailing the entire technology tree behind the confusion protocol (iO) mainline

Overview of Important Market Events on June 29

What you bought on CEX is really not US stocks: Analyzing the 94% liquidation monopoly and the evaporation of equity under a five-layer pipeline

Peeling back its smooth trading interface to examine the underlying legal relationships and settlement processes, you will find that this is far from a simple "RWA asset revolution," but rather a complex game of interests involving spot pricing, rights ownership, and the monopoly of underlying custo...

In such a crowded cross-border payment arena, where is the next stop for the future?

Only by stepping into the mud can one have the chance to touch gold.

Why Is Bitcoin Down in 2026? What We Can Learn From 2022

Why is Bitcoin down in 2026? Bitcoin has just recorded its worst first half since 2022, with back-to-back quarterly losses, record ETF outflows, and extreme fear. Here's what history says, how 2026 differs from the last bear market, and the three signals traders should wat

The large models in the United States are moving towards closure in the name of security

The government successfully inserted itself as an approver between commercial AI models and their users for the first time.

From the white-haired stock god to the billionaire fund mogul, the smart people shorting Nvidia are all getting rich using the same framework

Give up on heavily investing in Nvidia's "nine major bottlenecks"! This article analyzes the underlying logic behind top AI investors making billions: physical infrastructure such as electricity, HBM, and optical interconnects are the true keys to wealth in AI hardware.

Contents

Popular coins

Latest Crypto News

Read more
iconiconiconiconiconiconicon
Customer Support:@weikecs
Business Cooperation:@weikecs
Quant Trading & MM:bd@weex.com
VIP Program:support@weex.com