Ramp Labs Introduces Multi-Agent Memory Sharing Solution, Token Consumption Reduced by Up to 65%
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.
You may also like

Morning News | The draft amendment to the People's Bank of China Law aims to clarify the legal status of digital renminbi; South Korea will transfer about 40 unregistered virtual asset service providers to law enforcement agencies

The cryptocurrency industry has entered the "Show Me" era: merely relying on vision is no longer enough

Interpreting the Ethereum Foundation's new structure: Reaffirming self-sovereignty amid institutional trends

Former SpaceX engineer reconstructs the financial execution system using first principles

Tidal Investment: We still have a positive outlook on the AI industry chain, but the reasons have changed

Standard Chartered Bank sings a 50x rhapsody again, aiming for AAVE to reach 3500 USD

The interim executive director of the Ethereum Foundation speaks out: What is our mission?

Why does OKX want to start a new company with the parent company of the New York Stock Exchange?

Why Is PAXG Price Different From Gold? 5 Reasons Crypto Traders Should Know

WEEX OpenAPI 101: 5 Powerful Modules, AI Trading Tools, and Grab Up to 70% Revenue Opportunities
Learn how WEEX OpenAPI connects traders, developers, AI agents, and trading platforms. Discover WEEX API features, Binance-compatible integration, automated trading workflows, revenue opportunities, and ecosystem possibilities.

Interview with NDV Founder Jason Huang: Popping the AI Bubble and the Myth of Microstrategy, Seeking the Ultimate Ace in the Crypto Market

Morning Report | Former Ethereum Foundation researcher establishes Ethlabs; EU Parliament Economic Committee passes digital euro regulatory proposal

Dragonfly partner Haseeb: The fastest-growing companies in the future may all be stuck at 149 people

How xBubble Breaks the Deadlock in VC's Heavy Investment in the OPC Economy

The encrypted unicorn Blockstream is deeply embroiled in a serious fraud case

Morning Report | The South Korean Financial Services Commission plans to expand the regulatory sandbox to include virtual assets; the parent company of the New York Stock Exchange, ICE, has reached a partnership with OKX to jointly establish a cryptocu...

Exclusive Interview with Strategy CEO: Putting Aside the Sale of 32 BTC, the 60 Trillion AI Intelligence is the Ultimate Fate of Bitcoin






