PorTAL (1 minute read)

TLDR AI Papers

Summary

Introduces Latent Briefing, a method for multi-agent systems to communicate via KV cache compaction, achieving 31% fewer tokens with same accuracy and up to 20x speedup.

Ramp Labs has open-sourced PorTAL, a framework for shared task representations and cross-model LoRA adaptation. PorTAL learns a base-agnostic task latent and a light per-base alignment that generates ordinary per-layer LoRA weights. A task can be trained once, adapted to supported frozen base models, and exported as a standard Hugging Face PEFT adapter.
Original Article
View Cached Full Text

Cached at: 07/28/26, 06:22 PM

# Thread by @RampLabs on Thread Reader App Source: [https://threadreaderapp.com/thread/2081819550329327689.html](https://threadreaderapp.com/thread/2081819550329327689.html) Introducing Latent Briefing, a way for agents to quickly share their relevant memory directly\. Result: 31% fewer tokens used, same accuracy\. Multi\-agent systems are powerful, but can be wildly inefficient\. They pass context as tokens, so costs explode and signal gets lost\. We built an algorithm that allows agents to communicate KV cache to KV cache\. ![Video Poster](https://pbs.twimg.com/amplify_video_thumb/2042672677110841344/img/JGgMkS-knZ9LmohN.jpg) Agents need to share context, but doing it in token space has real tradeoffs: • LLM summaries: slow \(20–60s\), lossy, and often miss what the next agent actually needs • RAG: splits context into chunks, so relationships across documents get lost • Passing full context: expensive, noisy, and often hurts accuracy Our method skips tokens entirely\. We operate on the KV cache, using the worker's own attention patterns to extract what's relevant from the orchestrator's memory and discard the rest\. ![Video Poster](https://pbs.twimg.com/amplify_video_thumb/2042672775543033857/img/eoCbEzi329lRGmab.jpg) We adapted the Attention Matching \(AM\) KV cache compaction framework\. The AM algorithm compacts the KV cache \(C1, β, C2\) preserving attention outputs through a correction term\. We modified the algorithm to make it inference ready: 1\.Score tokens using the worker's task query, not self attention 2\.Global mask across all heads → enables massive batching 3\.MAD\-normalized thresholding for adaptive compression Result: 320 sequential solves → 2\-3 batched ops\. 20x speedup to a median of 1\.7 s\.[![Image](https://threadreaderapp.com/images/1px.png)](https://pbs.twimg.com/media/HFkIHUvbQAAdO-l.jpg)

Similar Articles

Prompt Caching In Agents

Lobsters Hottest

The article explains how prompt caching works in large language model agents, covering KV cache mechanics, prefill and decode phases, and the impact on latency, cost, and agent design.