Back to Devoured - July 28, 2026
DEVOURED
PorTAL

PorTAL

AI llm Ramp Labs
Ramp Labs open-sourced PorTAL, a framework that enables cross-model LoRA adaptation by learning a shared, base-agnostic task representation.
What: PorTAL allows a single task to be trained once and then adapted to different frozen base models by generating ordinary per-layer LoRA weights via a light alignment layer.
Why it matters: This approach aims to reduce the overhead of fine-tuning different models for the same task, offering a modular path to share specialized skills across model architectures.
Decoder
  • LoRA (Low-Rank Adaptation): A technique that fine-tunes large models by training small, rank-decomposition matrices while keeping the original model weights frozen.
  • PEFT (Parameter-Efficient Fine-Tuning): A collection of techniques to adapt large models to new tasks by updating only a small subset of parameters.
Original article

We’re open-sourcing PorTAL, our framework for shared task representations and cross model LoRA adaptation.

It now spans from hybrid attention models to multimodal systems including Gemma 4 E2B, Mistral 7B & Inkling.

Code: github.com/ramp-public/portallib

Models + data: huggingface.co/RampPublic

PyPI: pypi.org/project/portallib/

Introducing Latent Briefing

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.

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.

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.

Explore all 42 articles in today's digest

Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi