Multi-Stream LLMs: Unblocking Language Models with Parallel Streams of Thoughts, Inputs and Outputs

TL;DR

This paper proposes multi-stream language models that process parallel streams of thoughts, inputs, and outputs to overcome the limitations of sequential message exchanges in AI agents, enhancing efficiency and usability.

Contribution

It introduces a novel instruction-tuning approach for parallel streams in language models, enabling simultaneous reading, thinking, and acting within a single model pass.

Findings

Parallel streams improve model efficiency and responsiveness.

Multi-stream models address usability limitations of sequential message formats.

Enhanced separation of concerns increases security and monitorability.

Abstract

The continued improvements in language model capability have unlocked their widespread use as drivers of autonomous agents, for example in coding or computer use applications. However, the core of these systems has not changed much since early instruction-tuned models like ChatGPT. Even advanced AI agents function on message exchange formats, successively exchanging messages with users, systems, with itself (i.e. chain-of-thought) and tools in a single stream of computation. This bottleneck to a single stream in chat models leads to a number of limitations: the agent cannot act (generate output) while reading, and in reverse, cannot react to new information while writing. Similarly, the agent cannot act while thinking and cannot think while reading or acting on information. In this work, we show that models can be unblocked by switching from instruction-tuning for sequential message formats to instruction-tuning for multiple, parallel streams of computation, splitting each role into a separate stream. Every forward pass of the language model then simultaneously reads from multiple input streams and generates tokens in multiple output streams, all of which causally depend on earlier timesteps. We argue that this data-driven change remedies a number of usability limitations as outlined above, improves model efficiency through parallelization, improves model security through better separation of concerns and can further improve model monitorability.