---
title: "How to Guide Your Language Flow"
description: "Probe guidance, a novel method for guiding flow matching models, is introduced. It uses the frozen internal states of an existing diffusion model to construct a guidance signal, operating on a similar principle as autoguidance but removing the need for an additional forward pass at inference time. The method provides a reliable path ensuring the weak and strong models share similar dynamics. Probe guidance sets a new state-of-the-art performance on unconditional generation when benchmarked on continuous diffusion language models. Applied to a 1.7B diffusion language model, it consistently improves multiple choice question answering benchmarks."
---

# How to Guide Your Language Flow

[Apple](https://yomu.fyi/company/apple) · Sep 23, 2026

**Type:** Announcement

## Summary

Probe guidance, a novel method for guiding flow matching models, is introduced. It uses the frozen internal states of an existing diffusion model to construct a guidance signal, operating on a similar principle as autoguidance but removing the need for an additional forward pass at inference time. The method provides a reliable path ensuring the weak and strong models share similar dynamics. Probe guidance sets a new state-of-the-art performance on unconditional generation when benchmarked on continuous diffusion language models. Applied to a 1.7B diffusion language model, it consistently improves multiple choice question answering benchmarks.

## Approach / What changed

Probe guidance uses the frozen internal states of an existing diffusion model to construct a guidance signal, eliminating the need for an additional forward pass at inference time.

## Takeaways

- Probe guidance uses frozen internal states of an existing diffusion model to construct a guidance signal, eliminating the need for an additional forward pass at inference time.
- When applied to a 1.7B diffusion language model, probe guidance consistently improves multiple choice question answering benchmarks.
- A weak model used for autoguidance must come from a low-entropy region of training to ensure reliable performance.

**Tags:** [AI](https://yomu.fyi/topic/ai), [LLMs](https://yomu.fyi/topic/llm)

- Source: [Apple](https://machinelearning.apple.com/research/guide-language-flow)
- Source URL: https://machinelearning.apple.com/research/guide-language-flow
- Ingested by Yomu: 2026-09-24T00:00:13.657Z

[Read original post](https://machinelearning.apple.com/research/guide-language-flow)