Paper Detail

Subliminal Clocks: Latent Time Modelling in Diffusion Language Models

Maximo Eduardo Rulli, Thomas Vaitses Fontanari, Simone Petruzzi, Federico Alvetreti, Giorgio Strano, Donato Crisostomi, Giorgos Nikolaou, Tommaso Mencattini, Andrea Santilli, Emanuele Rodolà, Simone Scardapane, Alessio Devoto

huggingface Score 6.0

Published 2026-07-20 · First seen 2026-07-23

General AI

Abstract

Diffusion Language Models (DLMs) have recently emerged as a promising alternative to autoregressive models. Unlike standard diffusion-based approaches, DLMs are not explicitly conditioned on a timestep, raising a natural question: do these models internally represent denoising progress, and how is such information used downstream? In this work, we show that DLMs do in fact encode a latent representation related to the diffusion timestep within their residual streams. We find that this signal can be reliably extracted using probes across layers, indicating that denoising progress is decodable from internal activations. We further demonstrate that steering the model along a low-dimensional subspace associated with the inferred timestep allows us to systematically modulate its notion of denoising progress, leading to predictable changes in model confidence and entropy. Finally, we analyse the geometry of the identified representation, showing that it exhibits structured and interpretable properties in activation space, and shedding light on how such a signal is processed by these models.

Workflow Status

Review status
pending
Role
unreviewed
Read priority
later
Vote
Not set.
Saved
no
Collections
Not filed yet.
Next action
Not filled yet.

Reading Brief

No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.

Why It Surfaced

No ranking explanation is available yet.

Tags

No tags.

BibTeX

@misc{rulli2026subliminal,
  title = {Subliminal Clocks: Latent Time Modelling in Diffusion Language Models},
  author = {Maximo Eduardo Rulli and Thomas Vaitses Fontanari and Simone Petruzzi and Federico Alvetreti and Giorgio Strano and Donato Crisostomi and Giorgos Nikolaou and Tommaso Mencattini and Andrea Santilli and Emanuele Rodolà and Simone Scardapane and Alessio Devoto},
  year = {2026},
  abstract = {Diffusion Language Models (DLMs) have recently emerged as a promising alternative to autoregressive models. Unlike standard diffusion-based approaches, DLMs are not explicitly conditioned on a timestep, raising a natural question: do these models internally represent denoising progress, and how is such information used downstream? In this work, we show that DLMs do in fact encode a latent representation related to the diffusion timestep within their residual streams. We find that this signal can},
  url = {https://huggingface.co/papers/2607.01774},
  keywords = {huggingface daily},
  eprint = {2607.01774},
  archiveprefix = {arXiv},
}

Metadata

{}