Saved in:
Bibliographic Details
Main Author: Ruffenach, Yves
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2604.12513
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908962563031040
author Ruffenach, Yves
author_facet Ruffenach, Yves
contents We study whether a variational language model can support a minimal and measurable form of agentic control grounded in its own internal evidence. Our model combines local variational hidden computation (EVE), a homeostatic latent regulator, structurally aware checkpoint retention and a calibrated uncertainty-aware controller operating on top of the retained model. Rather than treating uncertainty as a passive diagnostic measured after prediction, we treat it as an operational signal that can regulate training, support checkpoint retention and guide inference-time intervention. The resulting framework is deliberately focused. It studies a closed-loop form of internal control in which structural and predictive signals become actionable. Empirically, the variational backbone improves over a matched deterministic reference on the language-modeling task while also exhibiting a richer and more usable uncertainty profile. On top of this backbone, the calibrated controller remains active, uses multiple actions under a full agentic evaluation and yields a positive quality-cost trade-off. These results support a precise claim: internal uncertainty can serve not only as a descriptive property of a variational language model, but also as a practical control interface for regulation, checkpoint retention and minimal agentic routing.
format Preprint
id arxiv_https___arxiv_org_abs_2604_12513
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Agentic Control in Variational Language Models
Ruffenach, Yves
Machine Learning
We study whether a variational language model can support a minimal and measurable form of agentic control grounded in its own internal evidence. Our model combines local variational hidden computation (EVE), a homeostatic latent regulator, structurally aware checkpoint retention and a calibrated uncertainty-aware controller operating on top of the retained model. Rather than treating uncertainty as a passive diagnostic measured after prediction, we treat it as an operational signal that can regulate training, support checkpoint retention and guide inference-time intervention. The resulting framework is deliberately focused. It studies a closed-loop form of internal control in which structural and predictive signals become actionable. Empirically, the variational backbone improves over a matched deterministic reference on the language-modeling task while also exhibiting a richer and more usable uncertainty profile. On top of this backbone, the calibrated controller remains active, uses multiple actions under a full agentic evaluation and yields a positive quality-cost trade-off. These results support a precise claim: internal uncertainty can serve not only as a descriptive property of a variational language model, but also as a practical control interface for regulation, checkpoint retention and minimal agentic routing.
title Agentic Control in Variational Language Models
topic Machine Learning
url https://arxiv.org/abs/2604.12513