Entropic-Time Inference: Self-Organizing Large Language Model Decoding Beyond Attention

TL;DR

This paper introduces entropic-time inference, a novel approach for large language model decoding that uses uncertainty flow instead of token index, enabling resource-efficient and stable generation.

Contribution

It presents a self-organizing inference architecture that integrates scheduling, attention sparsification, and temperature control under an entropy-based objective.

Findings

Enhanced inference stability near target entropy.

Improved resource allocation during decoding.

Extended vLLM with entropy-aware mechanisms.

Abstract

Modern large language model (LLM) inference engines optimize throughput and latency under fixed decoding rules, treating generation as a linear progression in token time. We propose a fundamentally different paradigm: entropic\-time inference, where decoding is governed by the flow of uncertainty rather than token index. We introduce a self\-organizing inference architecture that jointly couples scheduling, attention sparsification, and sampling temperature under a unified entropy control objective. Our method extends vLLM with entropy-aware scheduling, entropic pruning of paged attention blocks, and adaptive temperature control that stabilizes generation near a target entropy regime. This transforms inference into a resource\-intelligent thermodynamic process that allocates computation where uncertainty reduction is maximized. We present a concrete systems design, pseudocode, and integration plan, demonstrating how entropy can serve as a first\-class control signal for scalable LLM inference.