Information Abstraction for Data Transmission Networks based on Large Language Models

TL;DR

The paper introduces the Degree of Information Abstraction (DIA), a computable metric that jointly assesses information compression and semantic preservation to guide energy-efficient data transmission. DIA provides a principled alternative to mutual-information objectives, enabling multimodal representations through a shared latent space and aligning with semantic objectives via a KL-based semantic discrepancy term. The authors instantiate a DIA-based optimization framework (OPRO) for LLM-guided semantic video transmission, augmented by the Video Semantic Differential Stream (VSDS) to capture spatio-temporal semantic dynamics; they demonstrate substantial data-rate reductions while preserving semantic fidelity and show convergence with IB-inspired baselines. The work suggests broad implications across neural architecture design, semantic communication, neuromorphic computing, and joint sensing-communication systems, proposing practical paths for differentiable estimators, NAS integration, and cross-domain abstraction-aware optimization.

Abstract

Biological systems, particularly the human brain, achieve remarkable energy efficiency by abstracting information across multiple hierarchical levels. In contrast, modern artificial intelligence and communication systems often consume significant energy overheads in transmitting low-level data, with limited emphasis on abstraction. Despite its implicit importance, a formal and computational theory of information abstraction remains absent. In this work, we introduce the Degree of Information Abstraction (DIA), a general metric that quantifies how well a representation compresses input data while preserving task-relevant semantics. We derive a tractable information-theoretic formulation of DIA and propose a DIA-based information abstraction framework. As a case study, we apply DIA to a large language model (LLM)-guided video transmission task, where abstraction-aware encoding significantly reduces transmission volume by $99.75\%$, while maintaining semantic fidelity. Our results suggest that DIA offers a principled tool for rebalancing energy and information in intelligent systems and opens new directions in neural network design, neuromorphic computing, semantic communication, and joint sensing-communication architectures.

Figures (4)

• Figure 1: Two transmitter–receiver workflows in DIA-based semantic video transmission. \ref{['fig:dia_opro_workflows:a']} depicts the standard pipeline, where DIA provides scalar feedback for OPRO-driven configuration search. \ref{['fig:dia_opro_workflows:b']} augments the transmitter with a prediction module that locally generates a video conditioned on the candidate caption to assess its semantic adequacy before transmission, which is preferable when the transmitter has sufficient computation resources.
• Figure 2: SSIM versus optimization round for different optimization strategies.
• Figure 3: VMAF versus optimization round for different optimization strategies.
• Figure 4: DIA versus optimization round for different optimization strategies.