Semantic and Goal-Oriented Communication

This direction is merged into another direction: semantic communication, semantic-agnostic communication. Its last version stays readable and citable here.

Version 1 of 1103 papers105 conceptsfrozen 2026-10-04

For over seventy years, digital communication has been governed by Claude Shannon’s "Level A" problem: how to transmit symbols accurately over a noisy channel. In this paradigm, every bit is created equal. Whether a bit represents a crucial pixel in a medical scan or background noise in a video, the network treats them with the same urgency, striving for bit-perfect reconstruction.

As we approach the 6G era, this bit-centric approach is hitting a wall. The explosion of the Internet of Things (IoT), autonomous vehicles, and the Metaverse creates data demands that outstrip even the most advanced spectral efficiency gains. Researchers like Kai Niu and Ping Zhang argue for a fundamental paradigm shift: moving from transmitting raw data to transmitting meaning (Level B) and effectiveness (Level C).

The Shift from Reconstruction to Regeneration

Traditional communication systems focus on reconstruction—rebuilding the exact source signal at the receiver. Semantic communication (SemCom) shifts the focus to regeneration. Instead of sending a high-resolution image of a car, a semantic transmitter might extract the "meaning" (e.g., "a red sedan at coordinates X, Y") and send only that compact representation.

This is often achieved through Joint Source-Channel Coding (JSCC). Unlike traditional systems that compress data and then add error correction in two separate steps, DeepJSCC, pioneered by researchers like Deniz Gunduz and Mikolaj Jankowski, maps source features directly to wireless signals. This allows for "graceful degradation"—as the signal weakens, the image doesn't suddenly disappear into digital "cliff-effect" noise; instead, it loses fine detail while preserving the essential semantic content.

Goal-Oriented Filtering and the Information Bottleneck

Why transmit the color of a car if the receiver's only task is to count how many vehicles pass through an intersection? This is the core of Goal-Oriented Communication. By understanding the downstream task, the system can "take only what it needs," as described by Xiang Chen.

This filtering is often framed using the Information Bottleneck (IB) principle. As Francesco Pezone explains, the goal is to find a representation that is as compressed as possible while retaining maximum predictive power for the specific task. By discarding task-irrelevant information at the source, systems like PragComm can reduce communication costs by over 30,000× while maintaining high performance in collaborative perception for robotics.

Generative AI as a Shared Knowledge Base

The most radical efficiency gains come from integrating Generative AI (GenAI). In this framework, the transmitter and receiver share a "knowledge base"—a set of pre-trained models that understand the world. Yaping Sun and Stephen D. Liang demonstrate that by transmitting concise textual descriptions (e.g., via Vision-Language Models) and using tools like Stable Diffusion at the receiver, data volume can be reduced by 99.9%.

The receiver isn't just decoding bits; it is imagining the scene based on a few semantic "clues." To prevent the "hallucinations" common in GenAI, researchers like Shunpu Tang suggest using Retrieval-Augmented Generation (RAG), allowing the system to ground its regeneration in historical data or external knowledge graphs.

New Metrics for a New Paradigm

If bit-perfect accuracy is no longer the goal, how do we measure success? Traditional metrics like Bit Error Rate (BER) are insufficient because a bit error in a "meaningless" background pixel is less damaging than an error in a "critical" feature.

  1. Age of Incorrect Information (AoII): Introduced by Ali Maatouk, this metric combines data freshness with accuracy. It penalizes the system only when the information at the receiver is both "stale" and "wrong" relative to the goal.
  2. Classification Margin: Zhanwei Wang argues that we can exploit the inherent robustness of AI models. If a classifier can still correctly identify an object despite 5% bit errors, those errors are semantically irrelevant. This allows for ultra-low-latency transmission by intentionally skipping expensive error-correction cycles.

The Future: Semantic-Aware Networking

Ultimately, this research points toward Semantic-aware Networking, a concept championed by Guangming Shi and Yong Xiao. In this vision, the network is no longer a "dumb pipe" for bits. Instead, it is an intelligent fabric that understands the context of the data it carries, prioritizing information based on its utility and meaning to the end-user. By shifting the burden from transmission to computation, semantic communication paves the way for a 6G ecosystem that is not just faster, but significantly smarter.

Papers in this version

Showing 60 of 103 papers.

History & citationVersion 1 · 2026-10-04

Merged into another direction · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_semantic_and_goal_oriented_communication_0ce7feef,
  title = {Semantic and Goal-Oriented Communication},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/semantic-and-goal-oriented-communication/rdr_0ce7feef2dbe41fb85ab81dbd4c81c52},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_semantic_and_goal_oriented_communication_0ce7feef_v1,
  title = {Semantic and Goal-Oriented Communication},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/semantic-and-goal-oriented-communication/rdr_0ce7feef2dbe41fb85ab81dbd4c81c52/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}