# Compressing AI Traffic: Standardized Neural Network Coding of Visual-Token Representations in Split Vision-Language Inference

arXiv: https://arxiv.org/abs/2609.01200 · HTML: https://mlsanity.com/paper/2609.01200/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Visual token pruning and efficiency, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 7 of its 10 references that are in the map.
- First posted on arXiv 1 Sep 2026.

## Placed in

- Multimodal LLMs / vision-language models / [Visual token pruning and efficiency](https://mlsanity.com/mllm/visual-token-pruning-and-efficiency/)

## Builds on (most cited first, 7 of 7)

- [Qwen3-VL Technical Report](https://mlsanity.com/paper/2511.21631/) (2025)
- [Video-MME: The First-Ever Comprehensive Evaluation Benchmark of Multi-modal LLMs in Video Analysis](https://mlsanity.com/paper/2405.21075/) (2024)
- [MLVU: Benchmarking Multi-task Long Video Understanding](https://mlsanity.com/paper/2406.04264/) (2024)
- [BottleNet++: An End-to-End Approach for Feature Compression in Device-Edge Co-Inference Systems](https://mlsanity.com/paper/1910.14315/) (2019)
- [Video Coding for Machines: A Paradigm of Collaborative Compression and Intelligent Analytics](https://mlsanity.com/paper/2001.03569/) (2020)
- [Deep feature compression for collaborative object detection](https://mlsanity.com/paper/1802.03931/) (2018)
- [DeepCABAC: A Universal Compression Algorithm for Deep Neural Networks](https://mlsanity.com/paper/1907.11900/) (2019)

## Built on by (most cited first, 1 of 1)

- [FRESHLATENT: Channel-Aware Latent Adaptation for Resource-Constrained Embodied VLM Perception](https://mlsanity.com/paper/2609.30629/) (2026)

## Abstract (arXiv)

When the visual encoder and the language decoder of a vision-language model (VLM) run on different compute nodes, the intermediate visual-token embeddings become a communicated payload rather than an internal activation. We call such machine-consumed intermediate tensors AI traffic and ask how far they can be compressed with a standardized, training-free codec. We insert ISO/IEC 15938-17 Neural Network Coding (NNC) round trips on the complete visual interface of a Qwen3-VL-8B-Instruct video question answering pipeline, comprising the main visual-token representation and the DeepStack feature streams, while leaving weights, prompts, and generation untouched, and sweep the quantization parameter (QP) over a wide rate range. Closed-ended Video-MME accuracy remains close to the uncompressed reference up to a 98% reduction of the transmitted BF16 tensor and only then collapses; open-ended MLVU generation shows the same plateau-and-collapse profile under an LLM judge. This robustness is not due to near-lossless reconstruction: the decoded tensor is heavily discretized, carries substantial row-wise relative L2 error, and has a visibly steeper singular-value decay than its source. Downstream reasoning therefore depends on coarse structure and relative geometry rather than exact floating-point values, which argues for rate-task rather than rate-distortion optimization of AI traffic codecs.