发表机构
Aalto University; Nokia Technologies（阿尔托大学; 诺基亚技术公司）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究针对拆分式视觉-语言推理场景，采用ISO/IEC 15938-17神经网络编码压缩AI流量，在Qwen3-VL-8B-Instruct流水线中实现98%压缩率下的任务性能鲁棒，提出应采用码率-任务优化而非码率-失真优化。
AI 中文摘要
当视觉-语言模型（VLM）的视觉编码器与语言解码器运行在不同计算节点时，中间视觉令牌嵌入会成为需传输的有效载荷而非内部激活。我们将这类机器消耗的中间张量称为AI流量，并探究其在采用标准化、无训练编解码器的情况下可被压缩的程度。我们在Qwen3-VL-8B-Instruct视频问答流水线的完整视觉接口中插入ISO/IEC 15938-17神经网络编码（NNC）往返操作，该接口包含主要视觉令牌表示与DeepStack特征流，且保留权重、提示及生成过程不变，同时在宽码率范围内扫描量化参数（QP）。封闭集Video-MME准确率在传输的BF16张量压缩率达98%时仍接近未压缩参考值，仅当压缩率超过该值后才出现崩溃；开放集MLVU生成在LLM评判下呈现相同的平台-崩溃曲线。这种鲁棒性并非源于近无损重构：解码后的张量存在严重离散化，带有显著的逐行相对L2误差，且其奇异值衰减比源张量更陡峭。因此，下游推理依赖于粗略结构与相对几何而非精确浮点值，这表明AI流量编解码器应采用码率-任务优化而非码率-失真优化。
英文摘要
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.
Comments4 pages, 6 figures, 1 table
