# SAGE: Training-Free Semantic Evidence Composition for Edge-Cloud Inference under Hard Uplink Budgets

arXiv: https://arxiv.org/abs/2604.19623 · HTML: https://mlsanity.com/paper/2604.19623/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 27 of its 31 references that are in the map.
- First posted on arXiv 21 Apr 2026.

## Builds on (most cited first, 27 of 27)

- [Training data-efficient image transformers & distillation through attention](https://mlsanity.com/paper/2012.12877/) (2020)
- [Token Merging: Your ViT But Faster](https://mlsanity.com/paper/2210.09461/) (2022)
- [DynamicViT: Efficient Vision Transformers with Dynamic Token Sparsification](https://mlsanity.com/paper/2106.02034/) (2021)
- [Going deeper with Image Transformers](https://mlsanity.com/paper/2103.17239/) (2021)
- [Not All Patches are What You Need: Expediting Vision Transformers via Token Reorganizations](https://mlsanity.com/paper/2202.07800/) (2022)
- [AdaViT: Adaptive Tokens for Efficient Vision Transformer](https://mlsanity.com/paper/2112.07658/) (2021)
- [Deep Learning Enabled Semantic Communication Systems](https://mlsanity.com/paper/2006.10685/) (2020)
- [Edge Intelligence: Paving the Last Mile of Artificial Intelligence with Edge Computing](https://mlsanity.com/paper/1905.10083/) (2019)
- [Adaptive Token Sampling For Efficient Vision Transformers](https://mlsanity.com/paper/2111.15667/) (2021)
- [DivPrune: Diversity-based Visual Token Pruning for Large Multimodal Models](https://mlsanity.com/paper/2503.02175/) (2025)
- [Beyond Transmitting Bits: Context, Semantics, and Task-Oriented Communications](https://mlsanity.com/paper/2207.09353/) (2022)
- [Coverage-centric Coreset Selection for High Pruning Rates](https://mlsanity.com/paper/2210.15809/) (2022)
- [Split Computing and Early Exiting for Deep Learning Applications: Survey and Research Challenges](https://mlsanity.com/paper/2103.04505/) (2021)
- [SPINN: Synergistic Progressive Inference of Neural Networks over Device and Cloud](https://mlsanity.com/paper/2008.06402/) (2020)
- [6G Networks: Beyond Shannon Towards Semantic and Goal-Oriented Communications](https://mlsanity.com/paper/2011.14844/) (2020)
- [Learning Task-Oriented Communication for Edge Inference: An Information Bottleneck Approach](https://mlsanity.com/paper/2102.04170/) (2021)
- [BottleNet++: An End-to-End Approach for Feature Compression in Device-Edge Co-Inference Systems](https://mlsanity.com/paper/1910.14315/) (2019)
- [Semantic Communications for Future Internet: Fundamentals, Applications, and Challenges](https://mlsanity.com/paper/2207.00427/) (2022)
- [Less Data, More Knowledge: Building Next Generation Semantic Communication Networks](https://mlsanity.com/paper/2211.14343/) (2022)
- [Communication-Computation Trade-Off in Resource-Constrained Edge Inference](https://mlsanity.com/paper/2006.02166/) (2020)
- [Beyond Attentive Tokens: Incorporating Token Importance and Diversity for Efficient Vision Transformers](https://mlsanity.com/paper/2211.11315/) (2022)
- [Bandwidth-Agile Image Transmission with Deep Joint Source-Channel Coding](https://mlsanity.com/paper/2009.12480/) (2020)
- [Adaptive Semantic Token Selection for AI-native Goal-oriented Communications](https://mlsanity.com/paper/2405.02330/) (2024)
- [Attention-aware Semantic Communications for Collaborative Inference](https://mlsanity.com/paper/2404.07217/) (2024)
- [AgilePruner: An Empirical Study of Attention and Diversity for Adaptive Visual Token Pruning in Large Vision-Language Models](https://mlsanity.com/paper/2603.01236/) (2026)
- [Vision Transformer-based Semantic Communications With Importance-Aware Quantization](https://mlsanity.com/paper/2412.06038/) (2024)
- [Optimal Fronthaul Compression for Synchronization in the Uplink of Cloud Radio Access Networks](https://mlsanity.com/paper/1510.01545/) (2015)

## Abstract (arXiv)

Edge-cloud hybrid inference offloads difficult inputs to a powerful remote model, but the uplink channel imposes hard per-request constraints on the number of bits that can be transmitted. We show that selecting transmitted content based solely on attention-based importance, the standard approach in collaborative inference, is inherently limited under hard budgets. Two findings support this claim. First, replacing high-importance units with low-importance but complementary ones improves server accuracy. This shows that what matters is not individual importance but how well the transmitted set covers diverse aspects of the input. Second, spatially uniform selection without any content information achieves competitive accuracy at moderate budgets. This confirms that spatial coverage alone carries independent value. Based on this analysis, we propose SAGE (Semantic Attention-Guided Evidence), a principled, training-free method that combines importance filtering with embedding-diversity sampling. SAGE achieves 93% of the server ceiling in offloaded accuracy while transmitting fewer than half of the available evidence units on ImageNet-1K, substantially outperforming importance-only composition.