# Why Training-Free Token Reduction Collapses: The Inherent Instability of Pairwise Scoring Signals

arXiv: https://arxiv.org/abs/2604.16745 · HTML: https://mlsanity.com/paper/2604.16745/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Token pruning and merging for ViTs, one of 12 directions in Efficient LLM inference and serving.
- Builds on 42 of its 50 references that are in the map.
- First posted on arXiv 17 Apr 2026.

## Placed in

- Efficient LLM inference and serving / [Token pruning and merging for ViTs](https://mlsanity.com/inference/token-pruning-and-merging-for-vits/)

## Builds on (most cited first, 30 of 42)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [ImageNet Large Scale Visual Recognition Challenge](https://mlsanity.com/paper/1409.0575/) (2014)
- [Masked Autoencoders Are Scalable Vision Learners](https://mlsanity.com/paper/2111.06377/) (2021)
- [Training data-efficient image transformers & distillation through attention](https://mlsanity.com/paper/2012.12877/) (2020)
- [FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness](https://mlsanity.com/paper/2205.14135/) (2022)
- [Benchmarking Neural Network Robustness to Common Corruptions and Perturbations](https://mlsanity.com/paper/1903.12261/) (2019)
- [A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks](https://mlsanity.com/paper/1610.02136/) (2016)
- [Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention](https://mlsanity.com/paper/2006.16236/) (2020)
- [A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attacks](https://mlsanity.com/paper/1807.03888/) (2018)
- [ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness](https://mlsanity.com/paper/1811.12231/) (2018)
- [VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training](https://mlsanity.com/paper/2203.12602/) (2022)
- [Token Merging: Your ViT But Faster](https://mlsanity.com/paper/2210.09461/) (2022)
- [Learning Robust Global Representations by Penalizing Local Predictive Power](https://mlsanity.com/paper/1905.13549/) (2019)
- [Vision Transformers Need Registers](https://mlsanity.com/paper/2309.16588/) (2023)
- [DynamicViT: Efficient Vision Transformers with Dynamic Token Sparsification](https://mlsanity.com/paper/2106.02034/) (2021)
- [Do Vision Transformers See Like Convolutional Neural Networks?](https://mlsanity.com/paper/2108.08810/) (2021)
- [Measuring Robustness to Natural Distribution Shifts in Image Classification](https://mlsanity.com/paper/2007.00644/) (2020)
- [Intriguing Properties of Vision Transformers](https://mlsanity.com/paper/2105.10497/) (2021)
- [DeiT III: Revenge of the ViT](https://mlsanity.com/paper/2204.07118/) (2022)
- [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)
- [Evo-ViT: Slow-Fast Token Evolution for Dynamic Vision Transformer](https://mlsanity.com/paper/2108.01390/) (2021)
- [SPViT: Enabling Faster Vision Transformers via Soft Token Pruning](https://mlsanity.com/paper/2112.13890/) (2021)
- [Efficient Deep Learning: A Survey on Making Deep Learning Models Smaller, Faster, and Better](https://mlsanity.com/paper/2106.08962/) (2021)
- [Patch Slimming for Efficient Vision Transformers](https://mlsanity.com/paper/2106.02852/) (2021)
- [Token Fusion: Bridging the Gap between Token Pruning and Token Merging](https://mlsanity.com/paper/2312.01026/) (2023)

## Abstract (arXiv)

Training-free token reduction methods for Vision Transformers (ToMe, ToFu, PiToMe, and MCTF) employ different scoring mechanisms, yet they share a closely matched cliff-like collapse at high compression. This paper explains why. We develop a diagnostic framework with two tools, ranking consistency ρs and off-diagonal correlation ρoff, that decomposes the collapse into (1)a signal-agnostic error amplifier inherent to layer-wise reduction, predicting convex Pareto curves and rcrit propto 1/L; and (2)shared reliance on pairwise similarity signals whose ranking consistency degrades from ρs=0.88 to 0.27 in deep layers. Pairwise rankings are inherently unstable (O(Np2) joint perturbations) while unary signals enjoy greater stability (O(Np) perturbations, CLT). From three design principles derived from this diagnosis, we construct CATIS as a constructive validation: unary signals raise the trigger threshold, triage suppresses the gain. On ViT-Large at 63% FLOPs reduction, CATIS retains 96.9% of vanilla accuracy (81.0%) on ImageNet-1K where all baselines collapse to 43--65%.