# Learning Dynamic Evidence Routes for Vision Transformer Probing

arXiv: https://arxiv.org/abs/2605.00915 · HTML: https://mlsanity.com/paper/2605.00915/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 23 of its 25 references that are in the map.
- First posted on arXiv 29 Apr 2026.

## Builds on (most cited first, 23 of 23)

- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Masked Autoencoders Are Scalable Vision Learners](https://mlsanity.com/paper/2111.06377/) (2021)
- [Emerging Properties in Self-Supervised Vision Transformers](https://mlsanity.com/paper/2104.14294/) (2021)
- [Mamba: Linear-Time Sequence Modeling with Selective State Spaces](https://mlsanity.com/paper/2312.00752/) (2023)
- [Efficiently Modeling Long Sequences with Structured State Spaces](https://mlsanity.com/paper/2111.00396/) (2021)
- [BEiT: BERT Pre-Training of Image Transformers](https://mlsanity.com/paper/2106.08254/) (2021)
- [Sinkhorn Distances: Lightspeed Computation of Optimal Transportation Distances](https://mlsanity.com/paper/1306.0895/) (2013)
- [Deep Sets](https://mlsanity.com/paper/1703.06114/) (2017)
- [Understanding intermediate layers using linear classifier probes](https://mlsanity.com/paper/1610.01644/) (2016)
- [Vision Mamba: Efficient Visual Representation Learning with Bidirectional State Space Model](https://mlsanity.com/paper/2401.09417/) (2024)
- [Token Merging: Your ViT But Faster](https://mlsanity.com/paper/2210.09461/) (2022)
- [VMamba: Visual State Space Model](https://mlsanity.com/paper/2401.10166/) (2024)
- [NetVLAD: CNN architecture for weakly supervised place recognition](https://mlsanity.com/paper/1511.07247/) (2015)
- [HiPPO: Recurrent Memory with Optimal Polynomial Projections](https://mlsanity.com/paper/2008.07669/) (2020)
- [Focal Self-attention for Local-Global Interactions in Vision Transformers](https://mlsanity.com/paper/2107.00641/) (2021)
- [Learning Latent Permutations with Gumbel-Sinkhorn Networks](https://mlsanity.com/paper/1802.08665/) (2018)
- [Differentiable Ranks and Sorting using Optimal Transport](https://mlsanity.com/paper/1905.11885/) (2019)
- [Stochastic Optimization of Sorting Networks via Continuous Relaxations](https://mlsanity.com/paper/1903.08850/) (2019)
- [SoftSort: A Continuous Relaxation for the argsort Operator](https://mlsanity.com/paper/2006.16038/) (2020)
- [PPT: Token Pruning and Pooling for Efficient Vision Transformers](https://mlsanity.com/paper/2310.01812/) (2023)
- [Jigsaw-ViT: Learning Jigsaw Puzzles in Vision Transformer](https://mlsanity.com/paper/2207.11971/) (2022)
- [PuzzleTuning: Explicitly Bridge Pathological and Natural Image with Puzzles](https://mlsanity.com/paper/2311.06712/) (2023)

## Abstract (arXiv)

Probing frozen vision transformers typically uses permutation-invariant aggregation (GAP or texttt[CLS]), treating patch tokens as an unstructured set. Content-dependent probes such as self-attention are useful accuracy controls, but they do not expose a fixed token schedule or fixed position weights for auditing. We introduce SSMProbe, a probe of frozen vision transformers that learns a Sinkhorn patch route and reads it with a diagonal S4 decoder. The decoder's causal linear time-invariant (LTI) recurrence yields fixed position coefficients, exposing how patch arrangement interacts with sequential weighting. Held-out spatial predictability measures route organization beyond a location-bias null. On CUB and Cars, MAE, BEiT, DINOv2, and supervised ViT yield spatially organized routes with distinct readout profiles: DINOv2 is texttt[CLS]-dominant, while MAE benefits from learned routing over fixed scans. On MAE/ImageNet, Sinkhorn reaches 70.33% accuracy, 5.55 percentage points above the strongest fixed scan.