# Token pruning and merging for ViTs

Area: [Efficient LLM inference and serving](https://mlsanity.com/inference/index.md). HTML: https://mlsanity.com/inference/token-pruning-and-merging-for-vits/

## Key facts

- Token pruning and merging for ViTs: 108 papers on arXiv since 2021 (data through 14 Oct 2026).
- 75 of its 108 mapped papers (69%) cite “Token Merging: Your ViT But Faster” (2022, arXiv 2210.09461).
- 50 of its 108 mapped papers (46%) cite “SPViT: Enabling Faster Vision Transformers via Soft Token Pruning” (2021, arXiv 2112.13890).
- 5 new papers 24 Sep–14 Oct.

## Most cited here

- [Token Merging: Your ViT But Faster](https://mlsanity.com/paper/2210.09461/) (2022), cited by 75 papers here
- [SPViT: Enabling Faster Vision Transformers via Soft Token Pruning](https://mlsanity.com/paper/2112.13890/) (2021), cited by 50 papers here
- [Evo-ViT: Slow-Fast Token Evolution for Dynamic Vision Transformer](https://mlsanity.com/paper/2108.01390/) (2021), cited by 38 papers here
- [Token Fusion: Bridging the Gap between Token Pruning and Token Merging](https://mlsanity.com/paper/2312.01026/) (2023), cited by 29 papers here
- [DiffRate : Differentiable Compression Rate for Efficient Vision Transformers](https://mlsanity.com/paper/2305.17997/) (2023), cited by 22 papers here
- [Joint Token Pruning and Squeezing Towards More Aggressive Compression of Vision Transformers](https://mlsanity.com/paper/2304.10716/) (2023), cited by 21 papers here
- [Learned Token Pruning for Transformers](https://mlsanity.com/paper/2107.00910/) (2021), cited by 17 papers here
- [Zero-TPrune: Zero-Shot Token Pruning through Leveraging of the Attention Graph in Pre-Trained Transformers](https://mlsanity.com/paper/2305.17328/) (2023), cited by 16 papers here
- [Revisiting Token Pruning for Object Detection and Instance Segmentation](https://mlsanity.com/paper/2306.07050/) (2023), cited by 15 papers here
- [Dynamic Token Pruning in Plain Vision Transformers for Semantic Segmentation](https://mlsanity.com/paper/2308.01045/) (2023), cited by 14 papers here

## New (24 Sep–14 Oct)

- [Scaling Vision Transformers Without Training: Prune-and-Merge Token Compression](https://mlsanity.com/paper/2610.18039/) (2026-10-13)
- [Later Is Better: Token Reduction for ViTs Under Distribution Shift](https://mlsanity.com/paper/2610.07758/) (2026-10-06)
- [Less Context, Better Geometry: Masked Geometric Encoder for Robust 3D Foundation Models](https://mlsanity.com/paper/2610.06813/) (2026-10-05)
- [DORA: Dynamic Online Reinforcement Agent for Token Pruning in Vision Transformers](https://mlsanity.com/paper/2609.34325/) (2026-09-28)
- [FlashLoop: Fast and Memory-Efficient Looped Transformers via Lazy Updates](https://mlsanity.com/paper/2609.29812/) (2026-09-24)