# From Saliency to Discriminability: Rank-Preserving Visual Token Pruning for VLM Rerankers

arXiv: https://arxiv.org/abs/2609.00667 · HTML: https://mlsanity.com/paper/2609.00667/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Visual token pruning and efficiency, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 30 of its 43 references that are in the map.
- First posted on arXiv 1 Sep 2026.

## Placed in

- Multimodal LLMs / vision-language models / [Visual token pruning and efficiency](https://mlsanity.com/mllm/visual-token-pruning-and-efficiency/)
- Efficient LLM inference and serving / [Visual token pruning in multimodal models](https://mlsanity.com/inference/visual-token-pruning-in-multimodal-models/)

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

- [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014)
- [Qwen3-VL Technical Report](https://mlsanity.com/paper/2511.21631/) (2025)
- [Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling](https://mlsanity.com/paper/2412.05271/) (2024)
- [Deep Visual-Semantic Alignments for Generating Image Descriptions](https://mlsanity.com/paper/1412.2306/) (2014)
- [Qwen3-VL-Embedding and Qwen3-VL-Reranker: A Unified Framework for State-of-the-Art Multimodal Retrieval and Ranking](https://mlsanity.com/paper/2601.04720/) (2026)
- [MM-Embed: Universal Multimodal Retrieval with Multimodal LLMs](https://mlsanity.com/paper/2411.02571/) (2024)
- [Stop Looking for Important Tokens in Multimodal Language Models: Duplication Matters More](https://mlsanity.com/paper/2502.11494/) (2025)
- [Devils in Middle Layers of Large Vision-Language Models: Interpreting, Detecting and Mitigating Object Hallucinations via Attention Lens](https://mlsanity.com/paper/2411.16724/) (2024)
- [LamRA: Large Multimodal Model as Your Advanced Retrieval Assistant](https://mlsanity.com/paper/2412.01720/) (2024)
- [Attention in Large Language Models Yields Efficient Zero-Shot Re-Rankers](https://mlsanity.com/paper/2410.02642/) (2024)
- [UniME-V2: MLLM-as-a-Judge for Universal Multimodal Embedding Learning](https://mlsanity.com/paper/2510.13515/) (2025)
- [Feather the Throttle: Revisiting Visual Token Pruning for Vision-Language Model Acceleration](https://mlsanity.com/paper/2412.13180/) (2024)
- [Inference Optimal VLMs Need Fewer Visual Tokens and More Parameters](https://mlsanity.com/paper/2411.03312/) (2024)
- [HiDrop: Hierarchical Vision Token Reduction in MLLMs via Late Injection, Concave Pyramid Pruning, and Early Exit](https://mlsanity.com/paper/2602.23699/) (2026)
- [Are We Using the Right Benchmark: An Evaluation Framework for Visual Token Compression Methods](https://mlsanity.com/paper/2510.07143/) (2025)
- [TrimTokenator-LC: Towards Adaptive Visual Token Pruning for Large Multimodal Models with Long Contexts](https://mlsanity.com/paper/2512.22748/) (2025)
- [When Token Pruning is Worse than Random: Understanding Visual Token Information in VLLMs](https://mlsanity.com/paper/2512.07580/) (2025)
- [Beyond Intermediate States: Explaining Visual Redundancy through Language](https://mlsanity.com/paper/2503.20540/) (2025)
- [CARES: Context-Aware Resolution Selector for VLMs](https://mlsanity.com/paper/2510.19496/) (2025)
- [IDPruner: Harmonizing Importance and Diversity in Visual Token Pruning for MLLMs](https://mlsanity.com/paper/2602.13315/) (2026)
- [Gradient-Stable Attention Heads Signal LLM Correctness](https://mlsanity.com/paper/2602.13699/) (2026)
- [OSCAR: Online Soft Compression And Reranking](https://mlsanity.com/paper/2504.07109/) (2025)
- [Chain-of-Thought Re-ranking for Image Retrieval Tasks](https://mlsanity.com/paper/2509.14746/) (2025)
- [ToFu: Visual Tokens Reduction via Fusion for Multi-modal, Multi-patch, Multi-image Task](https://mlsanity.com/paper/2503.04444/) (2025)
- [When Vision Meets Texts in Listwise Reranking](https://mlsanity.com/paper/2601.20623/) (2026)
- [Look Where It Matters: High-Resolution Crops Retrieval for Efficient VLMs](https://mlsanity.com/paper/2603.16932/) (2026)
- [Where Reliability Lives in Vision-Language Models: A Mechanistic Study of Attention, Hidden States, and Causal Circuits](https://mlsanity.com/paper/2605.08200/) (2026)
- [Very Efficient Listwise Multimodal Reranking for Long Documents](https://mlsanity.com/paper/2605.11864/) (2026)
- [CAPA: Contribution-Aware Pruning and FFN Approximation for Efficient Large Vision-Language Models](https://mlsanity.com/paper/2602.00247/) (2026)
- [Evading Visual Aphasia: Contrastive Adaptive Semantic Token Pruning for Vision-Language Models](https://mlsanity.com/paper/2605.09429/) (2026)

## Abstract (arXiv)

Large vision-language models used as listwise rerankers must jointly process visual tokens from tens of candidates per query, making token pruning essential for practical deployment. Existing pruning methods retain tokens by attention saliency, yet we show that saliency is systematically misaligned with ranking contribution: visually prominent tokens often capture order-neutral patterns shared across candidates. This mismatch is layer-dependent: saliency becomes informative only where attention is concentrated, and normalized attention entropy diagnoses the reliability shift (Pearson r=0.87). We propose RaDiCal (Rank-Discriminative Calibration), a training-free framework that uses normalized attention entropy to decide when saliency can be trusted, fusing it with an attention-free rank-discriminative prior and selecting pruning layers from the same trust landscape. Across three retrieval benchmarks and multiple VLM architectures, RaDiCal matches Dense MRR@10 on Flickr30K and surpasses it on MSCOCO at a 20% token budget, ranks first among all pruning methods on FashionIQ, and holds within 1.2 pp on Flickr30K and MSCOCO at 10% retention. It cuts FLOPs by 39--45% and delivers 1.28--1.45× measured speedups across two VLM architectures without dataset-specific retuning.