# OrthoRank: Token Selection via Sink Token Orthogonality for Efficient LLM inference

arXiv: https://arxiv.org/abs/2507.03865 · HTML: https://mlsanity.com/paper/2507.03865/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Placed in KV cache optimization, one of 12 directions in Efficient LLM inference and serving.
- Builds on 35 of its 41 references that are in the map.
- First posted on arXiv 5 Jul 2025.

## Placed in

- Efficient LLM inference and serving / [KV cache optimization](https://mlsanity.com/inference/kv-cache-optimization/)

## Builds on (most cited first, 30 of 35)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge](https://mlsanity.com/paper/1803.05457/) (2018)
- [HellaSwag: Can a Machine Really Finish Your Sentence?](https://mlsanity.com/paper/1905.07830/) (2019)
- [Mistral 7B](https://mlsanity.com/paper/2310.06825/) (2023)
- [Pointer Sentinel Mixture Models](https://mlsanity.com/paper/1609.07843/) (2016)
- [PIQA: Reasoning about Physical Commonsense in Natural Language](https://mlsanity.com/paper/1911.11641/) (2019)
- [Efficient Streaming Language Models with Attention Sinks](https://mlsanity.com/paper/2309.17453/) (2023)
- [Mixtral of Experts](https://mlsanity.com/paper/2401.04088/) (2024)
- [LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding](https://mlsanity.com/paper/2308.14508/) (2023)
- [BloombergGPT: A Large Language Model for Finance](https://mlsanity.com/paper/2303.17564/) (2023)
- [PyramidKV: Dynamic KV Cache Compression based on Pyramidal Information Funneling](https://mlsanity.com/paper/2406.02069/) (2024)
- [ShortGPT: Layers in Large Language Models are More Redundant Than You Expect](https://mlsanity.com/paper/2403.03853/) (2024)
- [Confident Adaptive Language Modeling](https://mlsanity.com/paper/2207.07061/) (2022)
- [LayerSkip: Enabling Early Exit Inference and Self-Speculative Decoding](https://mlsanity.com/paper/2404.16710/) (2024)
- [BioMistral: A Collection of Open-Source Pretrained Large Language Models for Medical Domains](https://mlsanity.com/paper/2402.10373/) (2024)
- [Massive Activations in Large Language Models](https://mlsanity.com/paper/2402.17762/) (2024)
- [DuoAttention: Efficient Long-Context LLM Inference with Retrieval and Streaming Heads](https://mlsanity.com/paper/2410.10819/) (2024)
- [Mixture-of-Depths: Dynamically allocating compute in transformer-based language models](https://mlsanity.com/paper/2404.02258/) (2024)
- [Using an LLM to Help With Code Understanding](https://mlsanity.com/paper/2307.08177/) (2023)
- [When Attention Sink Emerges in Language Models: An Empirical View](https://mlsanity.com/paper/2410.10781/) (2024)
- [RazorAttention: Efficient KV Cache Compression Through Retrieval Heads](https://mlsanity.com/paper/2407.15891/) (2024)
- [SLEB: Streamlining LLMs through Redundancy Verification and Elimination of Transformer Blocks](https://mlsanity.com/paper/2402.09025/) (2024)
- [SkipDecode: Autoregressive Skip Decoding with Batching and Caching for Efficient LLM Inference](https://mlsanity.com/paper/2307.02628/) (2023)
- [Shortened LLaMA: Depth Pruning for Large Language Models with Comparison of Retraining Methods](https://mlsanity.com/paper/2402.02834/) (2024)
- [Fast and Robust Early-Exiting Framework for Autoregressive Language Models with Synchronized Parallel Decoding](https://mlsanity.com/paper/2310.05424/) (2023)
- [Unveiling and Harnessing Hidden Attention Sinks: Enhancing Large Language Models without Training through Attention Calibration](https://mlsanity.com/paper/2406.15765/) (2024)
- [EE-LLM: Large-Scale Training and Inference of Early-Exit Large Language Models with 3D Parallelism](https://mlsanity.com/paper/2312.04916/) (2023)
- [MathChat: Converse to Tackle Challenging Math Problems with LLM Agents](https://mlsanity.com/paper/2306.01337/) (2023)

## Built on by (most cited first, 8 of 8)

- [The Spike, the Sparse and the Sink: Anatomy of Massive Activations and Attention Sinks](https://mlsanity.com/paper/2603.05498/) (2026)
- [Attention Sink Forges Native MoE in Attention Layers: Sink-Aware Training to Address Head Collapse](https://mlsanity.com/paper/2602.01203/) (2026)
- [Surgery: Mitigating Harmful Fine-Tuning for Large Language Models via Attention Sink](https://mlsanity.com/paper/2602.05228/) (2026)
- [DTRNet: Dynamic Token Routing Network to Reduce Quadratic Costs in Transformers](https://mlsanity.com/paper/2509.00925/) (2025)
- [Geometric Analysis of Token Selection in Multi-Head Attention](https://mlsanity.com/paper/2602.01893/) (2026)
- [Token Sparse Attention: Efficient Long-Context Inference with Interleaved Token Selection](https://mlsanity.com/paper/2602.03216/) (2026)
- [Contribution Weights: A Geometrical Analysis of Self-Attention Transformers](https://mlsanity.com/paper/2606.07604/) (2026)
- [Retrieval Capacity of Self-Attention Under Competition](https://mlsanity.com/paper/2609.37879/) (2026)

## Abstract (arXiv)

Attention mechanisms are central to the success of large language models (LLMs), enabling them to capture intricate token dependencies and implicitly assign importance to each token. Recent studies have revealed the sink token, which receives disproportionately high attention despite their limited semantic role. In this paper, we first expand the relationship between the sink token and other tokens, moving beyond attention to explore their similarity in hidden states, considering the layer depth. We observe that as the layers get deeper, the cosine similarity between the normalized hidden states of the sink token and those of other tokens increases, and that the normalized hidden states of the sink token exhibit negligible changes. These imply that other tokens consistently are directed toward the sink token throughout the layers. Next, we propose a dynamic token selection method, called OrthoRank, using these findings to select important tokens. Specifically, in a certain layer, we define token importance by the speed at which the token moves toward the sink token. This is converted into orthogonality with the sink token, meaning that tokens that are more orthogonal to the sink token are assigned greater importance. Finally, through extensive experiments, we demonstrated that our method results in lower perplexity and higher zero-shot accuracy compared to layer pruning methods at the same sparsity ratio with comparable throughput, while also achieving superior performance on LongBench.