# Do Dynamic Routers Need Memory? HeRo: History-Aware Routing for Efficient LLM Inference

arXiv: https://arxiv.org/abs/2609.08189 · HTML: https://mlsanity.com/paper/2609.08189/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Serving systems and request scheduling, one of 12 directions in Efficient LLM inference and serving.
- Builds on 22 of its 30 references that are in the map.
- First posted on arXiv 8 Sep 2026.

## Placed in

- Efficient LLM inference and serving / [Serving systems and request scheduling](https://mlsanity.com/inference/serving-systems-and-request-scheduling/)

## Builds on (most cited first, 22 of 22)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge](https://mlsanity.com/paper/1803.05457/) (2018)
- [Parameter-Efficient Transfer Learning for NLP](https://mlsanity.com/paper/1902.00751/) (2019)
- [HellaSwag: Can a Machine Really Finish Your Sentence?](https://mlsanity.com/paper/1905.07830/) (2019)
- [PIQA: Reasoning about Physical Commonsense in Natural Language](https://mlsanity.com/paper/1911.11641/) (2019)
- [Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention](https://mlsanity.com/paper/2006.16236/) (2020)
- [Tulu 3: Pushing Frontiers in Open Language Model Post-Training](https://mlsanity.com/paper/2411.15124/) (2024)
- [ShortGPT: Layers in Large Language Models are More Redundant Than You Expect](https://mlsanity.com/paper/2403.03853/) (2024)
- [DeeBERT: Dynamic Early Exiting for Accelerating BERT Inference](https://mlsanity.com/paper/2004.12993/) (2020)
- [BERT Loses Patience: Fast and Robust Inference with Early Exit](https://mlsanity.com/paper/2006.04152/) (2020)
- [FastBERT: a Self-distilling BERT with Adaptive Inference Time](https://mlsanity.com/paper/2004.02178/) (2020)
- [LaCo: Large Language Model Pruning via Layer Collapse](https://mlsanity.com/paper/2402.11187/) (2024)
- [Transformers represent belief state geometry in their residual stream](https://mlsanity.com/paper/2405.15943/) (2024)
- [Adaptive Layer-skipping in Pre-trained LLMs](https://mlsanity.com/paper/2503.23798/) (2025)
- [Cross-Layer Retrospective Retrieving via Layer Attention](https://mlsanity.com/paper/2302.03985/) (2023)
- [Layerwise Recurrent Router for Mixture-of-Experts](https://mlsanity.com/paper/2408.06793/) (2024)
- [Accelerating Inference in Large Language Models with a Unified Layer Skipping Strategy](https://mlsanity.com/paper/2404.06954/) (2024)
- [Dr.LLM: Dynamic Layer Routing in LLMs](https://mlsanity.com/paper/2510.12773/) (2025)
- [SkipGPT: Dynamic Layer Pruning Reinvented with Token Awareness and Module Decoupling](https://mlsanity.com/paper/2506.04179/) (2025)
- [What Layers When: Learning to Skip Compute in LLMs with Residual Gates](https://mlsanity.com/paper/2510.13876/) (2025)
- [You Do Not Fully Utilize Transformer's Representation Capacity](https://mlsanity.com/paper/2502.09245/) (2025)

## Abstract (arXiv)

Dynamic layer routing reduces the inference cost of Large Language Models (LLMs) by learning to skip layers for individual tokens. Existing methods, however, treat each routing decision as a local operation conditioned solely on the current hidden state which is a formulation that overlooks the sequential, path-dependent nature of routing across depth: earlier decisions shape the representations seen by downstream routers, and the layer-usage objective couples all decisions jointly. We propose History-Aware Routing (HeRo), a dynamic routing framework that resolves this mismatch by introducing a router memory mechanism to maintain an explicit routing state across model depth. The memory is constructed via linear attention, incrementally aggregating preceding routing scores and their induced residual updates into a compact history representation. At each routed layer, the router conditions jointly on this accumulated state and the current hidden representation to select the executed branch. Instantiated for token-wise FFN routing, HeRo trains only lightweight routers and adapters on a frozen backbone, requiring no modification to pretrained parameters. Across Llama 3.1-8B, Llama 2-7B, and Llama 2-13B, HeRo consistently achieves the highest aggregate performance retention among ten baselines. On Llama 3.1-8B, it bypasses 26.87% of model parameters while achieving 100.24% of dense model performance across seven benchmarks, and retains 97.01% while bypassing 38.82% of model parameters under a tighter computation budget. Ablation studies confirm that removing routing history consistently degrades performance, most notably on multistep reasoning and code generation, validating that explicit routing memory enables more accurate and adaptive dynamic routing than solely conditioning on hidden state.