# Memory Retrieval and Consolidation in Large Language Models through Function Tokens

arXiv: https://arxiv.org/abs/2510.08203 · HTML: https://mlsanity.com/paper/2510.08203/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Builds on 30 of its 49 references that are in the map.
- First posted on arXiv 9 Oct 2025.

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

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Chain-of-Thought Prompting Elicits Reasoning in Large Language Models](https://mlsanity.com/paper/2201.11903/) (2022)
- [Large Language Models are Zero-Shot Reasoners](https://mlsanity.com/paper/2205.11916/) (2022)
- [Finetuned Language Models Are Zero-Shot Learners](https://mlsanity.com/paper/2109.01652/) (2021)
- [WebGPT: Browser-assisted question-answering with human feedback](https://mlsanity.com/paper/2112.09332/) (2021)
- [Sparse Autoencoders Find Highly Interpretable Features in Language Models](https://mlsanity.com/paper/2309.08600/) (2023)
- [Transformer Feed-Forward Layers Are Key-Value Memories](https://mlsanity.com/paper/2012.14913/) (2020)
- [Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small](https://mlsanity.com/paper/2211.00593/) (2022)
- [The FineWeb Datasets: Decanting the Web for the Finest Text Data at Scale](https://mlsanity.com/paper/2406.17557/) (2024)
- [Steering Llama 2 via Contrastive Activation Addition](https://mlsanity.com/paper/2312.06681/) (2023)
- [An overview of gradient descent optimization algorithms](https://mlsanity.com/paper/1609.04747/) (2016)
- [Toy Models of Superposition](https://mlsanity.com/paper/2209.10652/) (2022)
- [Phi-4 Technical Report](https://mlsanity.com/paper/2412.08905/) (2024)
- [Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoning](https://mlsanity.com/paper/2506.01939/) (2025)
- [Scaling and evaluating sparse autoencoders](https://mlsanity.com/paper/2406.04093/) (2024)
- [LIMO: Less is More for Reasoning](https://mlsanity.com/paper/2502.03387/) (2025)
- [Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2](https://mlsanity.com/paper/2408.05147/) (2024)
- [Persona Vectors: Monitoring and Controlling Character Traits in Language Models](https://mlsanity.com/paper/2507.21509/) (2025)
- [In-Context Learning Creates Task Vectors](https://mlsanity.com/paper/2310.15916/) (2023)
- [Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoders](https://mlsanity.com/paper/2407.14435/) (2024)
- [Massive Activations in Large Language Models](https://mlsanity.com/paper/2402.17762/) (2024)
- [Improving Dictionary Learning with Gated Sparse Autoencoders](https://mlsanity.com/paper/2404.16014/) (2024)
- [LLMs Can Easily Learn to Reason from Demonstrations Structure, not content, is what matters!](https://mlsanity.com/paper/2502.07374/) (2025)
- [SAEBench: A Comprehensive Benchmark for Sparse Autoencoders in Language Model Interpretability](https://mlsanity.com/paper/2503.09532/) (2025)
- [Circuit Component Reuse Across Tasks in Transformer Language Models](https://mlsanity.com/paper/2310.08744/) (2023)
- [Convergent Linear Representations of Emergent Misalignment](https://mlsanity.com/paper/2506.11618/) (2025)
- [SepLLM: Accelerate Large Language Models by Compressing One Segment into One Separator](https://mlsanity.com/paper/2412.12094/) (2024)
- [All for One: LLMs Solve Mental Math at the Last Token With Information Transferred From Other Tokens](https://mlsanity.com/paper/2509.09650/) (2025)

## Built on by (most cited first, 4 of 4)

- [Memory in the Age of AI Agents](https://mlsanity.com/paper/2512.13564/) (2025)
- [Pseudo-Unification: Entropy Probing Reveals Divergent Information Patterns in Unified Multimodal Models](https://mlsanity.com/paper/2604.10949/) (2026)
- [No More K-means: Single-Stage Sparse Coding for Efficient Multi-Vector Retrieval](https://mlsanity.com/paper/2605.30120/) (2026)
- [Attention Degradation, Function Token Anchoring, and the Limits of Attention-Based Intervention in Large Language Models](https://mlsanity.com/paper/2607.20524/) (2026)

## Abstract (arXiv)

The remarkable success of large language models (LLMs) stems from their ability to consolidate vast amounts of knowledge into the memory during pre-training and to retrieve it from the memory during inference, enabling advanced capabilities such as knowledge memorization, instruction-following and reasoning. However, the mechanisms of memory retrieval and consolidation in LLMs remain poorly understood. In this paper, we propose the function token hypothesis to explain the workings of LLMs: During inference, function tokens activate the most predictive features from context and govern next token prediction (memory retrieval). During pre-training, predicting the next tokens (usually content tokens) that follow function tokens increases the number of learned features of LLMs and updates the model parameters (memory consolidation). Function tokens here roughly correspond to function words in linguistics, including punctuation marks, articles, prepositions, and conjunctions, in contrast to content tokens. We provide extensive experimental evidence supporting this hypothesis. Using bipartite graph analysis, we show that a small number of function tokens activate the majority of features. Case studies further reveal how function tokens activate the most predictive features from context to direct next token prediction. We also find that during pre-training, the training loss is dominated by predicting the next content tokens following function tokens, which forces the function tokens to select the most predictive features from context.