# Lifted State Hypothesis in Large Language Models

arXiv: https://arxiv.org/abs/2607.19360 · HTML: https://mlsanity.com/paper/2607.19360/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 28 of its 57 references that are in the map.
- First posted on arXiv 2 Jun 2026.

## Builds on (most cited first, 28 of 28)

- [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)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Overcoming catastrophic forgetting in neural networks](https://mlsanity.com/paper/1612.00796/) (2016)
- [Locating and Editing Factual Associations in GPT](https://mlsanity.com/paper/2202.05262/) (2022)
- [Language Models as Knowledge Bases?](https://mlsanity.com/paper/1909.01066/) (2019)
- [Gradient Episodic Memory for Continual Learning](https://mlsanity.com/paper/1706.08840/) (2017)
- [Understanding Black-box Predictions via Influence Functions](https://mlsanity.com/paper/1703.04730/) (2017)
- [Calibrate Before Use: Improving Few-Shot Performance of Language Models](https://mlsanity.com/paper/2102.09690/) (2021)
- [Transformer Feed-Forward Layers Are Key-Value Memories](https://mlsanity.com/paper/2012.14913/) (2020)
- [Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!](https://mlsanity.com/paper/2310.03693/) (2023)
- [Towards Understanding Sycophancy in Language Models](https://mlsanity.com/paper/2310.13548/) (2023)
- [How Much Knowledge Can You Pack Into the Parameters of a Language Model?](https://mlsanity.com/paper/2002.08910/) (2020)
- [Editing Factual Knowledge in Language Models](https://mlsanity.com/paper/2104.08164/) (2021)
- [In-Context Learning Creates Task Vectors](https://mlsanity.com/paper/2310.15916/) (2023)
- [Theoretical Limitations of Self-Attention in Neural Sequence Models](https://mlsanity.com/paper/1906.06755/) (2019)
- [Evaluating the Ripple Effects of Knowledge Editing in Language Models](https://mlsanity.com/paper/2307.12976/) (2023)
- [Large Language Models as Commonsense Knowledge for Large-Scale Task Planning](https://mlsanity.com/paper/2305.14078/) (2023)
- [LoRA Fine-tuning Efficiently Undoes Safety Training in Llama 2-Chat 70B](https://mlsanity.com/paper/2310.20624/) (2023)
- [Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet](https://mlsanity.com/paper/2605.29358/) (2026)
- [The Dormant Neuron Phenomenon in Deep Reinforcement Learning](https://mlsanity.com/paper/2302.12902/) (2023)
- [Capturing Failures of Large Language Models via Human Cognitive Biases](https://mlsanity.com/paper/2202.12299/) (2022)
- [Transformers represent belief state geometry in their residual stream](https://mlsanity.com/paper/2405.15943/) (2024)
- [BeliefBank: Adding Memory to a Pre-Trained Language Model for a Systematic Notion of Belief](https://mlsanity.com/paper/2109.14723/) (2021)
- [Belief Revision: The Adaptability of Large Language Models Reasoning](https://mlsanity.com/paper/2406.19764/) (2024)
- [Why Does New Knowledge Create Messy Ripple Effects in LLMs?](https://mlsanity.com/paper/2407.12828/) (2024)
- [Coarse-to-Fine Grounded Memory for LLM Agent Planning](https://mlsanity.com/paper/2508.15305/) (2025)
- [Patterns Over Principles: The Fragility of Inductive Reasoning in LLMs under Noisy Observations](https://mlsanity.com/paper/2502.16169/) (2025)

## Abstract (arXiv)

Large language models (LLMs) adapt rapidly through fine-tuning and in-context learning, yet it remains unclear which inputs they treat as the same case and why their predictions change together. We introduce the Lifted State Hypothesis. Under fixed model parameters, context scope, and target computation, samples indistinguishable in their observed computation-relevant behavior form a computation-relative type. We hypothesize that compatible episodes activate a reusable latent component---a lifted state---that supports the target computation. State reuse enables type-level generalization but creates a non-monotonic revision problem. When later evidence distinguishes a subtype, revising a state still shared with its parent may affect members whose predictions should remain unchanged. The model must separate the subtype through rerouting, a new state, or input-specific compensation. We formalize this relation between generalization and revision. We introduce the NMR-Type Dataset to evaluate LLMs. The dataset first supports a broad modulo rule. It then provides conflicting supervision for a withheld subtype while replaying earlier examples. Across full fine-tuning, LoRA, and in-context learning, models often generalize the broad rule to the subtype but fail to localize its later revision. These results provide behavioral evidence consistent with the hypothesis and motivate further study of lifted-state formation and revision.