# Cross-Architecture Steering Transfer in Language Models: A Systematic Empirical Study

arXiv: https://arxiv.org/abs/2608.05164 · HTML: https://mlsanity.com/paper/2608.05164/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Activation steering methods, one of 13 directions in Mechanistic interpretability.
- Builds on 18 of its 26 references that are in the map.
- First posted on arXiv 26 May 2026.

## Placed in

- Mechanistic interpretability / [Activation steering methods](https://mlsanity.com/mechinterp/activation-steering-methods/)

## Builds on (most cited first, 18 of 18)

- [A Simple Framework for Contrastive Learning of Visual Representations](https://mlsanity.com/paper/2002.05709/) (2020)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [SQuAD: 100,000+ Questions for Machine Comprehension of Text](https://mlsanity.com/paper/1606.05250/) (2016)
- [UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction](https://mlsanity.com/paper/1802.03426/) (2018)
- [Character-level Convolutional Networks for Text Classification](https://mlsanity.com/paper/1509.01626/) (2015)
- [Similarity of Neural Network Representations Revisited](https://mlsanity.com/paper/1905.00414/) (2019)
- [Sparse Autoencoders Find Highly Interpretable Features in Language Models](https://mlsanity.com/paper/2309.08600/) (2023)
- [Representation Engineering: A Top-Down Approach to AI Transparency](https://mlsanity.com/paper/2310.01405/) (2023)
- [Hierarchical Neural Story Generation](https://mlsanity.com/paper/1805.04833/) (2018)
- [Scaling and evaluating sparse autoencoders](https://mlsanity.com/paper/2406.04093/) (2024)
- [Word Translation Without Parallel Data](https://mlsanity.com/paper/1710.04087/) (2017)
- [A Discourse-Aware Attention Model for Abstractive Summarization of Long Documents](https://mlsanity.com/paper/1804.05685/) (2018)
- [The Platonic Representation Hypothesis](https://mlsanity.com/paper/2405.07987/) (2024)
- [Do Wide and Deep Networks Learn the Same Things? Uncovering How Neural Network Representations Vary with Width and Depth](https://mlsanity.com/paper/2010.15327/) (2020)
- [Pile of Law: Learning Responsible Data Filtering from the Law and a 256GB Open-Source Legal Dataset](https://mlsanity.com/paper/2207.00220/) (2022)
- [Universal Neurons in GPT2 Language Models](https://mlsanity.com/paper/2401.12181/) (2024)
- [Towards Universality: Studying Mechanistic Similarity Across Language Model Architectures](https://mlsanity.com/paper/2410.06672/) (2024)
- [Quantifying Feature Space Universality Across Large Language Models via Sparse Autoencoders](https://mlsanity.com/paper/2410.06981/) (2024)

## Built on by (most cited first, 2 of 2)

- [Activation Steering Transfer to Agents: One Gain Ratio Does Not Identify Potency and Efficacy](https://mlsanity.com/paper/2607.09156/) (2026)
- [Do Emotion Concepts Generalize Across Sources, Modalities, and Architectures in Vision-Language Models?](https://mlsanity.com/paper/2609.34742/) (2026)

## Abstract (arXiv)

Independently trained large language models may develop shared internal representations of semantic concepts despite architectural differences -- but whether this geometric similarity has functional consequences for cross-model behavioural control remains untested. We present the first systematic evaluation of cross-model steering transfer and show that shared LLM geometry is functionally exploitable, conditionally: concept directions from one model can steer a different independently trained model when sufficient representational capacity exists. We study five open-weight models spanning three parameter scales (0.8B--8B) and two architectural lineages, training one Sparse Autoencoder per model across 15 semantic domains and testing alignment across all 20 directed model pairs. We observe a suggestive discontinuity near 1.7B parameters: at >= 1.7B scale, 47--49% of cross-model feature pairs validate (Pearson r >= 0.60, Procrustes cosines 0.895--0.956), while alignment degrades sharply below 0.8B. Cross-model steering vectors (B3-TI) achieve a 71.0% win rate across 15 supervised concepts versus 68.0% for same-model native vectors; a single universal vector achieves 67.3% in 4 of 5 models without any per-model supervision. Transfer degrades for models below 1.7B and for one model with generation instability, confirming that functional exploitability requires sufficient representational capacity. Our findings underscore the importance of scale thresholds in mechanistic interpretability: tools validated at 7B scale may not transfer to smaller models without revalidation. We provide the first functional complement to the Platonic Representation Hypothesis -- geometric convergence across independently trained LLMs supports cross-model behavioural control without fine-tuning, under the identified scale conditions.