aiqbench analysis
Capsule Lens offers a rigorous framework for mechanistic interpretability by modeling concept geometry as trackable capsules within both static and dynamic model representations. This approach moves beyond simple mapping to directly characterize how concepts occupy vector space, revealing distinct geometric dynamics across different training regimes like CLIP pretraining and RL post-training. Such granular visibility into internal representations is a critical step toward understanding and ensuring the trustworthy deployment of advanced AI systems.
Capsule Lens: Locating and Tracking Concept Geometry in Model Representations
arXiv:2609.05575v1 Announce Type: new Abstract: Understanding how concepts are encoded in the internal representations of machine learning models is a central problem in mechanistic interpretability, essential both for the science of deep learning and for the trustworthy deployment of increasingly capable models. Existing approaches to interpret model representations mainly map representations onto more interpretable spaces and do not directly characterize how concepts occupy representation space; various hypotheses have been proposed, but often lack of rigorous validation and largely focus on static representations. In this work, we introduce Capsule Lens, a framework that matches the region a concept occupies with a simple, trackable geometric form, a capsule, defined by several interpretable parameters, fitted in closed form to each concept's geometry and validated on held-out samples. We apply Capsule Lens in two major settings: static and dynamic representations. On static representations, we demonstrate how to locate concept geometry across various models, and how the span and norm curves uncover important geometric characteristics. On dynamic representations, we present three case studies tracking representation drifts induced by distinct training settings, CLIP pretraining, RL post-training on visual question answering, and RL post-training on mathematical reasoning. These analyses reveal qualitatively different geometric dynamics, ranging from broad network-wide restructuring in CLIP pretraining to localized and concept-specific changes in RL post-training. Our results include findings aligned with existing literature as well as novel observations. We believe Capsule Lens stands as a promising tool for locating, analyzing, and tracking concept geometry in both static and dynamic representations.
Abstract:Understanding how concepts are encoded in the internal representations of machine learning models is a central problem in mechanistic interpretability, essential both for the science of deep learning and for the trustworthy deployment of increasingly capable models. Existing approaches to interpret model representations mainly map representations onto more interpretable spaces and do not directly characterize how concepts occupy representation space; various hypotheses have been proposed, but often lack of rigorous validation and largely focus on static representations. In this work, we introduce Capsule Lens, a framework that matches the region a concept occupies with a simple, trackable geometric form, a capsule, defined by several interpretable parameters, fitted in closed form to each concept's geometry and validated on held-out samples. We apply Capsule Lens in two major settings: static and dynamic representations. On static representations, we demonstrate how to locate concept geometry across various models, and how the span and norm curves uncover important geometric characteristics. On dynamic representations, we present three case studies tracking representation drifts induced by distinct training settings, CLIP pretraining, RL post-training on visual question answering, and RL post-training on mathematical reasoning. These analyses reveal qualitatively different geometric dynamics, ranging from broad network-wide restructuring in CLIP pretraining to localized and concept-specific changes in RL post-training. Our results include findings aligned with existing literature as well as novel observations. We believe Capsule Lens stands as a promising tool for locating, analyzing, and tracking concept geometry in both static and dynamic representations.
Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
arXiv:2609.05575 [cs.LG]
(or arXiv:2609.05575v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2609.05575 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 343 words
Related coverage
How do neural networks regress cubic polynomials? Apparently, they use a trick invented in Milan 500 years ago
LessWrong · Jul 22, 2026
J++ Lens: Jacobian Filtering Enables More Faithful Workspace Lenses
LessWrong · Oct 8, 2026
The user in the LLM might be a moral patient too (a strange implication of the PSM and simulators)
LessWrong · Oct 15, 2026
PIRAMID: Progress and Plans
LessWrong · Jul 27, 2026