Vision-language-action (VLA) models are a promising approach to general purpose robot manipulation. While the internal computation of vision-language models, which compose the backbone of VLAs, has been studied extensively, less is known about how their perceptual and semantic representations are transformed into actio...
Aiden Swann, Mathilde Papillon, Mozes Jacobs et al.· 0 citations
Overall, this work lays out an initial phenomenology of realized geometry in VLA policies and demonstrates that a concept's shape carries as much importance as its readability.
Mozes Jacobs, Aiden Swann, Mathilde Papillon et al.· 0 citations
Sparse autoencoders are widely used to uncover interpretable features in neural networks, yet reliable recovery remains difficult when features overlap or activate infrequently. These challenges involve both inferring which features explain an input and learning the dictionary that represents them. Here, we unify infer...
Hadi Vafaii, Tejas Rao, David Chanin et al.· 0 citations
This work characterize precisely the geometry of computation in a number-comparison task, as an abstraction of comparison for decision making, and how models utilize geometry in an elegant fashion to implement it, demonstrating that the manifold hypothesis can co-exist with linear representations.
Sai Sumedh R. Hindupur, Hadas Orgad, Thomas Fel et al.· 0 citations
This work introduces SparKer, a sparse ensemble of Gaussian kernels trained in a semi-supervised Neyman-Pearson framework to locally model the likelihood ratio between a sample that may contain anomalies and an anomaly-free reference, and provides theoretical insights into the mechanisms driving detection and self-orga...
Gaia Grosso, Sai Sumedh R. Hindupur, Thomas Fel et al.· arXiv.org· 2 citations
This work finds that simple difference of means vectors coherently represent reward hacking in Kimi K3, GLM 5.2, and Qwen 3.8 Max, and provides evidence that simple, white-box methods can be used to scalably study and monitor reward hacking behaviors in frontier open source models.
Leon Bergen, Usha Bhalla, Andrew Lee et al.· 3 citations
This work introduces axis-aligned feature accentuation, which converts each model’s fitted encoding axis into graded stimulus perturbations that are predicted to parametrically control neural firing within and beyond the natural-image range.
Jacob S. Prince, Binxu Wang, Thomas Fel et al.· bioRxiv· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
