Model editing methods modify specific behaviors of Large Language Models by altering a small, targeted set of network weights and require very little data and compute. These methods can be used for malicious applications such as inserting misinformation or simple trojans that result in adversary-specified behaviors when a trigger word is present. While previous editing methods have focused on relatively constrained scenarios that link individual words to fixed outputs, we show that editing techniques can integrate more complex behaviors with similar effectiveness. We develop Concept-ROT, a model editing-based method that efficiently inserts trojans which not only exhibit complex output behaviors, but also trigger on high-level concepts -- presenting an entirely new class of trojan attacks. Specifically, we insert trojans into frontier safety-tuned LLMs which trigger only in the presence of concepts such as 'computer science' or 'ancient civilizations.' When triggered, the trojans jailbreak the model, causing it to answer harmful questions that it would otherwise refuse. Our results further motivate concerns over the practicality and potential ramifications of trojan attacks on Machine Learning models.

We describe a computational model for inferring 3D structure from the motion of projected 2D points in an image, with the aim of understanding how biological vision systems learn and internally represent 3D transformations from the statistics of their input. The model uses manifold transport operators to describe the action of 3D points in a scene as they undergo transformation. We show that the model can learn the generator of the Lie group for these transformations from purely 2D input, providing a proof-of-concept demonstration for how biological systems could adapt their internal representations based on sensory input. Focusing on a rotational model, we evaluate the ability of the model to infer depth from moving 2D projected points and to learn rotational transformations from 2D training stimuli. Finally, we compare the model performance to psychophysical performance on structure-from-motion tasks.

The manifold hypothesis states that high-dimensional data can be modeled as lying on or near a low-dimensional, nonlinear manifold. Variational Autoencoders (VAEs) approximate this manifold by learning mappings from low-dimensional latent vectors to high-dimensional data while encouraging a global structure in the latent space through the use of a specified prior distribution. When this prior does not match the structure of the true data manifold, it can lead to a less accurate model of the data. To resolve this mismatch, we introduce the Variational Autoencoder with Learned Latent Structure (VAELLS) which incorporates a learnable manifold model into the latent space of a VAE. This enables us to learn the nonlinear manifold structure from the data and use that structure to define a prior in the latent space. The integration of a latent manifold model not only ensures that our prior is well-matched to the data, but also allows us to define generative transformation paths in the latent space and describe class manifolds with transformations stemming from examples of each class. We validate our model on examples with known latent structure and also demonstrate its capabilities on a real-world dataset.

Deep generative networks have been widely used for learning mappings from a low-dimensional latent space to a high-dimensional data space. In many cases, data transformations are defined by linear paths in this latent space. However, the Euclidean structure of the latent space may be a poor match for the underlying latent structure in the data. In this work, we incorporate a generative manifold model into the latent space of an autoencoder in order to learn the low-dimensional manifold structure from the data and adapt the latent space to accommodate this structure. In particular, we focus on applications in which the data has closed transformation paths which extend from a starting point and return to nearly the same point. Through experiments on data with natural closed transformation paths, we show that this model introduces the ability to learn the latent dynamics of complex systems, generate transformation paths, and classify samples that belong on the same transformation path.

We develop a framework for localizing an unknown point w using paired comparisons of the form "w is closer to point xithan to xj" when the points lie in a union of known subspaces. This model, which extends a broad class of existing methods to exploit union of subspaces structure, provides a powerful framework for using the types of structure found in many practical applications. We divide the problem into two phases: (1) determining which subspace w lies in, and (2) localizing w within the identified subspace using existing techniques. We introduce two algorithms for determining the subspace in which an unknown point lies: the first admits a sample complexity guarantee demonstrating the advantage of the union of subspaces model, and the second improves performance in practice using an adaptive Bayesian strategy. We demonstrate the efficacy of our method with experiments on synthetic data and in an image search application.

We develop a method for generating causal post-hoc explanations of black-box classifiers based on a learned low-dimensional representation of the data. The explanation is causal in the sense that changing learned latent factors produces a change in the classifier output statistics. To construct these explanations, we design a learning framework that leverages a generative model and information-theoretic measures of causal influence. Our objective function encourages both the generative model to faithfully represent the data distribution and the latent factors to have a large causal influence on the classifier output. Our method learns both global and local explanations, is compatible with any classifier that admits class probabilities and a gradient, and does not require labeled attributes or knowledge of causal structure. Using carefully controlled test cases, we provide intuition that illuminates the function of our causal objective. We then demonstrate the practical utility of our method on image recognition tasks.