Sparse Activation Editing for Reliable Instruction Following in Narratives

Complex narrative contexts often challenge language models' ability to follow instructions, and existing benchmarks fail to capture these difficulties. To address this, we propose Concise-SAE, a training-free framework that improves instruction following by identifying and editing instruction-relevant neurons using only natural language instructions, without requiring labelled data. To thoroughly evaluate our method, we introduce FreeInstruct, a diverse and realistic benchmark of 1,212 examples that highlights the challenges of instruction following in narrative-rich settings. While initially motivated by complex narratives, Concise-SAE demonstrates state-of-the-art instruction adherence across varied tasks without compromising generation quality.

MIB: A Mechanistic Interpretability Benchmark

How can we know whether new mechanistic interpretability methods achieve real improvements? In pursuit of meaningful and lasting evaluation standards, we propose MIB, a benchmark with two tracks spanning four tasks and five models. MIB favors methods that precisely and concisely recover relevant causal pathways or specific causal variables in neural language models. The circuit localization track compares methods that locate the model components - and connections between them - most important for performing a task (e.g., attribution patching or information flow routes). The causal variable localization track compares methods that featurize a hidden vector, e.g., sparse autoencoders (SAEs) or distributed alignment search (DAS), and locate model features for a causal variable relevant to the task. Using MIB, we find that attribution and mask optimization methods perform best on circuit localization. For causal variable localization, we find that the supervised DAS method performs best, while SAE features are not better than neurons, i.e., standard dimensions of hidden vectors. These findings illustrate that MIB enables meaningful comparisons of methods, and increases our confidence that there has been real progress in the field.

Beyond Voice Assistants: Exploring Advantages and Risks of an In-Car Social Robot in Real Driving Scenarios

In-car Voice Assistants (VAs) play an increasingly critical role in automotive user interface design. However, existing VAs primarily perform simple 'query-answer' tasks, limiting their ability to sustain drivers' long-term attention. In this study, we investigate the effectiveness of an in-car Robot Assistant (RA) that offers functionalities beyond voice interaction. We aim to answer the question: How does the presence of a social robot impact user experience in real driving scenarios? Our study begins with a user survey to understand perspectives on in-car VAs and their influence on driving experiences. We then conduct non-driving and on-road experiments with selected participants to assess user experiences with an RA. Additionally, we conduct subjective ratings to evaluate user perceptions of the RA's personality, which is crucial for robot design. We also explore potential concerns regarding ethical risks. Finally, we provide a comprehensive discussion and recommendations for the future development of in-car RAs.

Erasure of Unaligned Attributes from Neural Representations

We present the Assignment-Maximization Spectral Attribute removaL (AMSAL) algorithm, which aims at removing information from neural representations when the information to be erased is implicit rather than directly being aligned to each input example. Our algorithm works by alternating between two steps. In one, it finds an assignment of the input representations to the information to be erased, and in the other, it creates projections of both the input representations and the information to be erased into a joint latent space. We test our algorithm on an extensive array of datasets, including a Twitter dataset with multiple guarded attributes, the BiasBios dataset and the BiasBench benchmark. The latter benchmark includes four datasets with various types of protected attributes. Our results demonstrate that bias can often be removed in our setup. We also discuss the limitations of our approach when there is a strong entanglement between the main task and the information to be erased.

Gold Doesn't Always Glitter: Spectral Removal of Linear and Nonlinear Guarded Attribute Information

We describe a simple and effective method (Spectral Attribute removaL; SAL) to remove guarded information from neural representations. Our method uses singular value decomposition and eigenvalue decomposition to project the input representations into directions with reduced covariance with the guarded information rather than maximal covariance as normally these factorization methods are used. We begin with linear information removal and proceed to generalize our algorithm to the case of nonlinear information removal through the use of kernels. Our experiments demonstrate that our algorithm retains better main task performance after removing the guarded information compared to previous methods. In addition, our experiments demonstrate that we need a relatively small amount of guarded attribute data to remove information about these attributes, which lowers the exposure to such possibly sensitive data and fits better low-resource scenarios.