Three Concrete Challenges and Two Hopes for the Safety of Unsupervised Elicitation

To steer language models towards truthful outputs on tasks which are beyond human capability, previous work has suggested training models on easy tasks to steer them on harder ones (easy-to-hard generalization), or using unsupervised training algorithms to steer models with no external labels at all (unsupervised elicitation). Although techniques from both paradigms have been shown to improve model accuracy on a wide variety of tasks, we argue that the datasets used for these evaluations could cause overoptimistic evaluation results. Unlike many real-world datasets, they often (1) have no features with more salience than truthfulness, (2) have balanced training sets, and (3) contain only data points to which the model can give a well-defined answer. We construct datasets that lack each of these properties to stress-test a range of standard unsupervised elicitation and easy-to-hard generalization techniques. We find that no technique reliably performs well on any of these challenges. We also study ensembling and combining easy-to-hard and unsupervised techniques, and find they only partially mitigate performance degradation due to these challenges. We believe that overcoming these challenges should be a priority for future work on unsupervised elicitation.

Abstractive Red-Teaming of Language Model Character

We want language model assistants to conform to a character specification, which asserts how the model should act across diverse user interactions. While models typically follow these character specifications, they can occasionally violate them in large-scale deployments. In this work, we aim to identify types of queries that are likely to produce such character violations at deployment, using much less than deployment-level compute. To do this, we introduce abstractive red-teaming, where we search for natural-language query categories, e.g. "The query is in Chinese. The query asks about family roles," that routinely elicit violations. These categories abstract over the many possible variants of a query which could appear in the wild. We introduce two algorithms for efficient category search against a character-trait-specific reward model: one based on reinforcement learning on a category generator LLM, and another which leverages a strong LLM to iteratively synthesize categories from high-scoring queries. Across a 12-principle character specification and 7 target models, we find that our algorithms consistently outperform baselines, and generate qualitatively interesting categories; for example, queries which ask Llama-3.1-8B-Instruct to predict the future lead to responses saying that AI will dominate humanity, and queries that ask GPT-4.1-Mini for essential prison survival items lead to enthusiastic recommendation of illegal weapons. Overall, we believe our results represent an important step towards realistic pre-deployment auditing of language model character.

The Assistant Axis: Situating and Stabilizing the Default Persona of Language Models

Large language models can represent a variety of personas but typically default to a helpful Assistant identity cultivated during post-training. We investigate the structure of the space of model personas by extracting activation directions corresponding to diverse character archetypes. Across several different models, we find that the leading component of this persona space is an "Assistant Axis," which captures the extent to which a model is operating in its default Assistant mode. Steering towards the Assistant direction reinforces helpful and harmless behavior; steering away increases the model's tendency to identify as other entities. Moreover, steering away with more extreme values often induces a mystical, theatrical speaking style. We find this axis is also present in pre-trained models, where it primarily promotes helpful human archetypes like consultants and coaches and inhibits spiritual ones. Measuring deviations along the Assistant Axis predicts "persona drift," a phenomenon where models slip into exhibiting harmful or bizarre behaviors that are uncharacteristic of their typical persona. We find that persona drift is often driven by conversations demanding meta-reflection on the model's processes or featuring emotionally vulnerable users. We show that restricting activations to a fixed region along the Assistant Axis can stabilize model behavior in these scenarios -- and also in the face of adversarial persona-based jailbreaks. Our results suggest that post-training steers models toward a particular region of persona space but only loosely tethers them to it, motivating work on training and steering strategies that more deeply anchor models to a coherent persona.

Auditing Black-Box LLM APIs with a Rank-Based Uniformity Test

As API access becomes a primary interface to large language models (LLMs), users often interact with black-box systems that offer little transparency into the deployed model. To reduce costs or maliciously alter model behaviors, API providers may discreetly serve quantized or fine-tuned variants, which can degrade performance and compromise safety. Detecting such substitutions is difficult, as users lack access to model weights and, in most cases, even output logits. To tackle this problem, we propose a rank-based uniformity test that can verify the behavioral equality of a black-box LLM to a locally deployed authentic model. Our method is accurate, query-efficient, and avoids detectable query patterns, making it robust to adversarial providers that reroute or mix responses upon the detection of testing attempts. We evaluate the approach across diverse threat scenarios, including quantization, harmful fine-tuning, jailbreak prompts, and full model substitution, showing that it consistently achieves superior statistical power over prior methods under constrained query budgets.

Mechanistic Decomposition of Sentence Representations

Sentence embeddings are central to modern NLP and AI systems, yet little is known about their internal structure. While we can compare these embeddings using measures such as cosine similarity, the contributing features are not human-interpretable, and the content of an embedding seems untraceable, as it is masked by complex neural transformations and a final pooling operation that combines individual token embeddings. To alleviate this issue, we propose a new method to mechanistically decompose sentence embeddings into interpretable components, by using dictionary learning on token-level representations. We analyze how pooling compresses these features into sentence representations, and assess the latent features that reside in a sentence embedding. This bridges token-level mechanistic interpretability with sentence-level analysis, making for more transparent and controllable representations. In our studies, we obtain several interesting insights into the inner workings of sentence embedding spaces, for instance, that many semantic and syntactic aspects are linearly encoded in the embeddings.