Francisco Ferreira da Silva

Note: This is a research update sharing preliminary results as part of ongoing work. Figure 1: Contrastive (difference-of-means, English→Mandarin) feature directions elicit a downstream response at much smaller perturbation magnitudes than SAE directions, which behave similarly to random directions. This holds across multiple models and experimental setups. Summary & Main Results Understanding how concepts are represented in LLM internals would be extremely useful for AI safety (generally understanding AI's thinking, AI control and monitoring, retargeting the search). We assume that concepts are represented as directions in activation space (“features”). The most popular feature-finding approach, sparse dictionary learning via sparse autoencoders, suffers from strong dataset dependence (e.g. Heap et al. 2025, Kissane et al. 2024, Heimersheim 2024). It is thus unclear whether the directions they find reflect the model's computations or properties of the training distribution. Instead, we propose identifying feature directions by perturbing activations along a direction and using the model's response to judge whether the direction corresponds to a feature (following Mack & Turner 2024, Heimersheim & Mendel, 2024). We hypothesize that perturbations into random directions are suppressed, as they only produce interference noise along feature directions, while perturbations into feature directions create a focused perturbation and elicit a strong model response. If this is the case, we should see a longer suppression region for random directions than for feature directions. Our novel contribution is that we test this hypothesis using contrastive probing/steering vectors. We construct contrastive vectors (difference-of-means) for language (English→Mandarin), programming language (Python→Haskell), and gender (male→female) across different model families (Gemma 2-9B, Llama 3.1-8B, and Qwen 3-1.7B). We then perturb the models into contrastive-, SAE feature- and random