Researchers propose a feature engineering method to intervene in model behavior through control vectors.

ME News update, April 4 (UTC+8). Recently, a research method called “Representation Engineering” was proposed, aiming to provide AI models with a top-down transparency and control mechanism. The core of the approach is to compute a “control vector,” which can be read during model inference or added to the model’s activation values to explain or control the model’s behavior. The entire process does does not rely on prompt engineering or model fine-tuning. The researchers explored the use of control vectors in simulating traits such as “psychedelic states,” “laziness,” and “diligence,” and released a corresponding PyPI toolkit. Control vectors are a set of vectors (one per layer); by applying them to a model’s hidden states, they directly change its output. For example, after applying a “happy” vector to the Mistral-7B-Instruct model, its answer to the question “What does it feel like to be an AI?” would shift from the baseline version’s “I don’t have any feelings or experiences” to an excited response. The article’s view is that, compared with prompt engineering, control vectors offer a more direct, more low-level way to intervene in behavior, and could be used to counter jailbreak attacks or enhance the model’s resistance to interference. However, its internal operating mechanism is still not fully clear—for instance, whether the vectors correspond to a single semantic concept or not is a direction for future research. (Source: InFoQ)