AI models have their own internal representations of knowledge or concepts that are often difficult to discern, even as they are critical to the models’ output. For instance, knowing more about a model’s representation of a concept would help explain why an AI model might “hallucinate” information, or why certain prompts can trick it into responses that dodge its built-in safeguards. Daniel Beaglehole and colleagues now introduce a robust method to extract these representations of concepts, which works across several large-scale language, reasoning, and vision AI models. Their technique uses a feature extraction algorithm called the Recursive Feature Machine. By extracting concept representations with this technique, Beaglehole et al. were able to monitor these models in ways that exposed some of their vulnerabilities to behaviors like hallucinations and to steer them toward improved output responses. Surprisingly, the concept representations were transferable between different languages and could be combined with other concept representations for multi-concept steering, the researchers noted. “Together, these results suggest that the models know more than they express in responses and that understanding internal representations could lead to fundamental performance and safety improvements,” the authors write.

A team of researchers has found a way to steer the output of large language models by manipulating specific concepts inside these models. The new method could lead to more reliable, more efficient, and less computationally expensive training of LLMs. But it also exposes potential vulnerabilities. The researchers present their findings in the Feb. 19, 2026, issue of the journal Science. 

Mass General Brigham researchers explore how innovative technologies can improve surgical experiences for Spanish-speaking patients.

Researchers from the Department of Computer Science at Bar-Ilan University and from NVIDIA’s AI research center in Israel have developed a new method that significantly improves how artificial intelligence models understand spatial instructions when generating images – without retraining or modifying the models themselves.