Search papers, labs, and topics across Lattice.
This paper introduces PersonaDose, a method for precisely controlling the intensity of persona traits in LLMs by specializing a FLAS controller and calibrating its activation flow time against measured trait expression. PersonaDose achieves significant increases in core-trait expression across multiple LLM architectures, outperforming contrastive activation addition. The work demonstrates the ability to achieve fine-grained control over persona expression, separating the learned behavioral range from the accuracy of specific intensity requests.
LLMs can now be precisely "dosed" for specific persona traits, achieving graded control over expression intensity beyond simple on/off activation.
An activation-steering coefficient sets intervention strength, but requesting a particular degree of persona expression requires a behavioral scale. We study persona dosing: controlling a language model through a trait description and a requested mean intensity. PersonaDose specializes a shared, description-conditioned FLAS controller on persona responses, then calibrates its flow time against measured trait expression. Training responses are not paired with requested target intensities. Across Llama-3.1-8B, Qwen3-8B, and Gemma-3-4B, PersonaDose raises core-trait expression at the Persona Vectors coherence floor of 75 by 33.2, 18.3, and 17.8 points over contrastive activation addition. Calibration-selected settings retain an expression advantage on held-out questions, although the coherence floor does not hold for every trait there. Across seven trained traits, calibrated requests yield mean targeting errors of 4.7-6.2 points over 14-22 calibration-reachable targets out of 28 per model. These results separate the behavioral range learned by a controller from the accuracy of requests within that range.
