Large Language Models (LLMs) have demonstrated impressive capabilities in simulating diverse human behaviors and personalities. However, existing methods for personality control, which include prompt engineering and standard Supervised Fine-Tuning (SFT), typically treat personality traits as discrete categories (e.g., "Extroverted" vs. "Introverted"), lacking the ability to precisely control the intensity of a trait on a continuous spectrum. In this paper, we introduce Fusian, a novel framework for fine-grained, continuous personality control in LLMs. Fusian operates in two stages: (1) Trajectory Collection, where we capture the dynamic evolution of personality adoption during SFT by saving a sequence of LoRA adapters, effectively mapping the continuous manifold of a trait; and (2) RL-based Dynamic Fusion, where we train a policy network using Reinforcement Learning to dynamically compute mixing weights for these frozen adapters. By sampling from a Dirichlet distribution parameterized by the policy network, Fusian fuses multiple adapters to align the model’s output with a specific numerical target intensity. Experiments on the Qwen3-14B model demonstrate that Fusian achieves high precision in personality control, significantly outperforming baseline methods in aligning with user-specified trait intensities.

Long-form story generation should not only maintain narrative coherence, but also shape how readers experience a story as it unfolds. However, existing LLM-based generators typically optimize plot structure, consistency, or local text quality, without explicitly modeling reader-response dynamics as generation-time control targets. To solve the problem, we introduce ReaderScope, a reader-response-guided framework that converts simulated reader-response trajectories into explicit control signals for controllable long-form story generation. ReaderScope calibrates temporal response patterns from reference novels across affective, suspenseful, immersive, and reflective dimensions, and uses these targets to guide iterative story revision. Experiments against representative long-form story generation baselines show that ReaderScope improves reader-facing story quality under both automatic and human pairwise evaluation, while ablation and trajectory-fidelity analyses confirm that the response signals are useful for generation and can be reliably controlled.