Controllable Affective Generation via Latent Vector Steering

2026-08-26

AI Summary

为解决大型语言模型产生情感平淡响应的问题，提出EmoVec框架，通过潜在向量导向控制情感生成，增强情感显著性同时保持语义内容、流畅性和连贯性。

Abstract

Large Language Models (LLMs) often produce emotionally flattened responses after alignment, limiting their effectiveness in affect-sensitive applications. In this paper, we propose EmoVec, a lightweight framework for controllable affective generation via latent vector steering. EmoVec extracts emotion-specific directions from paired neutral and emotion-conditioned responses using contrastive activation addition, and further refines them through task-specific debiasing and principal subspace removal. During inference, these vectors are injected into the final residual stream with static or scenario-adaptive scaling, enabling continuous control over emotional intensity without updating model weights. Experiments across three LLMs and eight emotions show that EmoVec consistently improves emotional salience while largely preserving semantic content, fluency, and coherence. Ablation studies and human evaluation further confirm the effectiveness of vector purification and adaptive scaling, establishing EmoVec as a practical inference-time method for affective control in deployed LLMs.

Problem

• Large Language Models
• emotionally flattened responses
• affect-sensitive applications

Innovation

• Latent Vector Steering
• Contrastive Activation Addition
• Scenario-Adaptive Scaling