Emotion2Skill: Model-Internal Emotion Signals for Adaptive Skill Selection and Evolution

2026-08-10

AI Summary

This work proposes Emotion2Skill, a novel framework that leverages interpretable internal emotional states of large language models (LLMs) to enhance skill selection and evolution. Unlike existing skill-based agents that rely solely on external textual cues, Emotion2Skill extracts a 27-dimensional linear emotion vector from the LLM’s residual stream to guide decision-making. It injects emotion summaries into routing prompts for adaptive skill selection and detects abrupt shifts in emotion trajectories to identify problematic skill invocations, thereby driving fine-grained refinement of standard operating procedures (SOPs). This approach transcends conventional methods limited to task descriptions or binary feedback by enabling optimization grounded in internal state dynamics. Evaluated on WebShop and ALFWorld, Emotion2Skill with Qwen3-8B achieves success rate improvements of 26.9% and 25.5% over zero-shot baselines, outperforming prior methods, with consistent gains observed on Qwen3-14B. Co-activation analysis further confirms semantic alignment between extracted emotions and selected skills.

Abstract

Skill-based LLM agents select reusable procedures from an external library to solve complex tasks, yet their routing decisions rely entirely on text-level signals such as task descriptions, verbal reflections, and experience-derived rules, while the model's own internal representational state remains unobserved. Recent interpretability work has shown that LLMs maintain linear emotion representations that causally influence behavior; however, these representations have been exploited only for post-hoc analysis or direct output steering, and have not been used to inform agent-level decision-making. We propose Emotion2Skill, a framework that extracts LLM-internal emotion vectors and incorporates them into both skill selection and skill evolution. At each decision step, a 27-dimensional emotion state is extracted from the residual stream and mapped to a confidence-gated summary injected into the routing prompt. Beyond online selection, emotion trajectories are analyzed for abrupt internal-state shifts to pinpoint problematic skill invocations, guiding targeted SOP rewriting that replaces the coarse binary outcome signal of prior methods. On WebShop and ALFWorld, Emotion2Skill with Qwen3-8B improves over the Zero-Shot baseline by +26.9% success rate and +25.5% average success respectively, outperforming all baselines on both benchmarks with consistent gains on Qwen3-14B. Co-activation analysis further reveals semantically coherent emotion--skill pairings, confirming that the routing improvements reflect meaningful internal-state signals rather than opaque statistical correlations. These results establish LLM-internal emotion representations as an effective decision-level signal for orchestrating agent skill systems, extending their utility beyond interpretability and output steering. The code is available at https://github.com/BoHan-LIN04/Emotion2Skill.

Problem

• skill-based agents
• internal emotion representations
• LLM decision-making
• adaptive skill selection
• emotion signals

Innovation

• emotion representations
• skill selection
• LLM interpretability
• adaptive skill evolution
• internal-state signals