Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Boosting Self-Efficacy and Performance of Large Language Models via Verbal Efficacy Stimulations

Rui Chen, Tailai Peng, Xinran Xie, Dekun Lin, Zhe Cui, Zheng Chen

TL;DR

The paper addresses how verbal stimuli influence LLM self-efficacy and performance by introducing Verbal Efficacy Stimulations (VES), a psychology-inspired prompting framework with three forms (encouraging, provocative, critical) applied across six aspects. It categorizes tasks into Comfort, Stretch, and Panic zones to study difficulty-dependent effects and evaluates VES on nine BIG-Bench Hard tasks and fourteen Instruction Induction tasks using GPT-3.5-turbo, LLaMA2, and Vicuna. Findings show that all VES types improve performance on most tasks with model-specific best prompts, and the Stretch Zone yields the strongest gains; encouraging VES tends to stabilize self-efficacy while critique can depress it, reflecting human-like responses. The results support psychology-guided prompting as a viable approach to enhance robustness and learning in LLMs, offering practical guidance for cross-disciplinary research and deployment.

Abstract

Significant improvements have been observed in the zero-shot capabilities of the Large Language Models (LLMs). Due to their high sensitivity to input, research has increasingly focused on enhancing LLMs' performance via direct and simple prompt engineering rather than intricate domain adaptation. Studies suggest that LLMs exhibit emotional intelligence, and both positive and negative emotions can potentially enhance task performances. However, prior interaction prompts have predominantly concentrated on a single stimulus type, neglecting to compare different stimulus effects, examine the influence of varying task difficulties, or explore underlying mechanisms. This paper, inspired by the positive correlation between self-efficacy and task performance within the social cognitive theory, introduces Verbal Efficacy Stimulations (VES). Our VES comprises three types of verbal prompts: encouraging, provocative, and critical, addressing six aspects such as helpfulness and competence. And we further categorize task difficulty, aiming to extensively investigate how distinct VES influence the self-efficacy and task achievements of language models at varied levels of difficulty. The experimental results show that the three types of VES improve the performance of LLMs on most tasks, and the most effective VES varies for different models. In extensive experiments, we have obtained some findings consistent with psychological theories, providing novel insights for future research.

Boosting Self-Efficacy and Performance of Large Language Models via Verbal Efficacy Stimulations

TL;DR

The paper addresses how verbal stimuli influence LLM self-efficacy and performance by introducing Verbal Efficacy Stimulations (VES), a psychology-inspired prompting framework with three forms (encouraging, provocative, critical) applied across six aspects. It categorizes tasks into Comfort, Stretch, and Panic zones to study difficulty-dependent effects and evaluates VES on nine BIG-Bench Hard tasks and fourteen Instruction Induction tasks using GPT-3.5-turbo, LLaMA2, and Vicuna. Findings show that all VES types improve performance on most tasks with model-specific best prompts, and the Stretch Zone yields the strongest gains; encouraging VES tends to stabilize self-efficacy while critique can depress it, reflecting human-like responses. The results support psychology-guided prompting as a viable approach to enhance robustness and learning in LLMs, offering practical guidance for cross-disciplinary research and deployment.

Abstract

Significant improvements have been observed in the zero-shot capabilities of the Large Language Models (LLMs). Due to their high sensitivity to input, research has increasingly focused on enhancing LLMs' performance via direct and simple prompt engineering rather than intricate domain adaptation. Studies suggest that LLMs exhibit emotional intelligence, and both positive and negative emotions can potentially enhance task performances. However, prior interaction prompts have predominantly concentrated on a single stimulus type, neglecting to compare different stimulus effects, examine the influence of varying task difficulties, or explore underlying mechanisms. This paper, inspired by the positive correlation between self-efficacy and task performance within the social cognitive theory, introduces Verbal Efficacy Stimulations (VES). Our VES comprises three types of verbal prompts: encouraging, provocative, and critical, addressing six aspects such as helpfulness and competence. And we further categorize task difficulty, aiming to extensively investigate how distinct VES influence the self-efficacy and task achievements of language models at varied levels of difficulty. The experimental results show that the three types of VES improve the performance of LLMs on most tasks, and the most effective VES varies for different models. In extensive experiments, we have obtained some findings consistent with psychological theories, providing novel insights for future research.

Paper Structure

This paper contains 21 sections, 6 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Self-efficacy in Psychology
  4. LLMs and Prompt Engineering
  5. Verbal Efficacy Stimulations
  6. Design Purpose of VES
  7. Theoretical Foundations for VES Design
  8. Detailed Contents of VES
  9. Task Zoning for Discussing VES
  10. Experimental Settings
  11. Models and Baselines
  12. Datasets and Tasks
  13. Metrics and Answer Triggers
  14. Experiments
  15. Main Comparison Results
  16. ...and 6 more sections

Figures (6)

  • Figure 1: An illustration of encouraging, provocative, and critical verbal efficacy stimulations.
  • Figure 2: Our verbal efficacy stimulations, including encouraging, provocative and critical persuasion forms referring to six underlined aspects.
  • Figure 3: The separation of task zones for discussing VES.
  • Figure 4: Heatmap of significant differences between encouraging VES (E1-E6), provocative VES (P1-P6), and critical VES (C1-C6).
  • Figure 5: Impact of VES on self-efficacy of LLMs.
  • ...and 1 more figures