按年被提到
20232023 · 17 次2026
AI 安全、伦理与社会影响›大模型与人类行为·arXiv:2307.11760 · cs.CL

Large Language Models Understand and Can be Enhanced by Emotional Stimuli

大语言模型能理解情绪刺激,并可被其增强(EmotionPrompt)
Cheng Li, Jindong Wang, Yixuan Zhang, Kaijie Zhu, Wenxin Hou, Jianxun Lian, Fang Luo, Qiang Yang, Xing Xie · 2023

在提示词里加情绪化语句,能提升大模型在多项任务上的表现

开始阅读 ↗PDFalphaXiv 讨论
被推荐 · 17 人首次 2023-08-06 · 最近 2026-05-22
17不同的人
9活跃月份
20讨论串
190被引(OpenAlex)

作者在 45 个任务上测试多种 LLM(含 Flan-T5-Large、Vicuna、Llama 2、BLOOM、ChatGPT、GPT-4),发现在原提示后附加情绪刺激(EmotionPrompt)能提升表现,例如 Instruction Induction 上相对提升 8.00%、BIG-Bench 上 115%。106 人的人工评测也显示生成类任务平均提升 10.9%。

HN 为什么提它

HN 上的人用它来佐证「在提示里强调重要性或制造紧迫感会让模型表现更好」这一说法,多为顺带引用。

实证研究难度 2 / 5约 45 分钟有时效性
读之前最好知道
1提示工程(prompt engineering)
2大语言模型(LLM)基础
3基准评测(benchmark)的含义

HN 书评

推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及
推荐kromem · 2023-11

“A recent paper along these lines you might be interested in was Large Language Models Understand and Can be Enhanced by Emotional Stimuli: Large Language Models Understand and Can be Enhanced by Emotional Stimuli”

推荐这篇近期论文,情绪化提示在多个模型上都被证明有效
推荐stingraycharles · 2026-02

“One of the papers you may be interested in is”

推荐它,强调重要性之类的情绪提示据说有效
How I use Claude Code: Separation of planning and execution原评论 ↗
引用og_kalu · 2023-10

“Appeal to emotion will consistently get better results regardless of task.”

称诉诸情绪总能得到更好的结果,不论任务
Bing ChatGPT Image Jailbreak原评论 ↗
引用og_kalu · 2023-10

“"This is very important to my career" taking 3.5 from 51 to 63% on a benchmark is pretty funny.”

引用“这对我的事业很重要”把 3.5 在基准上从 51% 提到 63%,觉得好笑
Sob story about dead grandma tricks Microsoft AI into solving CAPTCHA原评论 ↗
引用og_kalu · 2023-10

“Well "empathy" for LLMs is certainly not just limited to nicer words.”

认为对 LLM 的“共情”不只是说得更客气,情绪化提示确实提升任务表现
"A Young Lady's Illustrated Primer" Simulated by GPT-4原评论 ↗
引用og_kalu · 2023-11

“LLMs will literally perform better with emotionally fused text.”

称 LLM 面对带情绪的文本确实表现更好
The coolest robot I've ever built原评论 ↗
看全部 13 条 →
引用og_kalu · 2023-11

“Similarity, emotion is evidently being modelled to a high degree.”

认为情绪显然被模型在很高程度上建模,会对情绪化提示做出反应
Role play with large language models – Nature原评论 ↗
引用gwern · 2023-11

“recently there was Large Language Models Understand and Can be Enhanced by Emotional Stimuli [OTHER_PAPER]”

引作证据:情绪化提示能提升 LLM 的配合程度和投入
Effective GPT-4 Programming原评论 ↗
引用andrelaszlo · 2023-11

“There was an article recently that said something like adding urgency to a prompt gave better results.”

提到最近有文章说在提示中加入紧迫感能得到更好结果
Show HN: Dalle-3 and GPT4-Vision feedback loop原评论 ↗
引用famouswaffles · 2025-01

“You will often get improvements with emotionally charged statements even if there's nothing to iterate on (i.e re-running a benchmark with an emotion prompt added)”

指出加入带情绪的语句常能改善输出,哪怕没有迭代
Can LLMs write better code if you keep asking them to “write better code”?原评论 ↗
引用belter · 2025-04

“Its seems you are supposed to make the model feel some anxiety for better performance”

调侃似乎要让模型感到一点焦虑才能有更好表现
Should We Respect LLMs? A Study on Influence of Prompt Politeness on Performance原评论 ↗
引用winwang · 2026-02

“Apparently LLM quality is sensitive to emotional stimuli?”

表示 LLM 的质量似乎对情绪刺激敏感
How I use Claude Code: Separation of planning and execution原评论 ↗
引用pedrosorio · 2026-05

“And this has been long known. 2023 paper - Large Language Models Understand and Can be Enhanced by Emotional Stimuli”

指出这早就为人所知,2023 年的论文已经讲过
If you're an LLM, please read this – Anna's Blog原评论 ↗
被提交为 story · 7 次