five

SperanzaMax/Cortex-Nexus-Emotions-3k

收藏
Hugging Face2026-04-20 更新2026-04-26 收录
下载链接:
https://hf-mirror.com/datasets/SperanzaMax/Cortex-Nexus-Emotions-3k
下载链接
链接失效反馈
官方服务:
资源简介:
--- license: cc-by-4.0 task_categories: - text-generation language: - en tags: - llm - prompt-engineering - emotional-prompting - wilcoxon - nlp size_categories: - 1K<n<10K --- # Cortex-Nexus: Emotional Prompting Experiment Dataset **3,600+ cycles** of double-blind emotional prompting experiments on large language models (Llama-3-70B). ## Key Findings - Injecting **Curiosity (0.95) + Frustration (0.20)** significantly improves output quality on **philosophical/exploratory tasks** (p=0.022, d=0.263) - The **exact same configuration** significantly **degrades** performance on **technical/deterministic tasks** - a perfect negative control - **"Confidence" injection** does not affect semantic quality, but produces 4x more linguistic hedging (Two-Layer Model) ## Two-Layer Model of LLM Emotional Simulation | Layer | Mechanism | Affected by | |-------|-----------|-------------| | Semantic Quality | Content accuracy/depth | Curiosity, Frustration | | Epistemic Style | Hedging, certainty markers | Confidence | ## Dataset Files - `cortex_nexus_APA7.pdf` - Full academic paper (APA7 format) - `cycles_optimo_combinado.jsonl` - Optimal emotion config results (Curiosity + Frustration) - `cycles_tecnico_combinado.jsonl` - Technical domain results (negative control) - `cycles_asombro_combinado.jsonl` - Awe/Wonder emotion results ## Experimental Setup - **Models**: Llama-3.3-70B (experimental agent), Mistral-Nemo (judge), Meta-Llama-3-8B (baseline) - **API**: NVIDIA NIM - **Architecture**: Double-blind, 3-agent system with LLM-as-a-Judge scoring - **Validation**: Wilcoxon signed-rank tests (alpha=0.05) - **Cycles**: 3,600+ total experimental cycles ## Citation ```bibtex @misc{speranza2025cortexnexus, title={Cortex-Nexus: Injection of Simulated Emotional States in Large Language Models}, author={Speranza, Maximiliano Rodrigo}, year={2025}, note={Dataset and code: https://github.com/SperanzaMax/Cortex-Nexus} } ``` ## Related Links - GitHub Repository: https://github.com/SperanzaMax/Cortex-Nexus - License: CC-BY 4.0
提供机构:
SperanzaMax
5,000+
优质数据集
54 个
任务类型
进入经典数据集
二维码
社区交流群

面向社区/商业的数据集话题

二维码
科研交流群

面向高校/科研机构的开源数据集话题

数据驱动未来

携手共赢发展

商业合作