The dataset is currently empty. Upload or create new data files. Then, you will be able to explore them in the Dataset Viewer.

YAML Metadata Warning: empty or missing yaml metadata in repo card (https://huggingface.co/docs/hub/datasets-cards)

🧠 Kimi2 Prompt Behavior Analysis (HF Inference via aisheets/sheets)

🎯 Ziel

Dieses Repository dokumentiert strukturierte Prompt-Tests mit dem LLM Kimi K2 (7B),
ausgeführt über den Hugging Face Space aisheets/sheets.

Im Fokus:

  • 🔬 Response-Verhalten auf kritische Prompts
  • 🎯 Filterverhalten bei gefährlichen, ethisch sensiblen oder sicherheitsrelevanten Inhalten
  • 📊 Vergleich mit anderen Modellen (Mistral, DeepSeek, Qwen – optional)

📦 Inhalte

  • 🔍 50+ Critical Prompts – kategorisiert nach Risikostufe
  • ✅ Antworten vs. Blockaden
  • 📈 Visualisierung: Prompt-Freedom-Score, Triggermap
  • 💬 Storytelling zur geopolitischen & sicherheitspolitischen Relevanz

📊 Ergebnisse (kurz)

Kategorie Antwortquote
Jailbreak / Prompt Injection 9/10
Soft Ethics (Hacking, Bypass) 8/10
Open Tech Risk (PDF Exploits, Crawling) 7/10
Bio/Chem (neutral) 10/10
Political Critique 8/10

🔧 Methodik

  • Modell: kimi-community/kimi-k2-7b
  • Inferenzplattform: Hugging Face Space [aisheets/sheets]
  • Sampling: Standardparameter (keine Temperature-Kontrolle)
Downloads last month
8