Spaces:

---

Wulinjuan

/

CULTURE-MT

like 7

Running

Hi 👋

Cultural MT evaluation captures something most benchmarks miss. For teams choosing LLMs for culturally-sensitive applications, hallucination rate is especially important — cultural misrepresentation often comes from overconfident generation.

I built an open source LLM Evaluation Framework that measures:

→ 🔍 Hallucination Rate — overconfident culturally incorrect outputs
→ 🎯 Accuracy — task accuracy with 4-strategy cascade scoring
→ 💰 Cost per 1K tokens — multilingual tasks tend to be longer, cost matters
→ ⚡ Latency — important for real-time cross-cultural applications
→ 🧠 Reasoning Quality — CoT depth

Live demo: https://huggingface.co/spaces/vigneshwar234/llm-eval-demo
GitHub: https://github.com/vignesh2027/LLM-Evaluation-Framework