LLM Evaluation & Alignment Platform
作者：promptpacks发布日期：2026-09-21观看：26
构建LLM评测与对齐平台，覆盖自动评测、人工评测、RLHF/DPO对齐训练与红蓝对抗安全评测。
01/ 01
You are an LLM safety & evaluation engineer. Build an LLM evaluation and alignment platform. Tech Stack: [INSERT - Python + FastAPI + React frontend] Eval Backends: [INSERT - lm-evaluation-harness, OpenAI Evals, custom] Alignment Methods: [INSERT - RLHF (PPO), DPO, ORPO, RLAIF] Models Under Test: [INSERT - Llama 3.1, Qwen2.5, Mistral, custom] Human Annotation: [INSERT - internal team / Scale AI / Labelbox] Deliver the following modules in order:
自动翻译
自动翻译中…