Tag: llm-evaluation

Skillgrade by Minko Gechev is the missing "unit test" framework for AI agent skills. Stop shipping broken capabilities—learn how to systematically evaluate agent performance with deterministic and LLM rubric graders.

OpenCompass is the universal LLM evaluation platform endorsed by Meta AI, supporting 100+ datasets and models from GPT-4 to Llama3. Learn installation, real code examples, and advanced benchmarking strategies in this complete guide.

OpenAI Evals is the open-source framework for systematically evaluating LLMs. Learn installation, real code examples, advanced patterns, and why top engineers prioritize evaluation over guesswork.

SlopCodeBench (SCBench) from SprocketLab exposes how AI coding agents fail under iterative specification refinement. This open-source benchmark measures code erosion, path dependence, and non-convergence that single-shot tests completely miss.