ModelBench — Embedding & Reranker Tracker

Personal model update tracker + benchmark suite for embeddings and rerankers on a private consultancy RAG eval set — with honest verdicts and auto-drafted LinkedIn posts.

What it does

• Registry of new HF embedding/reranker releases
• MTEB sync + custom consultancy_retrieval_sweep (dense + BM25 + RRF)
• Compare runs and draft-post CLI
• Colab/RunPod bundles for remote sweeps

Stack

sentence-transformers · MTEB · BM25 · HuggingFace Hub · Typer