Self-Evaluation Is Already There: Eliciting Latent Judge Calibration in Base LLMs with Minimal Data

Part of

LLM-Enhanced Document Retrieval And Embeddings