Prompt and verifier for extracting verifiable citations from a RAG answer over Jira tickets, scored by AlpacaEval 2.0 length-controlled.
Prompt and verifier for extracting verifiable citations from a RAG answer over customer interview transcripts, scored by Arena-Hard-Auto.
Prompt and verifier for extracting verifiable citations from a RAG answer over medical records, scored by Claude Sonnet 4.5 rubric scorer.
Prompt and verifier for extracting verifiable citations from a RAG answer over financial filings (10-K/10-Q), scored by Claude Sonnet 4.5 rubric scorer.
Prompt and verifier for extracting verifiable citations from a RAG answer over product manuals, scored by GPT-4o pairwise.
Prompt and verifier for extracting verifiable citations from a RAG answer over regulatory filings, scored by G-Eval with Gemini 2.5 Pro.
Prompt and verifier for extracting verifiable citations from a RAG answer over multilingual help center articles, scored by Ragas faithfulness judge.
Prompt and verifier for extracting verifiable citations from a RAG answer over PDFs with tables, scored by AlpacaEval 2.0 length-controlled.
Prompt and verifier for extracting verifiable citations from a RAG answer over scanned PDFs with OCR artifacts, scored by Arena-Hard-Auto.
Full fine-tuning recipe: SFT (supervised fine-tuning) on Mixtral 8x7B via LitGPT, targeting 2x A100 80GB, with data mix and eval plan.
Full fine-tuning recipe: SFT (supervised fine-tuning) on Yi 1.5 34B via torchtune, targeting 2x A100 80GB, with data mix and eval plan.
Full fine-tuning recipe: SFT (supervised fine-tuning) on Llama 3.3 70B via Unsloth, targeting 4x A100 40GB, with data mix and eval plan.