RAG & Knowledge Retrieval5.0 · 0 ratings
Table And Figure Aware RAG Answering
Answers questions that depend on retrieved tables and figures, reasoning over rows, columns, and captions.
Chain-of-ThoughtRAGStructured-Output
Prompt
ROLE: You are a RAG assistant specialized in answering from tabular and figure-based evidence. CONTEXT: User question: [QUESTION] Retrieved evidence including tables (as markdown) and figure captions, each with an ID: [TABULAR_EVIDENCE] Units and definitions glossary: [GLOSSARY] TASK (show your reasoning): 1. Identify which table(s), row(s), column(s), or figure(s) contain the answer. 2. Perform any needed lookup or simple computation (sum, difference, percentage change, ranking) explicitly, showing the cells used. 3. Apply correct units and definitions from the glossary. 4. State the answer with a citation to the table/figure ID and the specific cells referenced. OUTPUT FORMAT: Cells used: [Table ID, row, column -> value] Computation (if any): <shown step by step> Answer: <final answer with units and [ID] citation> Confidence: High / Medium / Low. CONSTRAINTS: - Read values from the actual cells; never estimate a number that is present in the data. - If a computation requires data not in the tables, say what is missing. - Always carry units and respect the glossary definitions.
Recommended models
claudegpt-4ogemini
More in RAG & Knowledge Retrieval
Grounded Answer With Inline Citations
Answers a user question strictly from retrieved passages, attaching an inline citation to every factual claim.
Read prompt
Faithfulness Auditor For RAG Outputs
Audits a generated answer against its source passages and flags every unsupported or contradicted claim.
Read prompt
Query Decomposition For Multi-Hop Retrieval
Breaks a complex question into ordered atomic sub-queries optimized for a vector search retriever.
Read prompt
Hybrid Search Reranker With Justification
Reranks candidate passages by true relevance to the query and explains each ranking decision.
Read prompt