TCAR-Gen: Temporal Graph Retrieval with Evidence Fusion for Knowledge-Grounded Generation

In three linesTCAR-Gen combines query-conditioned graph neural networks, temporal evidence fusion, and chain-of-trees reasoning for retrieval-augmented generation. Achieves 0.3738 Recall@5 on Victorian Crime Diaries benchmark, outperforming Vanilla RAG, Temporal RAG, and GraphRAG variants. Cross-model evaluation across GPT-OSS 20B to TinyLlama 1.1B shows robust retrieval coverage at smaller scales.

Summary generated by Claude — human-verified