Graph‑S3 Boosts LLM Textual Graph Retrieval with Stepwise Supervision

Graph‑S3 adds stepwise supervision to LLM graph retrieval, raising accuracy ~8% and F1 ~9.7% on three textual‑graph QA benchmarks, especially multi‑hop queries. Code will be open‑source. getnews.me/graph-s3-boosts-llm-text... #graphretrieval #llm

GRIL: Integrated Graph Retrieval Boosts Large Language Model Reasoning

GRIL links graph retrieval to LLMs with expansion and pruning that learns from logits, achieving results on multi‑hop QA benchmarks and reducing hallucinations. Read more: getnews.me/gril-integrated-graph-re... #gril #graphretrieval #llm