eval/reference/answer_1b.json

{
  "config": {
    "dataset": "hotpotqa-distractor",
    "model": "llama3.2:1b",
    "n_docs": 491,
    "n_queries": 50
  },
  "stacks": {
    "Stack 1 — Vectoriel (FAISS)": {
      "ndcg@10": 0.8031,
      "em": 0.0,
      "f1": 0.0793,
      "n_queries": 50
    },
    "Stack 2 — Hybride (BM25 + RRF)": {
      "ndcg@10": 0.8123,
      "em": 0.0,
      "f1": 0.0586,
      "n_queries": 50
    },
    "Stack 3 — Graphe (networkx)": {
      "ndcg@10": 0.7786,
      "em": 0.02,
      "f1": 0.0673,
      "n_queries": 50
    }
  }
}