Metraipolis › archive › dc_d671d9fccec0

Per-Query Gating of LLM Rerankers for Multi-Hop Retrieval — arXiv:2609.22880

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.22880
License (as archived)
CC0-1.0
Archived
2026-10-09T06:50:29.419Z
SHA-256 of the archived text
90a8199e624512fd095fc910f6f7e219abb796f06498dedc9268597f91d66209
Stable URL of this page
https://metraipolis.com/documento/dc_d671d9fccec0

How to cite

Per-Query Gating of LLM Rerankers for Multi-Hop Retrieval — arXiv:2609.22880 — arXiv (en), https://arxiv.org/abs/2609.22880 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d671d9fccec0 (Metraipolis archive dc_d671d9fccec0, sha256 90a8199e624512fd095fc910f6f7e219abb796f06498dedc9268597f91d66209)
BibTeX
@misc{dcd671d9fccec0,
  title = {Per-Query Gating of LLM Rerankers for Multi-Hop Retrieval — arXiv:2609.22880},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.22880},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d671d9fccec0; sha256 90a8199e624512fd095fc910f6f7e219abb796f06498dedc9268597f91d66209},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.22880. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

LLM rerankers add of the order of \$0.2-0.3 per 1,000 queries and about a second of tail latency on top of a graph-augmented dense pipeline such as HippoRAG2, and on three multi-hop benchmarks they improve final-hop top-K coverage on seven of nine (dataset, K) cells, by up to +34.8 pp. We ask whether a learned per-query gate can skip the reranker where it will not help, using only features available before the LLM call (27 score and lexical statistics of the two retrieval lists plus a PCA of a small query embedding) with an executable fallback. Every choice, including the fallback and the threshold, is made inside the training fold and applied once to held-out queries, and harmful skips (the rerank would have found the target, the fallback did not) are reported next to the aggregate coverage. Across nine cells on 2WikiMultiHopQA, MuSiQue and HotpotQA the gate skips 51% of calls at an average held-out LastHop@K cost of 1.2 pp; four cells meet a pre-registered 1 pp rule, harmful skips occur in eight (190 harmful against 136 beneficial), and a random gate at the same skip rate loses 2 to 11 pp on the high-lift cells. A second rule sets each cell's threshold from a pre-specified budget on the expected harmful-skip rate over Platt-calibrated harm probabilities (ECE 0.025 after calibration, 0.094 before): at a 1 pp budget the gate skips 42% at -0.8 pp with 66 harmful skips and six cells within 1 pp, but realised harm exceeds the promise in six cells (mean 1.45 vs 0.83 pp), a selection optimism we quantify; a 0.5 pp budget realises about 1 pp. The harm probabilities are calibrated but barely discriminative (AUC 0.16 to 0.70). An earlier version reported 73% "lossless" savings; that figure rested on an oracle fallback and a wrong MuSiQue target, and we document both.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-22 in cs.CL).
Categorie arXiv: Information Retrieval (cs.IR) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.22880 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-09T06:48:45.107Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-22?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).