Metraipolis › archive › dc_4da69bf895fd
The Magnitude Mirage: Rethinking Confidence for Reasoning-Intensive Retrieval — arXiv:2609.15578
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.15578
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T01:52:25.856Z
- SHA-256 of the archived text
89703f881d0762fe182f8cd58662a79600ae95785255ba84f046f12c84c51f29- Stable URL of this page
https://metraipolis.com/documento/dc_4da69bf895fd
How to cite
The Magnitude Mirage: Rethinking Confidence for Reasoning-Intensive Retrieval — arXiv:2609.15578 — arXiv (en), https://arxiv.org/abs/2609.15578 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4da69bf895fd (Metraipolis archive dc_4da69bf895fd, sha256 89703f881d0762fe182f8cd58662a79600ae95785255ba84f046f12c84c51f29)
BibTeX
@misc{dc4da69bf895fd,
title = {The Magnitude Mirage: Rethinking Confidence for Reasoning-Intensive Retrieval — arXiv:2609.15578},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.15578},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4da69bf895fd; sha256 89703f881d0762fe182f8cd58662a79600ae95785255ba84f046f12c84c51f29},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.15578. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Many production RAG systems implement retrieval abstention by thresholding raw similarity scores, implicitly treating score magnitude as a confidence signal. We demonstrate that this practice degrades systematically as queries require reasoning beyond semantic matching. Across 11 retrieval architectures and 28 datasets, neural retrievers consistently assign high similarity scores to semantically related but constraint-violating documents, causing magnitude-based thresholds to collapse toward near-random abstention performance on logical and temporal reasoning tasks---a failure we term the Magnitude Mirage. To address this without computationally expensive alternatives, we conduct a large-scale empirical study of six zero-cost Query Performance Prediction (QPP) metrics across three cognitive tiers: semantic matching (BEIR), logical reasoning (BRIGHT), and temporal reasoning (TEMPO). Our central finding is that the key improvement comes from abandoning magnitude in favor of score-distribution signals: the gain from this shift exceeds the differences among distributional alternatives by a factor of 5-10$\times$. In particular, Score Gap ($s_1 - s_k$) and a practical adaptation of Score Magnitude and Variance (LSMV) improve abstention AUROC by up to 0.16 in settings where magnitude-based confidence provides little discriminative power. These methods require no additional inference, retraining, or latency, making them a practical zero-cost replacement for magnitude thresholding in deployed RAG systems. Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-15 in cs.IR). Categorie arXiv: Information Retrieval (cs.IR). Pagina del lavoro: https://arxiv.org/abs/2609.15578 — 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 2026-10-16T01:23:14.593Z dalla pagina https://arxiv.org/catchup/cs.IR/2026-09-15?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).