analiza aiqbench
This arXiv preprint proposes Xetrieval, a method for explaining why dense retrieval models score documents as relevant. It uses auxiliary "feature elicitor" views to make document meaning more explicit before sparse autoencoders extract interpretable features, and it reports gains over an SAE baseline on feature influence and retrieval steering. For readers, the practical question is whether these results hold beyond the paper's benchmarks, and whether the interventions on embeddings reliably reflect the model's actual scoring behavior.
Xetrieval: Explaining Dense Retrieval through Feature Elicitation
arXiv:2605.29507v2 Announce Type: replace Abstract: Dense retrieval enables effective semantic matching, yet the semantic factors underlying its relevance scores remain opaque. Sparse autoencoders (SAEs) support feature-level explanations by decomposing dense embeddings into sparse semantic features. However, the document semantics underlying a retriever's relevance scores may not be sufficiently explicit in the embeddings to be recovered as interpretable SAE features. We introduce Xetrieval, a framework for explaining dense retrieval through feature elicitation. Feature Elicitors make retrieval-relevant aspects of document meaning more explicit through auxiliary views, facilitating their identification as interpretable SAE features. A shared sparse dictionary learns from these varied representations to express query--document relations through describable features. We assess their influence on relevance scores through interventions on the original document embeddings. Experiments show that Xetrieval yields more interpretable features than an SAE baseline and identifies feature sets with a stronger influence on the retriever's scores. Task-level feature modulation further demonstrates more effective retrieval steering. These findings highlight the value of feature elicitation for relevance attribution and retrieval steering.
Abstract:Dense retrieval enables effective semantic matching, yet the semantic factors underlying its relevance scores remain opaque. Sparse autoencoders (SAEs) support feature-level explanations by decomposing dense embeddings into sparse semantic features. However, the document semantics underlying a retriever's relevance scores may not be sufficiently explicit in the embeddings to be recovered as interpretable SAE features. We introduce Xetrieval, a framework for explaining dense retrieval through feature elicitation. Feature Elicitors make retrieval-relevant aspects of document meaning more explicit through auxiliary views, facilitating their identification as interpretable SAE features. A shared sparse dictionary learns from these varied representations to express query--document relations through describable features. We assess their influence on relevance scores through interventions on the original document embeddings. Experiments show that Xetrieval yields more interpretable features than an SAE baseline and identifies feature sets with a stronger influence on the retriever's scores. Task-level feature modulation further demonstrates more effective retrieval steering. These findings highlight the value of feature elicitation for relevance attribution and retrieval steering.
Artificial Intelligence (cs.AI); Information Retrieval (cs.IR)
arXiv:2605.29507 [cs.AI]
(or arXiv:2605.29507v2 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2605.29507 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 258 cuvinte
Subiecte conexe
From Dyad to Triad: Eliciting XAI Requirements in Stroke Rehabilitation
arXiv · 29 iul. 2026
Toward Evidence-Driven Human-Agent-Robot Teaming for Earth-Independent Anomaly Triage
arXiv · 8 oct. 2026
RES-MR: Risk-Aware Reasoning for Explainable and Safe Medication Recommendation
arXiv · 15 oct. 2026
Causal Chain of Thought: Mitigating Attention Leakage via Context Isolation
arXiv · 23 oct. 2026