Relevance Is Not Permission: Warranted Attention for Value Contributions

2026-06-29

AI Summary

Although attention mechanisms can identify relevant key-value pairs, they do not guarantee that the corresponding values actually influence predictions, resulting in a “warrant gap.” This work formally characterizes this issue for the first time and introduces the Warrant mechanism, which retains original attention weights while incorporating learnable query-item warrant factors $g_{ij}$. These factors enable path-localized warrant gating to dynamically regulate value propagation paths, activating predictive contributions only when explicitly warranted. The proposed mechanism is seamlessly integrated into diverse models—including CTDG, MTPP, RAG, STPP, and TKG—and evaluated across 32 experimental settings. It achieves improvements in the primary metric in 27 cases (10 significantly), with notable gains such as a 0.1076 increase in CTDG AUC and a 0.0683 improvement in TKG MRR, demonstrating both the effectiveness and broad applicability of warrant-based path localization.

Abstract

Relevance is not permission. Attention lets a model read key-value items related to the current query, but it does not guarantee that the value contribution of such an item becomes prediction evidence. A retrieved passage may be relevant to a question without being supporting evidence, and a historical fact or temporal neighbor may even blur true-tail ranking or the current edge score. This paper formalizes this gap as a permission problem for the weighted value term alpha_ij * v_j that is actually added to the prediction path. We propose Warrant, a path-localized interface that preserves attention relevance alpha_ij, exposes the value path leading to the primary metric, and, in the full model, turns alpha_ij * v_j into alpha_ij * g_ij * v_j through learned query-item permission g_ij. We place the same operator on the metric-defining value paths of CTDG link prediction, MTPP next-mark ranking, RAG supporting evidence selection, STPP next-location forecasting, and TKG tail prediction. Across 32 paired comparisons, 3 seeds, and 192 total runs, Warrant improves the primary metric in 27 comparisons; practical tiers consist of 10 substantial effects, 1 marginal effect, 8 positive but uncertain effects, 8 tie/negligible effects, and 5 drops. In the path-localization check, correct-path placement outperforms direction-aware Base performance in every domain and exceeds generic attention placement by +0.1076 AUC in CTDG and +0.0683 MRR in TKG. Ablations show that most TKG gains come from historical-tail value path exposure, whereas the core CTDG gain comes from edge-conditioned query-item permission. In conclusion, prediction evidence is not attention mass. A weighted value term becomes evidence only when it is warranted on the path to the metric.

Problem

• attention
• value contribution
• prediction evidence
• relevance
• permission

Innovation

• Warrant
• attention permission
• value contribution
• path-localized interface
• evidence-aware prediction