# The Override Gap: A Magnitude Account of Knowledge Conflict Failure in Hypernetwork-Based Instant LLM Adaptation Source: [The Override Gap: A Magnitude Account of Knowledge Conflict Failure in Hypernetwork-Based Instant LLM Adaptation]() Collection index: [llms.txt]() Data updated: 2026-05-13T05:00:27.562Z Authors: Shuaizhi Cheng, Xiang Shi, Zhiwei Zhang, Mingwei Li Published: 2026-04-26T14:59:14.000Z ## Abstract Hypernetwork-based methods such as Doc-to-LoRA internalize a document into an LLM's weights in a single forward pass, but they fail systematically on conflicts: when the document contradicts pretraining knowledge, accuracy collapses to 46.4% on the deepest facts. We show the failure is a magnitude problem rather than a representational one. The hypernetwork already targets the right layers, but its adapter margin is approximately constant across documents while the pretrained margin grows with training frequency, so deep conflicts lose by construction. The account predicts that failure should track prior strength: sorting 194 conflicts by the base model's log-probability on the contradicted fact, baseline accuracy falls from 68% on weak-prior questions to 16% on strong-prior ones, a 52 percentage-point gap. The cure is amplitude. Selective Layer Boosting scales the adapter at its top-norm layers, and Conflict-Aware Internalization triggers boosting only when the base model is confident. Both are training-free; together they raise deep-conflict accuracy from 46.4% to 71.0% on Gemma-2B and from 53.6% to 72.5% on Mistral-7B while preserving novel-knowledge recall, and beat vanilla retrieval-augmented generation on medium conflicts by 18 percentage points despite operating entirely in parameter space. We release KID-Bench, a 489-question benchmark that separates novel recall, cross-knowledge combination, and prior-graded conflicts. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Knowledge Conflict Resolution on RippleEdits style 40 q](): Knowledge Conflict Resolution; RippleEdits style 40 q; Accuracy: 77.5. - [Knowledge Conflict Resolution on Held-out 30 q](): Knowledge Conflict Resolution; Held-out 30 q; Accuracy: 76.7. - [Knowledge Conflict Resolution on Counterfact-style 40 q](): Knowledge Conflict Resolution; CounterFact-style 40 q; Accuracy: 67.5. - [Knowledge Conflict Resolution on CounterFact 500](): Knowledge Conflict Resolution; CounterFact 500; Accuracy: 96.6. - [Knowledge Conflict Resolution on KID-Bench Category C v2](): Knowledge Conflict Resolution; KID-Bench Category C v2; Accuracy (C-Light): 78.1. - [Novel Knowledge Recall on KID-Bench Category A v2](): Novel Knowledge Recall; KID-Bench Category A v2; Accuracy: 97.1. - [Knowledge Conflict Resolution on KID-Bench v2](): Knowledge Conflict Resolution; KID-Bench v2; Performance (Difficulty A): 97.1. - [Knowledge Combination on KID-Bench Category B v2](): Knowledge Combination; KID-Bench Category B v2; Accuracy: 68. ## Original sources and implementation [Original paper]() [PDF]()