Combating Instruction Conflict via Energy-Driven Latent Conflict Detection

2026-09-08

AI Summary

为解决大型语言模型响应漂移问题，提出ELCD方法，通过构建复合隐藏状态表示并优化成对边缘排序目标，在生成后进行潜在冲突检测。

Abstract

Large Language Models (LLMs) are increasingly deployed with hierarchical instructions, yet they remain vulnerable to conflicts in which user directives override system-level constraints. Existing defense mechanisms predominantly focus on static input inspection and therefore fail to detect Response Drift, a phenomenon in which the model's final response violates system-level constraints despite seemingly compliant inputs. To bridge this gap, we introduce ELCD, a response-level latent conflict detector for post-generation, pre-delivery verification. Given the full generated output, ELCD constructs a composite hidden-state representation by concatenating the final-token embedding with the mean-pooled response embedding. It then optimizes a pairwise margin ranking objective to separate compliant and drifting responses in latent space. Extensive experiments across five mainstream LLMs ranging from 1.5B to 14B parameters demonstrate that ELCD significantly outperforms competitive baselines. Notably, it improves the PR-AUC on Llama-2-7B by approximately 30 percentage points and reduces the False Positive Rate at 95% TPR (FPR95) on Mistral-7B to 2.67%. These results suggest that ELCD provides a promising approach for latent instruction-conflict detection in open-weight or self-hosted LLM deployments.

Problem

• Instruction Conflict
• Response Drift
• Large Language Models
• System-level Constraints

Innovation

• ELCD
• Latent Conflict Detection
• Pairwise Margin Ranking
• Response Drift
• Hierarchical Instructions