Instruction Following by Principled Boosting Attention of Large Language Models

ArXi:2506.13734v3 Announce Type: replace-cross Large language models' behavior is often shaped by instructions such as system prompts, refusal boundaries, privacy constraints, and tool-use rules that must hold at inference time. Yet in practice these constraints can be violated under long contexts or when user-provided context conflicts with them, creating reliability and safety risks. This motivates inference-time interventions that strengthen instruction influence without re