paper-with-me

Papers

IHDec: Divergence-Steered Contrastive Decoding for Securing Multi-Turn Instruction Hierarchies

2026-06-29 · Nicole Geumheon Liu, Haeun Jang, Yonghyun Jun, Hwanhee Lee arxiv

Large Language Models (LLMs) often fail to maintain instruction hierarchies (IH) when processing multi-source inputs with varying role-level priorities, paradoxically adhering to lower-priority directives during conflicts. While existing defenses mitigate this issue, they are largely restricted to single-turn scenarios and require expensive fine-tuning. In this paper, we formalize this failure mode in multi-turn contexts via a Jensen-Shannon Divergence (JSD) framework, uncovering a pervasive role-influence inversion phenomenon where subordinate inputs override superior roles. To rectify this without training, we propose IHDec (Instruction Hierarchy-steered Decoding). IHDec leverages JSD to automatically detect token-level hierarchy violations and dynamically executes contrastive decoding to suppress misaligned subordinate roles. Extensive evaluations demonstrate that IHDec outperforms training-based baselines in multi-turn conflicts while fully preserving general response quality. Furthermore, IHDec strengthens safety against adversarial prompt injections and exhibits a robust scaling synergy with larger models. The Code is available at https://github.com/nxcolelxu/IHDec.git

📄 PDF Abstract BibTeX arXiv:2606.29960

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Divergence Decoding: Inference-Time Unlearning via Auxiliary Models

2026-05-29 · Humzah Merchant, Bradford Levy arxiv

Large Language Models (LLMs) frequently memorize sensitive training data thereby creating significant privacy and copyright risks. Addressing these risks, i.e., removing such knowledge from an existing model checkpoint, …

Text Generation

Mechanistic Indicators of Steering Effectiveness in Large Language Models

2026-02-02 · Mehdi Jafari, Hao Xue, Flora Salim arxiv

Activation-based steering enables Large Language Models (LLMs) to exhibit targeted behaviors by intervening on intermediate activations without retraining. Despite its widespread use, the mechanistic factors that govern …

APCD: Adaptive Path-Contrastive Decoding for Reliable Large Language Model Generation

2026-05-10 · Tianyu Zheng, Hong Wu, Jiaji Zhong arxiv

Large language models (LLMs) often suffer from hallucinations due to error accumulation in autoregressive decoding, where suboptimal early token choices misguide subsequent generation. Although multi-path decoding can im…

Enhancing Learned Knowledge in LoRA Adapters Through Efficient Contrastive Decoding on Ascend NPUs

2025-05-20 · Morgan Lindsay Heisler, Linzi Xing, Ge Shi, Hanieh Sadri 외

Huawei Cloud users leverage LoRA (Low-Rank Adaptation) as an efficient and scalable method to fine-tune and customize large language models (LLMs) for application-specific needs. However, tasks that require complex reaso…

Attention-Guided Layer Selection for Contrastive Decoding in Large Language Models

2026-07-25 · Yusuke Sakai, Natthawut Kertkeidkachorn, Kiyoaki Shirai arxiv

Contrastive decoding methods such as DoLa improve the factuality of Large Language Models (LLMs) by contrasting the output distributions of mature and premature layers. However, DoLa's dynamic layer selection relies sole…