mlsanityData to 7 Oct 2026

Beyond "I cannot fulfill this request": Alleviating Rigid Rejection in LLMs via Label Enhancement

arXiv 2605.07883 ↗First posted 8 May 2026cs.CLcited by 0

On the map: LLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks

also in 1 more
Abstract (arXiv)

Large Language Models (LLMs) rely on safety alignment to obey safe requests while refusing harmful ones. However, traditional refusal mechanisms often lead to "rigid rejection," where a general template (e.g., "I cannot fulfill this request") indiscriminately triggers refusals and severely undermines the naturalness of interactions between humans and LLMs. To address this issue, LANCE is proposed in this paper to ensure safe yet flexible and natural responses via label enhancement. Specifically, LANCE employs variational inference to perform label enhancement, predicting a continuous distribution across multiple rejection categories. These fine-grained rejection distributions provide multi-way textual gradients for a refinement model to neutralize the hazardous elements in the prompt, so that the LLMs could generate safe responses that avoid rigid rejections while preserving the naturalness of interactions. Experiments demonstrate that LANCE significantly alleviates the rigid rejection problem while maintaining high security standards, significantly outperforming existing baseline models in terms of helpfulness and naturalness of responses.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Safety alignment and jailbreak attacks, one of 10 directions in LLM alignment / RLHF / preference optimisation. Builds on 25 of its 27 references that are in the map. First posted on arXiv 8 May 2026.

Builds on

25 of 27 references.

Closest ancestors · 4

Same direction, newest first

  1. 2025Understanding Refusal in Language Models with Sparse Autoencoders29 May
  2. 2024OR-Bench: An Over-Refusal Benchmark for Large Language Models31 May
  3. 2023Universal Jailbreak Backdoors from Poisoned Human Feedback24 Nov
  4. 2023BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Dataset10 Jul

Other references · 21

  1. 2022Training language models to follow instructions with human feedbackcited by 15,712
  2. 2024The Llama 3 Herd of Modelscited by 14,192
  3. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
  4. 2023Self-Refine: Iterative Refinement with Self-Feedbackcited by 3,317
  5. 2024Qwen2 Technical Reportcited by 2,076
  6. 2023Representation Engineering: A Top-Down Approach to AI Transparencycited by 1,239
  7. 2023Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversationscited by 1,124
  8. 2023Steering Llama 2 via Contrastive Activation Additioncited by 1,009
  9. 2024Refusal in Language Models Is Mediated by a Single Directioncited by 960
  10. 2023Large Language Models as Optimizerscited by 830
  11. 2023Instruction Tuning for Large Language Models: A Surveycited by 617
  12. 2023XSTest: A Test Suite for Identifying Exaggerated Safety Behaviours in Large Language Modelscited by 477
9 more
  1. 2024TextGrad: Automatic "Differentiation" via Textcited by 250
  2. 2023Evaluating Large Language Models: A Comprehensive Surveycited by 200
  3. 2023SafetyBench: Evaluating the Safety of Large Language Modelscited by 189
  4. 2024Refuse Whenever You Feel Unsafe: Improving Safety in LLMs via Decoupled Refusal Trainingcited by 67
  5. 2024ShieldLM: Empowering LLMs as Aligned, Customizable and Explainable Safety Detectorscited by 48
  6. 2022One Positive Label is Sufficient: Single-Positive Multi-Label Learning with Label Enhancementcited by 22
  7. 2025DLPO: Towards a Robust, Efficient, and Generalizable Prompt Optimization Framework from a Deep-Learning Perspectivecited by 7
  8. 2025Prompt Optimization and Evaluation for LLM Automated Red Teamingcited by 1
  9. 2025Learning to Extract Context for Context-Aware LLM Inferencecited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025From Threat to Tool: Leveraging Refusal-Aware Injection Attacks for Safety Alignment
  2. 2026LLMs Can Unlearn Refusal with Only 1,000 Benign Samples
  3. 2025The Geometry of Refusal in Large Language Models: Concept Cones and Representational Independence
  4. 2025FalseReject: A Resource for Improving Contextual Safety and Mitigating Over-Refusals in LLMs via Structured Reasoning
  5. 2026Discern Truth from Falsehood: Reducing Over-Refusal via Contrastive Refinement
  6. 2026Latent-space Attacks for Refusal Evasion in Language Models
  7. 2025Just Enough Shifts: Mitigating Over-Refusal in Aligned Language Models with Targeted Representation Fine-Tuning
  8. 2024Self and Cross-Model Distillation for LLMs: Effective Methods for Refusal Pattern Alignment