Adaptive Content Restriction for Large Language Models via Suffix Optimization

ACL ARR 2025 July Submission542 Authors

Abstract: Large Language Models (LLMs) have achieved remarkable success across various applications. However, restricting their outputs remains challenging due to the vastness of their generation space. While supervised fine-tuning (SFT) is commonly used to align models and prevent harmful content, content restriction needs often vary by user, evolve over time, and may not align with universal definitions of harm. Motivated by this need, we propose a new task called \textit{Adaptive Content Restriction} (AdaCoRe), which focuses on lightweight strategies -- methods without model fine-tuning -- to prevent deployed LLMs from generating restricted terms for specific use cases. We propose the first method for AdaCoRe, named \textit{Suffix Optimization (SOP)}, which appends a short, optimized suffix to any prompt to a) prevent a target LLM from generating a set of restricted terms, while b) preserving the output quality. To evaluate AdaCoRe approaches, including our SOP, we create a new \textit{Content Restriction Benchmark} (CoReBench), which contains 400 prompts for 80 restricted terms across 8 carefully selected categories. We demonstrate the effectiveness of SOP on CoReBench, which outperforms the system-level baselines such as system suffix by 15\%, 17\%, 10\%, 9\%, and 6\% on average restriction rates for Gemma2-2B, Mistral-7B, Vicuna-7B, Llama3-8B, and Llama3.1-8B, respectively. We also demonstrate that SOP is effective on POE, an online platform hosting various commercial LLMs, highlighting its practicality in real-world scenarios.

Paper Type: Long

Research Area: Machine Learning for NLP

Research Area Keywords: Adaptive Content Restriction, Suffix Optimization, Content Restriction Benchmark

Contribution Types: NLP engineering experiment

Languages Studied: English

Reassignment Request Area Chair: This is not a resubmission

Reassignment Request Reviewers: This is not a resubmission

A1 Limitations Section: This paper has a limitations section.

A2 Potential Risks: Yes

A2 Elaboration: In Appendix A

B Use Or Create Scientific Artifacts: No

B1 Cite Creators Of Artifacts: N/A

B2 Discuss The License For Artifacts: N/A

B3 Artifact Use Consistent With Intended Use: N/A

B4 Data Contains Personally Identifying Info Or Offensive Content: No

B5 Documentation Of Artifacts: N/A

B6 Statistics For Data: N/A

C Computational Experiments: Yes

C1 Model Size And Budget: Yes

C1 Elaboration: In Section 6.1

C2 Experimental Setup And Hyperparameters: Yes

C2 Elaboration: In Section 6.1

C3 Descriptive Statistics: N/A

C4 Parameters For Packages: N/A

D Human Subjects Including Annotators: No

D1 Instructions Given To Participants: No

D2 Recruitment And Payment: No

D3 Data Consent: No

D4 Ethics Review Board Approval: N/A

D5 Characteristics Of Annotators: N/A

E Ai Assistants In Research Or Writing: No

E1 Information About Use Of Ai Assistants: No

Author Submission Checklist: yes

Submission Number: 542