mlsanityData to 14 Oct 2026

Locket: Robust Feature-Locking Technique for Language Models

arXiv 2510.12117 ↗First posted 14 Oct 2025cs.CR cs.LGcited by 1

Not on the map. Why?

Abstract (arXiv)

Chatbot service providers (e.g., OpenAI) rely on tiered subscription plans to generate revenue, offering black-box access to basic models for free users and advanced models to paying subscribers. However, this approach is unprofitable and inflexible. A pay-to-unlock scheme for premium features (e.g., math, coding) offers a more sustainable alternative. Enabling such a scheme requires a feature-locking technique (FLoTE) that is (i) effective in refusing locked features, (ii) utility-preserving for unlocked features, (iii) robust against evasion or unauthorized credential sharing, and (iv) scalable to multiple features and clients. Existing FLoTEs (e.g., password-locked models) fail to meet these criteria. To fill this gap, we present Locket, the first robust and scalable FLoTE to enable pay-to-unlock schemes. We develop a framework for adversarial training and merging of feature-locking adapters, which enables Locket to selectively disable specific features of a model. Evaluation shows that Locket is effective (100% refusal rate), utility-preserving (≤ 7% utility degradation), robust (≤ 5% attack success rate), and scalable to multiple features and clients.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 32 of its 53 references that are in the map. First posted on arXiv 14 Oct 2025.

Builds on

32 of 53 references.

References · 32

  1. 2022Training language models to follow instructions with human feedbackcited by 15,971
  2. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 9,379
  3. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,944
  4. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,371
  5. 2024DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligencecited by 1,181
  6. 2023Enhancing Chat Language Models by Scaling High-quality Instructional Conversationscited by 839
  7. 2017Seq2SQL: Generating Structured Queries from Natural Language using Reinforcement Learningcited by 687
  8. 2024RouteLLM: Learning to Route LLMs with Preference Datacited by 682
  9. 2019SAMSum Corpus: A Human-annotated Dialogue Dataset for Abstractive Summarizationcited by 535
  10. 2024Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearningcited by 504
  11. 2024Improving Alignment and Robustness with Circuit Breakerscited by 322
  12. 2023Jailbreak in pieces: Compositional Adversarial Attacks on Multi-Modal Language Modelscited by 273
20 more
  1. 2023A Rank Stabilization Scaling Factor for Fine-Tuning with LoRAcited by 243
  2. 2024AutoDAN-Turbo: A Lifelong Agent for Strategy Self-Exploration to Jailbreak LLMscited by 199
  3. 2023Survey of Vulnerabilities in Large Language Models Revealed by Adversarial Attackscited by 194
  4. 2024Latent Adversarial Training Improves Robustness to Persistent Harmful Behaviors in LLMscited by 173
  5. 2024Task Singular Vectors: Reducing Task Interference in Model Mergingcited by 147
  6. 2023Understanding Catastrophic Forgetting in Language Models via Implicit Inferencecited by 144
  7. 2023Effective Prompt Extraction from Language Modelscited by 85
  8. 2022Adversarial Training for High-Stakes Reliabilitycited by 60
  9. 2024On Large Language Model Continual Unlearningcited by 52
  10. 2024LoRA Soups: Merging LoRAs for Practical Skill Composition Taskscited by 52
  11. 2024Persistent Pre-Training Poisoning of LLMscited by 48
  12. 2024Stress-Testing Capability Elicitation With Password-Locked Modelscited by 46
  13. 2025The Elicitation Game: Evaluating Capability Elicitation Techniquescited by 18
  14. 2025STAR: Spectral Truncation and Rescale for Model Mergingcited by 11
  15. 2025Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defensecited by 7
  16. 2024ModelLock: Locking Your Model With a Spellcited by 6
  17. 2024Pricing4SaaS: Towards a pricing model to drive the operation of SaaScited by 2
  18. 2025Identity Lock: Locking API Fine-tuned LLMs With Identity-based Wake Wordscited by 2
  19. 2025HORIZON: a Classification and Comparison Framework for Pricing-driven Feature Togglingcited by 1
  20. 2025Staining and locking computer vision models without retrainingcited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Toward Open Weight Models Without Risks: Separating Public and Private Capabilities in LLMscited by 0