# Privacy-preserved LLM Cascade via CoT-enhanced Policy Learning

arXiv: https://arxiv.org/abs/2410.08014 · HTML: https://mlsanity.com/paper/2410.08014/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Builds on 35 of its 60 references that are in the map.
- First posted on arXiv 10 Oct 2024.

## Builds on (most cited first, 30 of 35)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context](https://mlsanity.com/paper/2403.05530/) (2024)
- [FrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performance](https://mlsanity.com/paper/2305.05176/) (2023)
- [Do Prompt-Based Models Really Understand the Meaning of their Prompts?](https://mlsanity.com/paper/2109.01247/) (2021)
- [RouterBench: A Benchmark for Multi-LLM Routing System](https://mlsanity.com/paper/2403.12031/) (2024)
- [Large Language Model Routing with Benchmark Datasets](https://mlsanity.com/paper/2309.15789/) (2023)
- [Large Language Model Cascades with Mixture of Thoughts Representations for Cost-efficient Reasoning](https://mlsanity.com/paper/2310.03094/) (2023)
- [Towards Safer Large Language Models through Machine Unlearning](https://mlsanity.com/paper/2402.10058/) (2024)
- [Language Model Cascades: Token-level uncertainty and beyond](https://mlsanity.com/paper/2404.10136/) (2024)
- [Vidur: A Large-Scale Simulation Framework For LLM Inference](https://mlsanity.com/paper/2405.05465/) (2024)
- [On-Device Language Models: A Comprehensive Review](https://mlsanity.com/paper/2409.00088/) (2024)
- [Towards LLM-based Fact Verification on News Claims with a Hierarchical Step-by-Step Prompting Method](https://mlsanity.com/paper/2310.00305/) (2023)
- [Fly-Swat or Cannon? Cost-Effective Language Model Choice via Meta-Modeling](https://mlsanity.com/paper/2308.06077/) (2023)
- [Offline Multi-Action Policy Learning: Generalization and Optimization](https://mlsanity.com/paper/1810.04778/) (2018)
- [When Does Confidence-Based Cascade Deferral Suffice?](https://mlsanity.com/paper/2307.02764/) (2023)
- [LLM-based Medical Assistant Personalization with Short- and Long-Term Memory Coordination](https://mlsanity.com/paper/2309.11696/) (2023)
- [LLMCad: Fast and Scalable On-device Large Language Model Inference](https://mlsanity.com/paper/2309.04255/) (2023)
- [LLaMoCo: Instruction Tuning of Large Language Models for Optimization Code Generation](https://mlsanity.com/paper/2403.01131/) (2024)
- [Towards Optimizing the Costs of LLM Usage](https://mlsanity.com/paper/2402.01742/) (2024)
- [Personalized LLM Response Generation with Parameterized Memory Injection](https://mlsanity.com/paper/2404.03565/) (2024)
- [Reflection-Tuning: Data Recycling Improves LLM Instruction-Tuning](https://mlsanity.com/paper/2310.11716/) (2023)
- [Exposing Privacy Gaps: Membership Inference Attack on Preference Data for LLM Alignment](https://mlsanity.com/paper/2407.06443/) (2024)
- [EmailSum: Abstractive Email Thread Summarization](https://mlsanity.com/paper/2107.14691/) (2021)
- [Learning to Refuse: Towards Mitigating Privacy Risks in LLMs](https://mlsanity.com/paper/2407.10058/) (2024)
- [Cascade-Aware Training of Language Models](https://mlsanity.com/paper/2406.00060/) (2024)
- [No Free Lunch Theorem for Privacy-Preserving LLM Inference](https://mlsanity.com/paper/2405.20681/) (2024)
- [Efficient Contextual LLM Cascades through Budget-Constrained Policy Learning](https://mlsanity.com/paper/2404.13082/) (2024)

## Built on by (most cited first, 8 of 8)

- [A Survey on Collaborating Small and Large Language Models for Performance, Cost-effectiveness, Cloud-edge Privacy, and Trustworthiness](https://mlsanity.com/paper/2510.13890/) (2025)
- [PrivAct: Internalizing Contextual Privacy Preservation via Multi-Agent Preference Training](https://mlsanity.com/paper/2602.13840/) (2026)
- [Harnessing Chain-of-Thought Metadata for Task Routing and Adversarial Prompt Detection](https://mlsanity.com/paper/2503.21464/) (2025)
- [Economic Evaluation of LLMs](https://mlsanity.com/paper/2507.03834/) (2025)
- [Honesty over Accuracy: Trustworthy Language Models through Reinforced Hesitation](https://mlsanity.com/paper/2511.11500/) (2025)
- [Is Escalation Worth It? On the Depth of LLM Cascades](https://mlsanity.com/paper/2605.06350/) (2026)
- [When Cloud Agents Meet Device Agents: Lessons from Hybrid Multi-Agent Systems](https://mlsanity.com/paper/2605.30102/) (2026)
- [SEAL-Tag: Self-Tag Evidence Aggregation with Probabilistic Circuits for PII-Safe Retrieval-Augmented Generation](https://mlsanity.com/paper/2603.17292/) (2026)

## Abstract (arXiv)

Large Language Models (LLMs) have gained significant attention in on-device applications due to their remarkable performance across real-world tasks. However, on-device LLMs often suffer from suboptimal performance due to hardware limitations. A promising solution to this challenge is cascading a weaker local (on-device) LLM with a more powerful server LLM. While existing research on LLM cascade primarily optimizes the performance-cost trade-off, real-world applications impose additional requirements, such as privacy preservation, which remain largely unaddressed. In this work, we move beyond existing confidence- and logit-based LLM cascade methods and propose mathbfP3Defer, a novel Chain-of-Thought (CoT)-enhanced policy learning framework for privacy-preserved deferral decision-making. Our approach effectively improves cascade efficiency while mitigating privacy risks. Extensive experiments on three benchmark datasets demonstrate the effectiveness and superiority of mathbfP3Defer over existing methods.