mlsanityData to 14 Oct 2026

FedDetox: Robust Federated SLM Alignment via On-Device Data Sanitization

arXiv 2604.06833 ↗First posted 8 Apr 2026cs.CR cs.LGcited by 0

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

also in 1 more
Abstract (arXiv)

As high quality public data becomes scarce, Federated Learning (FL) provides a vital pathway to leverage valuable private user data while preserving privacy. However, real-world client data often contains toxic or unsafe information. This leads to a critical issue we define as unintended data poisoning, which can severely damage the safety alignment of global models during federated alignment. To address this, we propose FedDetox, a robust framework tailored for Small Language Models (SLMs) on resource-constrained edge devices. We first employ knowledge distillation to transfer sophisticated safety alignment capabilities from large scale safety aligned teacher models into light weight student classifiers suitable for resource constrained edge devices. Specifically, during federated learning for human preference alignment, the edge client identifies unsafe samples at the source and replaces them with refusal templates, effectively transforming potential poisons into positive safety signals. Experiments demonstrate that our approach preserves model safety at a level comparable to centralized baselines without compromising general utility.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Safety alignment and jailbreak attacks, one of 10 directions in LLM alignment / RLHF / preference optimisation. Builds on 33 of its 34 references that are in the map. First posted on arXiv 8 Apr 2026.

Builds on

33 of 34 references.

Closest ancestors · 6

Same direction, newest first

  1. 2024PKU-SafeRLHF: Towards Multi-Level Safety Alignment for LLMs with Human Preference20 Jun
  2. 2024SORRY-Bench: Systematically Evaluating Large Language Model Safety Refusal20 Jun
  3. 2024Is poisoning a real threat to LLM alignment? Maybe more so than you think17 Jun
  4. 2024Emerging Safety Attack and Defense in Federated Instruction Tuning of Large Language Models15 Jun
  5. 2023Universal and Transferable Adversarial Attacks on Aligned Language Models27 Jul
  6. 2022Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned23 Aug

Other references · 27

  1. 2022Training language models to follow instructions with human feedbackcited by 15,971
  2. 2017Proximal Policy Optimization Algorithmscited by 15,215
  3. 2024The Llama 3 Herd of Modelscited by 14,715
  4. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  5. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 8,065
  6. 2020Measuring Massive Multitask Language Understandingcited by 7,963
  7. 2016Communication-Efficient Learning of Deep Networks from Decentralized Datacited by 7,669
  8. 2024Qwen2.5 Technical Reportcited by 4,942
  9. 2021TruthfulQA: Measuring How Models Mimic Human Falsehoodscited by 2,975
  10. 2023Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversationscited by 1,207
  11. 2023Tree of Attacks: Jailbreaking Black-Box LLMs Automaticallycited by 672
  12. 2023XSTest: A Test Suite for Identifying Exaggerated Safety Behaviours in Large Language Modelscited by 505
15 more
  1. 2023"Do Anything Now": Characterizing and Evaluating In-The-Wild Jailbreak Prompts on Large Language Modelscited by 489
  2. 2020MobileBERT: a Compact Task-Agnostic BERT for Resource-Limited Devicescited by 482
  3. 2023Orca: Progressive Learning from Complex Explanation Traces of GPT-4cited by 379
  4. 2022A Holistic Approach to Undesired Content Detection in the Real Worldcited by 356
  5. 2023ToxicChat: Unveiling Hidden Challenges of Toxicity Detection in Real-World User-AI Conversationcited by 231
  6. 2023Towards Building the Federated GPT: Federated Instruction Tuningcited by 185
  7. 2023Mitigating the Alignment Tax of RLHFcited by 183
  8. 2024Improving LoRA in Privacy-preserving Federated Learningcited by 161
  9. 2021Challenges and Approaches for Mitigating Byzantine Attacks in Federated Learningcited by 47
  10. 2024LoRA-FAIR: Federated LoRA Fine-Tuning with Aggregation and Initialization Refinementcited by 38
  11. 2023Backdoor Attacks and Defenses in Federated Learning: Survey, Challenges and Future Research Directionscited by 33
  12. 2024A Survey of Small Language Modelscited by 31
  13. 2017Byzantine-Tolerant Machine Learningcited by 19
  14. 2025PluralLLM: Pluralistic Alignment in LLMs via Federated Learningcited by 14
  15. 2026Revisiting Backdoor Threat in Federated Instruction Tuning from a Signal Aggregation Perspectivecited by 5

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026AlignFed: Alignment-Aware Asynchronous Federated Fine-Tuning for Large Language Models in Heterogeneous Edge Environments
  2. 2026Defending Against Backdoor Attacks via Alignment Checking in Model-Contrastive Federated Learning
  3. 2025EASE: Practical and Efficient Safety Alignment for Small Language Models
  4. 2026GradShield: Alignment Preserving Finetuning
  5. 2026Aggregation Alignment for Federated Learning with Mixture-of-Experts under Data Heterogeneity
  6. 2025Oblivionis: A Lightweight Learning and Unlearning Framework for Federated Large Language Models
  7. 2022FedCC: Robust Federated Learning against Model Poisoning Attacks
  8. 2025FedSODA: Federated Fine-tuning of LLMs via Similarity Group Pruning and Orchestrated Distillation Alignment