# An Empirical Study of Output-to-Input Loops for Black-Box Backdoor Detection in Fine-Tuned Open-Weight LLMs

arXiv: https://arxiv.org/abs/2608.11348 · HTML: https://mlsanity.com/paper/2608.11348/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Backdoor and data poisoning attacks, one of 12 directions in LLM safety and security.
- Builds on 24 of its 32 references that are in the map.
- First posted on arXiv 11 Aug 2026.

## Placed in

- LLM safety and security / [Backdoor and data poisoning attacks](https://mlsanity.com/safety/backdoor-and-data-poisoning-attacks/)

## Builds on (most cited first, 24 of 24)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [QLoRA: Efficient Finetuning of Quantized LLMs](https://mlsanity.com/paper/2305.14314/) (2023)
- [Mistral 7B](https://mlsanity.com/paper/2310.06825/) (2023)
- [Gemma 3 Technical Report](https://mlsanity.com/paper/2503.19786/) (2025)
- [LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models](https://mlsanity.com/paper/2403.13372/) (2024)
- [BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain](https://mlsanity.com/paper/1708.06733/) (2017)
- [Phi-4 Technical Report](https://mlsanity.com/paper/2412.08905/) (2024)
- [Fine-Pruning: Defending Against Backdooring Attacks on Deep Neural Networks](https://mlsanity.com/paper/1805.12185/) (2018)
- [Great, Now Write an Article About That: The Crescendo Multi-Turn LLM Jailbreak Attack](https://mlsanity.com/paper/2404.01833/) (2024)
- [The Curse of Recursion: Training on Generated Data Makes Models Forget](https://mlsanity.com/paper/2305.17493/) (2023)
- [ONION: A Simple and Effective Defense Against Textual Backdoor Attacks](https://mlsanity.com/paper/2011.10369/) (2020)
- [Composite Backdoor Attacks Against Large Language Models](https://mlsanity.com/paper/2310.07676/) (2023)
- [BITE: Textual Backdoor Attacks with Iterative Trigger Injection](https://mlsanity.com/paper/2205.12700/) (2022)
- [CROW: Eliminating Backdoors from Large Language Models via Internal Consistency Regularization](https://mlsanity.com/paper/2411.12768/) (2024)
- [Chain-of-Scrutiny: Detecting Backdoor Attacks for Large Language Models](https://mlsanity.com/paper/2406.05948/) (2024)
- [CleanGen: Mitigating Backdoor Attacks for Generation Tasks in Large Language Models](https://mlsanity.com/paper/2406.12257/) (2024)
- [Test-time Backdoor Mitigation for Black-Box Large Language Models with Defensive Demonstrations](https://mlsanity.com/paper/2311.09763/) (2023)
- [When Backdoors Speak: Understanding LLM Backdoor Attacks Through Model-Generated Explanations](https://mlsanity.com/paper/2411.12701/) (2024)
- [IMBERT: Making BERT Immune to Insertion-based Backdoor Attacks](https://mlsanity.com/paper/2305.16503/) (2023)
- [Mitigating Backdoor Threats to Large Language Models: Advancement and Challenges](https://mlsanity.com/paper/2409.19993/) (2024)

## Abstract (arXiv)

Anyone can upload a fine-tuned large language model (LLM) to a public repository and claim it is safe. A backdoored model behaves normally on ordinary inputs until a hidden trigger fires, and a user with no training data, clean reference weights, or the trigger phrase has no clear way to check the model before using it. We introduce and empirically evaluate self-feeding, a black-box test method that feeds a model's own output back as its next input, so the text drifts away from the starting prompt and toward the data the model was fine-tuned on. We test self-feeding against a repeated same-prompt baseline on six open-weight LLMs (3B-15B parameters), each fine-tuned with backdoors spanning eleven attack categories, using twenty ordinary starting prompts and chains of up to ten steps. Self-feeding finds backdoors in five of six models at 92.0% pooled precision, while the same-prompt baseline succeeds on only one of 120 prompt-model pairs; chains that begin with a joke request, an arithmetic question, or a coffee recipe all reach a trigger within a few steps. Recall per prompt is low (19.2%), and we show why it still adds up to much higher detection at the model level once several starting prompts are used. We also report where the method falls short: one model was never triggered, and self-feeding produced two false positives that the same-prompt baseline cannot produce. Cutting the chains to four steps keeps every model-level detection at 100% precision while using 60% fewer queries. Needing only text-level query access and a way to recognize malicious output, self-feeding offers a cheap first check on a downloaded model.