TurnGate is a response-aware defense mechanism designed to detect and mitigate hidden malicious intent in multi-turn dialogue systems. Defending state-of-the-art multi-turn malicious attacks like
CKA-Agent.
TurnGate is a specialized monitor designed to detect hidden malicious intent in multi-turn dialogues. Unlike traditional filters that look at queries in isolation, TurnGate is response-aware: it inspects the assistant's candidate response in the context of the full dialogue history to identify the precise "closure turn" where a harmful objective becomes actionable.
This repository contains the weights for TurnGate-0.1, a model trained on the Multi-Turn Intent Dataset (MTID) and optimized via reinforcement learning with turn-level process rewards.
If you find this repository useful for your research, please consider citing the following paper:
1@misc{shen2026turnlateresponseawaredefense,
2 title={One Turn Too Late: Response-Aware Defense Against Hidden Malicious Intent in Multi-Turn Dialogue},
3 author={Xinjie Shen and Rongzhe Wei and Peizhi Niu and Haoyu Wang and Ruihan Wu and Eli Chien and Bo Li and Pin-Yu Chen and Pan Li},
4 year={2026},
5 eprint={2605.05630},
6 archivePrefix={arXiv},
7 primaryClass={cs.CL},
8 url={https://arxiv.org/abs/2605.05630},
9}
