# D-REX: A Benchmark for Detecting Deceptive Reasoning in Large Language Models

arXiv: https://arxiv.org/abs/2509.17938 · HTML: https://mlsanity.com/paper/2509.17938/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Placed in Reasoning for agents, planning and code, one of 17 directions in LLM reasoning.
- Builds on 14 of its 24 references that are in the map.
- First posted on arXiv 22 Sep 2025.

## Placed in

- LLM reasoning / [Reasoning for agents, planning and code](https://mlsanity.com/lr/reasoning-for-agents-planning-and-code/)
- LLM alignment / RLHF / preference optimisation / [Safety alignment and jailbreak attacks](https://mlsanity.com/alignment/safety-alignment-and-jailbreak-attacks/)
- LLM safety and security / [Jailbreaking](https://mlsanity.com/safety/jailbreaking/)

## Builds on (most cited first, 14 of 14)

- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2501.12948/) (2025)
- [Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities](https://mlsanity.com/paper/2507.06261/) (2025)
- [Universal and Transferable Adversarial Attacks on Aligned Language Models](https://mlsanity.com/paper/2307.15043/) (2023)
- [Ignore Previous Prompt: Attack Techniques For Language Models](https://mlsanity.com/paper/2211.09527/) (2022)
- [Safe RLHF: Safe Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2310.12773/) (2023)
- [Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training](https://mlsanity.com/paper/2401.05566/) (2024)
- [A StrongREJECT for Empty Jailbreaks](https://mlsanity.com/paper/2402.10260/) (2024)
- [MasterKey: Automated Jailbreak Across Multiple Large Language Model Chatbots](https://mlsanity.com/paper/2307.08715/) (2023)
- [Constitutional Classifiers: Defending against Universal Jailbreaks across Thousands of Hours of Red Teaming](https://mlsanity.com/paper/2501.18837/) (2025)
- [AILuminate: Introducing v1.0 of the AI Risk and Reliability Benchmark from MLCommons](https://mlsanity.com/paper/2503.05731/) (2025)
- [Virology Capabilities Test (VCT): A Multimodal Virology Q&A Benchmark](https://mlsanity.com/paper/2504.16137/) (2025)
- [BeHonest: Benchmarking Honesty in Large Language Models](https://mlsanity.com/paper/2406.13261/) (2024)
- [OpenDeception: Learning Deception and Trust in Human-AI Interaction via Multi-Agent Simulation](https://mlsanity.com/paper/2504.13707/) (2025)

## Built on by (most cited first, 7 of 7)

- [MoReBench: Evaluating Procedural and Pluralistic Moral Reasoning in Language Models, More than Outcomes](https://mlsanity.com/paper/2510.16380/) (2025)
- [How Vulnerable Are AI Agents to Indirect Prompt Injections? Insights from a Large-Scale Public Competition](https://mlsanity.com/paper/2603.15714/) (2026)
- [Toward Reliable, Safe, and Secure LLMs for Scientific Applications](https://mlsanity.com/paper/2603.18235/) (2026)
- [Tatemae: Detecting Alignment Faking via Tool Selection in LLMs](https://mlsanity.com/paper/2604.26511/) (2026)
- [Spot The Ball: A Benchmark for Visual Social Inference](https://mlsanity.com/paper/2511.00261/) (2025)
- [A Model of Multi-turn Human Persuadability Using Probabilistic Belief Tracing](https://mlsanity.com/paper/2606.05330/) (2026)
- [Towards Mitigating Deceptive Safety Alignment in Large Reasoning Models](https://mlsanity.com/paper/2609.36254/) (2026)

## Abstract (arXiv)

The safety and alignment of Large Language Models (LLMs) are critical for their responsible deployment. Current evaluation methods predominantly focus on identifying and preventing overtly harmful outputs. However, they often fail to address a more insidious failure mode: models that produce benign-appearing outputs while operating on malicious or deceptive internal reasoning. This vulnerability, often triggered by sophisticated system prompt injections, allows models to bypass conventional safety filters, posing a significant, underexplored risk. To address this gap, we introduce the Deceptive Reasoning Exposure Suite (D-REX), a novel dataset designed to evaluate the discrepancy between a model's internal reasoning process and its final output. D-REX was constructed through a competitive red-teaming exercise where participants crafted adversarial system prompts to induce such deceptive behaviors. Each sample in D-REX contains the adversarial system prompt, an end-user's test query, the model's seemingly innocuous response, and, crucially, the model's internal chain-of-thought, which reveals the underlying malicious intent. Our benchmark facilitates a new, essential evaluation task: the detection of deceptive alignment. We demonstrate that D-REX presents a significant challenge for existing models and safety mechanisms, highlighting the urgent need for new techniques that scrutinize the internal processes of LLMs, not just their final outputs.