# Language-Structured Relational Q-Learning for Threat-Aware Control in Safety-Critical Driving

arXiv: https://arxiv.org/abs/2608.11498 · HTML: https://mlsanity.com/paper/2608.11498/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in RL for autonomous driving and traffic, one of 15 directions in Deep reinforcement learning methods.
- Builds on 17 of its 23 references that are in the map.
- First posted on arXiv 11 Aug 2026.

## Placed in

- Deep reinforcement learning methods / [RL for autonomous driving and traffic](https://mlsanity.com/deeprl/rl-for-autonomous-driving-and-traffic/)

## Builds on (most cited first, 17 of 17)

- [Graph Attention Networks](https://mlsanity.com/paper/1710.10903/) (2017)
- [CARLA: An Open Urban Driving Simulator](https://mlsanity.com/paper/1711.03938/) (2017)
- [VectorNet: Encoding HD Maps and Agent Dynamics from Vectorized Representation](https://mlsanity.com/paper/2005.04259/) (2020)
- [TrafficSim: Learning to Simulate Realistic Multi-Agent Behaviors](https://mlsanity.com/paper/2101.06557/) (2021)
- [Generating Useful Accident-Prone Driving Scenarios via a Learned Traffic Prior](https://mlsanity.com/paper/2112.05077/) (2021)
- [AdvSim: Generating Safety-Critical Scenarios for Self-Driving Vehicles](https://mlsanity.com/paper/2101.06549/) (2021)
- [Language-Guided Traffic Simulation via Scene-Level Diffusion](https://mlsanity.com/paper/2306.06344/) (2023)
- [ChatScene: Knowledge-Enabled Safety-Critical Scenario Generation for Autonomous Vehicles](https://mlsanity.com/paper/2405.14062/) (2024)
- [Language Conditioned Traffic Generation](https://mlsanity.com/paper/2307.07947/) (2023)
- [Scenic: A Language for Scenario Specification and Scene Generation](https://mlsanity.com/paper/1809.09310/) (2018)
- [Learning to Collide: An Adaptive Safety-Critical Scenarios Generating Method](https://mlsanity.com/paper/2003.01197/) (2020)
- [SafeBench: A Benchmarking Platform for Safety Evaluation of Autonomous Vehicles](https://mlsanity.com/paper/2206.09682/) (2022)
- [CAT: Closed-loop Adversarial Training for Safe End-to-End Driving](https://mlsanity.com/paper/2310.12432/) (2023)
- [CtRL-Sim: Reactive and Controllable Driving Agents with Offline Reinforcement Learning](https://mlsanity.com/paper/2403.19918/) (2024)
- [Enhancing Autonomous Vehicle Training with Language Model Integration and Critical Scenario Generation](https://mlsanity.com/paper/2404.08570/) (2024)
- [Text2Scenario: Text-Driven Scenario Generation for Autonomous Driving Test](https://mlsanity.com/paper/2503.02911/) (2025)
- [GOOSE: Goal-Conditioned Reinforcement Learning for Safety-Critical Scenario Generation](https://mlsanity.com/paper/2406.03870/) (2024)

## Abstract (arXiv)

Natural-language-based scenario generation offers an intuitive means of describing rare and complex driving interactions, yet it is still uncertain whether training with language-structured data leads to truly adaptive control policies. We propose Language-Structured Relational Q-Learning, instantiated through an Ego-Centric Relational Q-Network (ERQ-Net), which jointly learns inter-vehicle relevance and action values from dynamic traffic graphs. Language descriptions define surrounding-vehicle behaviours during training, while prompts and semantic actor roles are hidden from the policy. ERQ-Net must therefore infer threat relevance solely from observable kinematics and interactions. Across 2,500 safety-critical scenarios, language-structured training improves test success from 49-52% to 55-58% and increases adversary-focused attention from 1.2x to 2.1x, demonstrating emergent threat awareness. However, this representational gain does not consistently translate into adaptive control: trained policies perform similarly to the best constant action, while a portfolio of simple policies solves 76% of scenarios. We formalise this discrepancy as a recognition-control gap and show that reward reweighting and margin shaping do not eliminate the resulting policy collapse. Evaluations of realism, criticality, semantic accuracy, and transfer of state-interface representations to CARLA further highlight both the strengths and the constraints of language-structured relational policy learning in safety-critical driving scenarios.