mlsanityData to 14 Oct 2026

Do LLM Modules Generalize? A Study on Motion Generation for Autonomous Driving

arXiv 2509.02754 ↗First posted 2 Sep 2025cs.AIcited by 4

On the map: End-to-end autonomous driving → Large language models for driving

Abstract (arXiv)

Recent breakthroughs in large language models (LLMs) have not only advanced natural language processing but also inspired their application in domains with structurally similar problems--most notably, autonomous driving motion generation. Both domains involve autoregressive sequence modeling, token-based representations, and context-aware decision making, making the transfer of LLM components a natural and increasingly common practice. However, despite promising early attempts, a systematic understanding of which LLM modules are truly transferable remains lacking. In this paper, we present a comprehensive evaluation of five key LLM modules--tokenizer design, positional embedding, pre-training paradigms, post-training strategies, and test-time computation--within the context of motion generation for autonomous driving. Through extensive experiments on the Waymo Sim Agents benchmark, we demonstrate that, when appropriately adapted, these modules can significantly improve performance for autonomous driving motion generation. In addition, we identify which techniques can be effectively transferred, analyze the potential reasons for the failure of others, and discuss the specific adaptations needed for autonomous driving scenarios. We evaluate our method on the Sim Agents task and achieve competitive results.

Key facts. Cited by 4 papers in the map (data through 14 Oct 2026). Placed in Large language models for driving, one of 9 directions in End-to-end autonomous driving. Builds on 49 of its 63 references that are in the map. First posted on arXiv 2 Sep 2025.

Builds on

49 of 63 references.

References · 49

  1. 2017Attention Is All You Needcited by 60,361
  2. 2022Training language models to follow instructions with human feedbackcited by 15,971
  3. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 14,132
  4. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  5. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 9,379
  6. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 8,065
  7. 2020Scaling Laws for Neural Language Modelscited by 6,586
  8. 2021RoFormer: Enhanced Transformer with Rotary Position Embeddingcited by 4,703
  9. 2017Deep reinforcement learning from human preferencescited by 3,912
  10. 2022Training Compute-Optimal Large Language Modelscited by 3,019
  11. 2019The Curious Case of Neural Text Degenerationcited by 2,916
  12. 2025DAPO: An Open-Source LLM Reinforcement Learning System at Scalecited by 2,810
37 more
  1. 2023A Survey of Large Language Modelscited by 2,576
  2. 2022Emergent Abilities of Large Language Modelscited by 2,512
  3. 2016Asynchronous Methods for Deep Reinforcement Learningcited by 2,258
  4. 2020DeBERTa: Decoding-enhanced BERT with Disentangled Attentioncited by 2,181
  5. 2019Fine-Tuning Language Models from Human Preferencescited by 2,150
  6. 2018SentencePiece: A simple and language independent subword tokenizer and detokenizer for Neural Text Processingcited by 1,989
  7. 2019Transformer-XL: Attentive Language Models Beyond a Fixed-Length Contextcited by 1,850
  8. 2018Hierarchical Neural Story Generationcited by 1,193
  9. 2022MaskGIT: Masked Generative Image Transformercited by 1,099
  10. 2019DialoGPT: Large-Scale Generative Pre-training for Conversational Response Generationcited by 842
  11. 2023Instruction Tuning for Large Language Models: A Surveycited by 623
  12. 2020VectorNet: Encoding HD Maps and Agent Dynamics from Vectorized Representationcited by 530
  13. 2021ByT5: Towards a token-free future with pre-trained byte-to-byte modelscited by 509
  14. 2023Statistical Rejection Sampling Improves Preference Optimizationcited by 323
  15. 2022Motion Transformer with Global Intention Localization and Local Movement Refinementcited by 270
  16. 2023Parting with Misconceptions about Learning-based Vehicle Motion Planningcited by 242
  17. 2021Charformer: Fast Character Transformers via Gradient-based Subword Tokenizationcited by 160
  18. 2023Waymax: An Accelerated, Data-Driven Simulator for Large-Scale Autonomous Driving Researchcited by 160
  19. 2023MotionLM: Multi-Agent Motion Forecasting as Language Modelingcited by 156
  20. 2022Imitation Is Not Enough: Robustifying Imitation with Reinforcement Learning for Challenging Driving Scenarioscited by 129
  21. 2022Self-conditioned Embedding Diffusion for Text Generationcited by 117
  22. 2023The Waymo Open Sim Agents Challengecited by 86
  23. 2024PLUTO: Pushing the Limit of Imitation Learning-based Planning for Autonomous Drivingcited by 84
  24. 2025Robust Autonomy Emerges from Self-Playcited by 67
  25. 2023Trajeglish: Traffic Modeling as Next-Token Predictioncited by 60
  26. 2023TrafficBots: Towards World Models for Autonomous Driving Simulation and Motion Predictioncited by 56
  27. 2024BehaviorGPT: Smart Agent Simulation for Autonomous Driving with Next-Patch Predictioncited by 50
  28. 2024Solving Motion Planning Tasks with a Scalable Generative Modelcited by 40
  29. 2024Closed-Loop Supervised Fine-Tuning of Tokenized Traffic Modelscited by 40
  30. 2023QCNeXt: A Next-Generation Framework For Joint Multi-Agent Trajectory Predictioncited by 31
  31. 2024SIMPL: A Simple and Efficient Multi-agent Motion Prediction Baseline for Autonomous Drivingcited by 27
  32. 2024CtRL-Sim: Reactive and Controllable Driving Agents with Offline Reinforcement Learningcited by 27
  33. 2024Improving Agent Behaviors with RL Fine-tuning for Autonomous Drivingcited by 27
  34. 2024KiGRAS: Kinematic-Driven Generative Model for Realistic Agent Simulationcited by 17
  35. 2025UniMM: A Unified Mixture Model Framework for Multi-Agent Simulationcited by 13
  36. 2020Character-Level Translation with Self-attentioncited by 12
  37. 2025DRoPE: Directional Rotary Position Embedding for Efficient Agent Interaction Modelingcited by 9

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 2; 2026: 2.

4 papers.

  1. 2025SPACeR: Self-Play Anchoring with Centralized Reference Modelscited by 7
  2. 2025DecompGAIL: Learning Realistic Traffic Behaviors with Decomposed Multi-Agent Generative Adversarial Imitation Learningcited by 5
  3. 2026AutoWorld: Learning Multi-Agent Traffic Simulation with Self-Supervised World Modelscited by 2
  4. 2026Bridging Local Observation and Global Simulation in Closed-Loop Traffic Modelingcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2023DriveMLM: Aligning Multi-Modal Large Language Models with Behavioral Planning States for Autonomous Driving
  2. 2023Driving with LLMs: Fusing Object-Level Vector Modality for Explainable Autonomous Driving
  3. 2024Large Language Models for Human-like Autonomous Driving: A Survey
  4. 2023LLM4Drive: A Survey of Large Language Models for Autonomous Driving
  5. 2024LLM4AD: Large Language Models for Autonomous Driving -- Concept, Review, Benchmark, Experiments, and Future Trends
  6. 2023LanguageMPC: Large Language Models as Decision Makers for Autonomous Driving
  7. 2024LimSim++: A Closed-Loop Platform for Deploying Multimodal LLMs in Autonomous Driving
  8. 2024Generalizing End-To-End Autonomous Driving In Real-World Environments Using Zero-Shot LLMs