% Key papers for Theory of autoregressive reasoning (scime atlas, https://atlas.scime.ai). Authors are not included: add them from the paper pages.

@inproceedings{neurips2023_language_67831,
  title = {{Language Models can Solve Computer Tasks}},
  booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
  year = {2023},
  url = {https://nips.cc/virtual/2023/poster/71929},
}

@inproceedings{aaai2020_hybrid_103246,
  title = {{Hybrid Compositional Reasoning for Reactive Synthesis from Finite-Horizon Specifications}},
  booktitle = {AAAI Conference on Artificial Intelligence (AAAI)},
  year = {2020},
  url = {https://ojs.aaai.org/index.php/AAAI/article/view/6528},
}

@inproceedings{neurips2023_guiding_68320,
  title = {{Guiding Large Language Models via Directional Stimulus Prompting}},
  booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
  year = {2023},
  url = {https://nips.cc/virtual/2023/poster/71484},
}

@inproceedings{neurips2025_t2i_76886,
  title = {{T2I-R1: Reinforcing Image Generation with Collaborative Semantic-level and Token-level CoT}},
  booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
  year = {2025},
  url = {https://openreview.net/forum?id=O7311P8YgP},
}

@inproceedings{icml2024_stay_44409,
  title = {{Stay on Topic with Classifier-Free Guidance}},
  booktitle = {International Conference on Machine Learning (ICML)},
  year = {2024},
  url = {https://icml.cc/virtual/2024/poster/34043},
}

@inproceedings{neurips2025_delving_79831,
  title = {{Delving into RL for Image Generation with CoT: A Study on DPO vs. GRPO}},
  booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
  year = {2025},
  url = {https://openreview.net/forum?id=tMf3keuPOl},
}

@inproceedings{iclr2024_amortizing_34350,
  title = {{Amortizing intractable inference in large language models}},
  booktitle = {International Conference on Learning Representations (ICLR)},
  year = {2024},
  url = {https://iclr.cc/virtual/2024/poster/18725},
}

@inproceedings{icml2024_reprompting_43809,
  title = {{Reprompting: Automated Chain-of-Thought Prompt Inference Through Gibbs Sampling}},
  booktitle = {International Conference on Machine Learning (ICML)},
  year = {2024},
  url = {https://icml.cc/virtual/2024/poster/34643},
}

@inproceedings{icml2024_auto_45083,
  title = {{Auto-Regressive Next-Token Predictors are Universal Learners}},
  booktitle = {International Conference on Machine Learning (ICML)},
  year = {2024},
  url = {https://icml.cc/virtual/2024/poster/33369},
}

@inproceedings{icml2025_demonstration_48815,
  title = {{Demonstration Selection for In-Context Learning via Reinforcement Learning}},
  booktitle = {International Conference on Machine Learning (ICML)},
  year = {2025},
  url = {https://icml.cc/virtual/2025/poster/43807},
}

@inproceedings{neurips2024_towards_72115,
  title = {{Towards a Theoretical Understanding of the 'Reversal Curse' via Training Dynamics}},
  booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
  year = {2024},
  url = {https://neurips.cc/virtual/2024/poster/95214},
}

@inproceedings{neurips2024_sled_74190,
  title = {{SLED: Self Logits Evolution Decoding for Improving Factuality in Large Language Models}},
  booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
  year = {2024},
  url = {https://neurips.cc/virtual/2024/poster/93358},
}

@inproceedings{neurips2025_language_78764,
  title = {{Language Models Can Predict Their Own Behavior}},
  booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
  year = {2025},
  url = {https://openreview.net/forum?id=i8IqEzpHaJ},
}

@inproceedings{iclr2025_limits_81326,
  title = {{Limits of Deep Learning: Sequence Modeling through the Lens of Complexity Theory}},
  booktitle = {International Conference on Learning Representations (ICLR)},
  year = {2025},
  url = {https://iclr.cc/virtual/2025/poster/30446},
}

@inproceedings{icml2025_task_48262,
  title = {{Task Generalization with Autoregressive Compositional Structure: Can Learning from \$D\$ Tasks Generalize to \$D^T\$ Tasks?}},
  booktitle = {International Conference on Machine Learning (ICML)},
  year = {2025},
  url = {https://icml.cc/virtual/2025/poster/44350},
}

@inproceedings{icml2026_frontier_49994,
  title = {{Frontier Models Can Take Actions at Low Probabilities}},
  booktitle = {International Conference on Machine Learning (ICML)},
  year = {2026},
  url = {https://openreview.net/forum?id=7yuPdMfO0M},
}

@inproceedings{icml2026_on_50239,
  title = {{On Robustness and Chain-of-Thought Consistency of RL-Finetuned VLMs}},
  booktitle = {International Conference on Machine Learning (ICML)},
  year = {2026},
  url = {https://openreview.net/forum?id=ANbnxDfcAD},
}

@inproceedings{icml2026_syntax_51507,
  title = {{Syntax vs. Semantics: How Transformers Learn Deep Dependencies}},
  booktitle = {International Conference on Machine Learning (ICML)},
  year = {2026},
  url = {https://openreview.net/forum?id=N0NjjQ9V3w},
}

@inproceedings{neurips2024_on_73834,
  title = {{On the Power of Decision Trees in Auto-Regressive Language Modeling}},
  booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
  year = {2024},
  url = {https://neurips.cc/virtual/2024/poster/93674},
}

@inproceedings{iclr2025_logicbreaks_83535,
  title = {{Logicbreaks: A Framework for Understanding Subversion of Rule-based Inference}},
  booktitle = {International Conference on Learning Representations (ICLR)},
  year = {2025},
  url = {https://iclr.cc/virtual/2025/poster/28281},
}

@inproceedings{iclr2026_token_84713,
  title = {{Token-Guard: Towards Token-Level Hallucination Control via Self-Checking Decoding}},
  booktitle = {International Conference on Learning Representations (ICLR)},
  year = {2026},
  url = {https://openreview.net/forum?id=5fCDEz43ya},
}

@inproceedings{iclr2026_self_87365,
  title = {{Self-Jailbreaking: Language Models Can Reason Themselves Out of Safety Alignment After Benign Reasoning Training}},
  booktitle = {International Conference on Learning Representations (ICLR)},
  year = {2026},
  url = {https://openreview.net/forum?id=akbtPEZnDZ},
}

@inproceedings{iclr2026_activationreasoning_87853,
  title = {{ActivationReasoning: Logical Reasoning in Latent Activation Spaces}},
  booktitle = {International Conference on Learning Representations (ICLR)},
  year = {2026},
  url = {https://openreview.net/forum?id=gGJh5AZTG7},
}