Equal contribution is indicated by an asterisk.
Peer-Reviewed Journal and Conference Papers
[1] Pan, X.*, Hahami, E.*, Fan, J.*, Xie, Z., & Sompolinsky, H. (2026). Demasking Fine-Tuning for Knowledge Injection in Autoregressive and Diffusion LLMs. Transactions on Machine Learning Research (TMLR) (accepted; to appear).
[2] Pan, X., Hahami, E., Siegelmann, R., & Sompolinsky, H. (2026). A Complementary Learning System for Continual Episodic Memory in Large Language Models. eLife (accepted; to appear). Earlier preprint
[3] Hoy, W., Wang, B.*, & Pan, X.* (2026). Matching Accuracy, Different Geometry: Evolution Strategies vs GRPO in LLM Post-Training. Conference on Language Modeling (COLM 2026).
[4] Wang, B.*, Fan, J.*, & Pan, X.* (2026). Circuit Mechanisms for Spatial Relation Generation in Diffusion Transformers. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2026), Highlight.
[5] Sun, Y., Han, X., Zhang, W., …, Pan, X., …, & Song, D. (2026). Agents’ Last Exam. Advances in Neural Information Processing Systems (NeurIPS 2026).
[6] Pan, X.*, Fan, J.*, Xiong, Z., Hahami, E., Overwiening, J., & Xie, Z. (2026). User-Assistant Bias in LLMs. Findings of the Association for Computational Linguistics (ACL 2026).
[7] Pan, X., Philip, A., Xie, Z., & Schwartz, O. (2024). Dissecting Query-Key Interaction in Vision Transformers. Advances in Neural Information Processing Systems (NeurIPS 2024), Spotlight.
[8] Pan, X., Coen-Cagli, R., & Schwartz, O. (2024). Probing the Structure and Functional Properties of the Dropout-Induced Correlated Variability in Convolutional Neural Networks. Neural Computation, 36(4), 621–644.
[9] Pan, X., DeForge, A., & Schwartz, O. (2023). Generalizing Biological Surround Suppression Based on Center Surround Similarity via Deep Neural Network Models. PLoS Computational Biology, 19(9), e1011486.
[10] James, E. B.*, Pan, X.*, Schwartz, O., & Wilson, A. C. C. (2022). SymbiQuant: A Machine Learning Object Detection Tool for Polyploid Independent Estimates of Endosymbiont Population Size. Frontiers in Microbiology, 13, 816608.
[11] Geng, S., Ren, G., Pan, X., Zysman, J., & Ogihara, M. (2021). Sequential Modeling of Temporal Timbre Series for Popular Music Sub-Genre Analyses Using Deep Bidirectional Encoder Representations from Transformers. Audio Engineering Society 150th Convention.
Invited Publications
[12] Pan, X., & Schwartz, O. (2024). Multimodal AI Needs Active Human Interaction. Nature Human Behaviour (Correspondence).
Peer-Reviewed Workshop Papers
[13] Hoy, W., Fan, J., Celik, N., & Pan, X. (2026). Asynchrony Is Nearly Free for Evolution Strategies on Long-Horizon Agentic Tasks. NeurIPS 2026 Workshop: Pre-to-Post.
[14] Pan, X., Sánchez Giraldo, L. G., Kartal, E., & Schwartz, O. (2021). Brain-Inspired Weighted Normalization for CNN Image Classification. ICLR 2021 Workshop: How Can Findings About The Brain Improve AI Systems?
Preprints and Manuscripts Under Review
[15] Hoy, W., Fan, J., Celik, N., & Pan, X. (2026). Continuous Context Management. arXiv:2609.35540
[16] Wu, S., Zhang, W., Yang, C., Fei, E. X., Overwiening, J., Pan, X., et al. (2026). PersonaScienceBench: Evaluating Scientific Agents with Diverse Simulated Researchers. arXiv:2610.32106.
[17] Geng, H., Huang, Z., Li, H., …, Pan, X., et al. (2026). ScienceIDE: Turning World’s Scientific Codebase into Agent Learnable Environments. arXiv:2609.19134.
[18] Li, X., Hao, Y., Hou, J., …, Pan, X., …, & Song, D. (2026). MatrAIx: Simulating the World with 8.3 Billion Persona Agents. arXiv:2608.04205. Project · Code
Selected media coverage: Numerama · Infobae (August 10, 2026).
Conference Abstracts and Posters
[19] Pan, X., DeForge, A., & Schwartz, O. (2022). Visualizing Surround Suppression in Deep Convolutional Neural Networks. Neuromatch Conference 2022.
[20] Claman, A.*, Pan, X.*, Aguiar-Pulido, V., & Schwartz, O. (2022). Comparing CNNs and the Brain: Sensitivity to Images Altered in the Frequency Domain. Neuromatch Conference 2022.
[21] DeForge, A.*, Pan, X.*, & Schwartz, O. (2022). Brain-Like Visual Surround Suppression in Generic CNNs: Successes and Limitations. Neuromatch Conference 2022.
[22] Pan, X., Coen-Cagli, R., & Schwartz, O. (2021). Modeling Neural Variability in Deep Networks with Dropout. Computational and Systems Neuroscience (COSYNE 2021).