Lune

Profiles2020-2026

Di He

Peking University

24Papers
7Last author
34Unique co-authors
0First author
0Solo

First and last author come from byline position, which some venues assign alphabetically.

Recent papers24

  1. Efficient Reasoning for Large Reasoning Language Models via Certainty-Guided Reflection SuppressionAuthor 5 of 6AAAI 2026 · 21 citations
  2. In-Place Test-Time TrainingAuthor 5 of 7ICLR 2026 · 15 citations
  3. Diagnosing and Improving Diffusion Models by Estimating the Optimal Loss ValueAuthor 4 of 5ICLR 2026 · 6 citations
  4. From Five Dimensions to Many: Large Language Models as Precise and Interpretable Psychological ProfilersAuthor 3 of 4ICLR 2026 · 6 citations
  5. Quotient-Space Diffusion ModelsAuthor 6 of 7ICLR 2026 · 1 citation
  6. SWE-Swiss: A Multi-Task Fine-Tuning and RL Recipe for High-Performance Issue ResolutionAuthor 9 of 11ACL Findings 2026
  7. Ted-Tok: Maintaining an Evolving Vocabulary for Lifelong LearningAuthor 5 of 5ACL 2026
  8. Theoretical Benefit and Limitation of Diffusion Language ModelAuthor 6 of 6NeurIPS 2025 · 52 citations
  9. How Numerical Precision Affects Arithmetical Reasoning Capabilities of LLMsAuthor 7 of 9ACL Findings 2025 · 5 citations
  10. Let the Code LLM Edit Itself When You Edit the CodeAuthor 6 of 6ICLR 2025
  11. Beyond Weisfeiler-Lehman: A Quantitative Framework for GNN ExpressivenessAuthor 5 of 6ICLR 2024 · 59 citations
  12. Do Efficient Transformers Really Save Computation?Author 8 of 9ICML 2024 · 29 citations
  13. Two Stones Hit One Bird: Bilevel Positional Encoding for Better Length ExtrapolationAuthor 9 of 9ICML 2024 · 25 citations
  14. GeoMFormer: A General Architecture for Geometric Molecular Representation LearningAuthor 3 of 6ICML 2024 · 9 citations
  15. Bridging Geometric States via Geometric Diffusion BridgeAuthor 3 of 6NeurIPS 2024 · 3 citations
  16. Towards Revealing the Mystery behind Chain of Thought: A Theoretical PerspectiveAuthor 5 of 6NeurIPS 2023 · 470 citations
  17. Robustness-Aware Word Embedding Improves Certified Robustness to Adversarial Word SubstitutionsAuthor 3 of 4ACL Findings 2023 · 8 citations
  18. Denoising Masked Autoencoders Help Robust ClassificationAuthor 6 of 6ICLR 2023 · 5 citations
  19. One Transformer Can Understand Both 2D & 3D Molecular DataAuthor 7 of 7ICLR 2023 · 4 citations
  20. Online Training Through Time for Spiking Neural NetworksAuthor 4 of 5NeurIPS 2022 · 121 citations
  21. Rethinking Lipschitz Neural Networks and Certified Robustness: A Boolean Function PerspectiveAuthor 3 of 4NeurIPS 2022 · 88 citations
  22. Your Transformer May Not be as Powerful as You ExpectAuthor 6 of 6NeurIPS 2022 · 69 citations
  23. Is L2L^2 Physics Informed Loss Always Suitable for Training Physics Informed Neural Network?Author 3 of 4NeurIPS 2022 · 36 citations
  24. MACER: Attack-free and Scalable Robust Training via Maximizing Certified RadiusAuthor 3 of 8ICLR 2020 · 195 citations
Dusk over the sea between two cliffs drawn in fine vertical lines