Profiles2024-2026
Lu Yin
University of Surrey
8Papers
2Last author
12Unique co-authors
1First author
0Solo
First and last author come from byline position, which some venues assign alphabetically.
Recent papers8
- One LR Doesn’t Fit All: Heavy-Tail Guided Layerwise Learning Rates for LLMsAuthor 4 of 5ICML 2026 · 3 citations
- Search or Accelerate: Confidence-Switched Position Beam Search for Diffusion Language ModelsAuthor 5 of 5ICML 2026
- AlphaDecay: Module-wise Weight Decay for Heavy-Tailed Balancing in LLMsAuthor 7 of 7NeurIPS 2025 · 14 citations
- GPAS: Accelerating Convergence of LLM Pretraining via Gradient-Preserving Activation ScalingAuthor 14 of 15NeurIPS 2025 · 2 citations
- Mix-LN: Unleashing the Power of Deeper Layers by Combining Pre-LN and Post-LNAuthor 2 of 3ICLR 2025
- TODO: Enhancing LLM Alignment with Ternary PreferencesAuthor 2 of 4ICLR 2025
- E2ENet: Dynamic Sparse Feature Fusion for Accurate and Efficient 3D Medical Image SegmentationAuthor 4 of 8NeurIPS 2024 · 29 citations
- Junk DNA Hypothesis: Pruning Small Pre-Trained Weights Irreversibly and Monotonically Impairs "Difficult" Downstream Tasks in LLMsAuthor 1 of 5ICML 2024 · 13 citations
