Defa Zhu

mathfinder's activity

• All
• Models
• Datasets
• Spaces
• Buckets
• Papers
• Collections
• Community
• Posts
• Upvotes
• Likes
• Articles

upvoted a paper 1 day ago

Hierarchical Residuals: A Minimalist Decoupling of Residual Width from Compute Width

Paper • 2610.41711 • Published 12 days ago • 8

upvoted a paper 8 days ago

Your LR Schedule May Not Be Principled: Effective Learning Rate Scheduling in Weight-Normalized Training

Paper • 2610.40362 • Published 13 days ago • 11

upvoted a paper 14 days ago

Rollback: Experience Backtracking in World Models for Policy Improvement

Paper • 2610.36068 • Published 16 days ago • 54

upvoted a paper 20 days ago

Cache Engineering for Agents

Paper • 2610.28577 • Published 22 days ago • 105

upvoted a paper 28 days ago

DoSA: Dots Sparse Attention

Paper • 2610.13990 • Published Sep 26 • 67

upvoted 2 papers about 1 month ago

TokenRouter: Efficient Serving System for Token-Level LLM Routing

Paper • 2610.12242 • Published Oct 8 • 119

Improving Test-Time Scaling with Adaptive Looped Transformers

Paper • 2609.35748 • Published Sep 28 • 64

upvoted a paper 3 months ago

Scaling Domain Data Repetition in LLM Pretraining

Paper • 2608.14071 • Published Aug 14 • 17

upvoted a paper 4 months ago

Seed2.0 Model Card: Towards Intelligence Frontier for Real-World Complexity

Paper • 2607.00248 • Published Jun 30 • 32

upvoted 2 papers 6 months ago

Negligible in Size, Significant in Effect: On Scale Vectors in Large Language Models

Paper • 2605.26895 • Published May 26 • 18

Linear-Time Global Visual Modeling without Explicit Attention

Paper • 2605.01711 • Published May 3 • 7

upvoted 2 papers 10 months ago

ConceptMoE: Adaptive Token-to-Concept Compression for Implicit Compute Allocation

Paper • 2601.21420 • Published Jan 29 • 42

mHC: Manifold-Constrained Hyper-Connections

Paper • 2512.24880 • Published Dec 31, 2025 • 336

authored 4 papers 12 months ago

Expert Race: A Flexible Routing Strategy for Scaling Diffusion Transformer with Mixture of Experts

Paper • 2503.16057 • Published Mar 20, 2025 • 15

Seed1.5-Thinking: Advancing Superb Reasoning Models with Reinforcement Learning

Paper • 2504.13914 • Published Apr 10, 2025 • 6

UltraMemV2: Memory Networks Scaling to 120B Parameters with Superior Long-Context Learning

Paper • 2508.18756 • Published Aug 26, 2025 • 36

Virtual Width Networks

Paper • 2511.11238 • Published Nov 14, 2025 • 39

upvoted a paper 12 months ago

Virtual Width Networks

Paper • 2511.11238 • Published Nov 14, 2025 • 39

upvoted a paper about 1 year ago

UltraMemV2: Memory Networks Scaling to 120B Parameters with Superior Long-Context Learning

Paper • 2508.18756 • Published Aug 26, 2025 • 36

upvoted a paper over 1 year ago

Genie Envisioner: A Unified World Foundation Platform for Robotic Manipulation

Paper • 2508.05635 • Published Aug 7, 2025 • 73