Shiwei Liu

Shiweiliuiiiiiii's activity

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

upvoted a paper 6 days ago

Scaling to Tens of Thousands of Test-Time Iterations with Loop-Native Attention Residuals

Paper • 2610.11570 • Published 8 days ago • 9

upvoted a paper 10 days ago

Looping Beyond Twice: A Scalable Recipe for Looped Mixture-of-Experts

Paper • 2610.01153 • Published 15 days ago • 24

upvoted a paper 14 days ago

FlashLoop: Fast and Memory-Efficient Looped Transformers via Lazy Updates

Paper • 2609.29812 • Published 22 days ago • 3

upvoted a paper 16 days ago

DepthBench: Measuring How Residual Connections Enable More Computational Depth

Paper • 2609.32534 • Published 20 days ago • 32

upvoted 4 papers 4 months ago

Why Diffusion Language Models Struggle with Truly Parallel (Non-Autoregressive) Decoding?

Paper • 2602.23225 • Published Feb 26 • 1

One LR Doesn't Fit All: Heavy-Tail Guided Layerwise Learning Rates for LLMs

Paper • 2605.22297 • Published May 27 • 1

AlphaQ: Calibration-Free Bit Allocation for Mixture-of-Experts Quantization

Paper • 2606.04980 • Published Jun 3 • 2

Learning from the Self-future: On-policy Self-distillation for dLLMs

Paper • 2606.18195 • Published Jun 16 • 82

upvoted a paper 11 months ago

The Path Not Taken: RLVR Provably Learns Off the Principals

Paper • 2511.08567 • Published Nov 11, 2025 • 37

upvoted a paper 12 months ago

The Art of Scaling Reinforcement Learning Compute for LLMs

Paper • 2510.13786 • Published Oct 15, 2025 • 34

upvoted a paper about 1 year ago

Diffusion Language Models Know the Answer Before Decoding

Paper • 2508.19982 • Published Aug 27, 2025 • 27

upvoted 7 papers over 1 year ago

Spatial Mental Modeling from Limited Views

Paper • 2506.21458 • Published Jun 26, 2025 • 14

LIFT the Veil for the Truth: Principal Weights Emerge after Rank Reduction for Reasoning-Focused Supervised Fine-Tuning

Paper • 2506.00772 • Published Jun 1, 2025 • 2

SoS1: O1 and R1-Like Reasoning LLMs are Sum-of-Square Solvers

Paper • 2502.20545 • Published Feb 27, 2025 • 22

Stable-SPAM: How to Train in 4-Bit More Stably than 16-Bit Adam

Paper • 2502.17055 • Published Feb 24, 2025 • 20

Mask-Enhanced Autoregressive Prediction: Pay Less Attention to Learn More

Paper • 2502.07490 • Published Feb 11, 2025 • 12

The Curse of Depth in Large Language Models

Paper • 2502.05795 • Published Feb 9, 2025 • 41

O1-Pruner: Length-Harmonizing Fine-Tuning for O1-Like Reasoning Pruning

Paper • 2501.12570 • Published Jan 22, 2025 • 28

upvoted 2 papers almost 2 years ago

SPAM: Spike-Aware Adam with Momentum Reset for Stable LLM Training

Paper • 2501.06842 • Published Jan 12, 2025 • 16

Mix-LN: Unleashing the Power of Deeper Layers by Combining Pre-LN and Post-LN

Paper • 2412.13795 • Published Dec 18, 2024 • 20