DepthBench: Measuring How Residual Connections Enable More Computational Depth Paper • 2609.32534 • Published 3 days ago • 20
FlashLoop: Fast and Memory-Efficient Looped Transformers via Lazy Updates Paper • 2609.29812 • Published 5 days ago • 1
AlphaQ: Calibration-Free Bit Allocation for Mixture-of-Experts Quantization Paper • 2606.04980 • Published Jun 3 • 2