leo

duckking032's activity

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

upvoted a paper 4 days ago

Compile by Training: Turning Natural-Language Specifications into Local Neural Functions

Paper • 2609.04199 • Published 28 days ago • 333

upvoted 3 papers 5 days ago

Why Do Video Diffusion Models Violate Physics? Unveiling the Flaws in Attention Mechanisms

Paper • 2609.23658 • Published 11 days ago • 30

Your Transformer Can Hold Two Thoughts at Once: Evidence of Linear Superposition in LLMs

Paper • 2609.29845 • Published 7 days ago • 99

Training Object Permanence in World Models

Paper • 2609.28654 • Published 8 days ago • 238

upvoted a paper 9 days ago

RRSI: Regularized Recursive Self-Improvement of Agent Harnesses

Paper • 2609.24972 • Published 10 days ago • 219

upvoted a paper 12 days ago

DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression

Paper • 2609.19969 • Published 14 days ago • 191

upvoted 2 papers 21 days ago

Φ-Bench: Can Large Language Models Engineer the Infrastructure That Powers Them?

Paper • 2609.10226 • Published 22 days ago • 19

NeoHorse-1: Towards Recursive Self-Improvement via Agentic Post-Training with Routing Harness

Paper • 2609.08183 • Published 23 days ago • 327

upvoted 2 papers 23 days ago

Unlocking Lossless Speedups in LLMs via Discrete Diffusion

Paper • 2609.04010 • Published 28 days ago • 114

Don't Drop Dropout: Optimizing Layer Sparsity for Efficient LLM Training and Inference

Paper • 2609.05275 • Published 27 days ago • 26

upvoted a paper 24 days ago

MaxKernel: Agentic Kernel Generation for TPUs

Paper • 2609.04523 • Published 28 days ago • 20

upvoted a paper 26 days ago

Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments

Paper • 2609.04148 • Published 28 days ago • 248

upvoted a paper 27 days ago

SMELT: Scaling Laws for Compute-Matched MoE Looped Transformers

Paper • 2609.01343 • Published 30 days ago • 91

upvoted a paper 29 days ago

On the Design of Qwen3.8-Next Architecture: Evaluation, Efficiency, and Training Stability

Paper • 2608.30320 • Published about 1 month ago • 63

upvoted 5 papers about 1 month ago

DSpark: Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generation

Paper • 2607.05147 • Published Jul 6 • 51

Don't Retrain, Align: Adapting Autoregressive LMs to Diffusion LMs via Representation Alignment

Paper • 2605.06885 • Published May 7 • 1

Prime Agent: A Self-Improving RLM Harness

Paper • 2608.23552 • Published Aug 24 • 56

ParaTempo: Efficient Parallel Reasoning via Temporal Confidence

Paper • 2608.16425 • Published Aug 17 • 41

TinyStories: How Small Can Language Models Be and Still Speak Coherent English?

Paper • 2305.07759 • Published May 12, 2023 • 49

upvoted a collection about 1 month ago

DFlash 2

Collection

Keep Drafting Parallel • 6 items • Updated Aug 28 • 48