Jiwon Song

jiwonsong's activity

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

upvoted a paper 1 day ago

Rotation-Aligned Key Channel Pruning for Efficient Vision-Language Model Inference

Paper • 2605.19218 • Published May 19 • 1

upvoted a paper 14 days ago

Loop the Loopies!

Paper • 2607.16051 • Published 20 days ago • 77

upvoted 2 papers 3 months ago

CompactAttention: Accelerating Chunked Prefill with Block-Union KV Selection

Paper • 2605.16839 • Published May 16 • 14

Teaching Language Models to Think in Code

Paper • 2605.07237 • Published May 11 • 31

upvoted 3 papers 5 months ago

Attention Residuals

Paper • 2603.15031 • Published Mar 16 • 193

LK Losses: Direct Acceptance Rate Optimization for Speculative Decoding

Paper • 2602.23881 • Published Feb 27 • 18

dLLM: Simple Diffusion Language Modeling

Paper • 2602.22661 • Published Feb 26 • 154

upvoted 5 papers 6 months ago

Squeezing Large-Scale Diffusion Models for Mobile

Paper • 2307.01193 • Published Jul 3, 2023 • 2

RelayGen: Intra-Generation Model Switching for Efficient Reasoning

Paper • 2602.06454 • Published Feb 6 • 12

L4Q: Parameter Efficient Quantization-Aware Training on Large Language Models via LoRA-wise LSQ

Paper • 2402.04902 • Published Feb 7, 2024 • 5

LRAgent: Efficient KV Cache Sharing for Multi-LoRA LLM Agents

Paper • 2602.01053 • Published Feb 1 • 8

Token Sparse Attention: Efficient Long-Context Inference with Interleaved Token Selection

Paper • 2602.03216 • Published Feb 3 • 14

upvoted 3 papers 9 months ago

Retrospective Sparse Attention for Efficient Long-Context Generation

Paper • 2508.09001 • Published Aug 12, 2025 • 3

Kimi Linear: An Expressive, Efficient Attention Architecture

Paper • 2510.26692 • Published Oct 30, 2025 • 137

Exploring Conditions for Diffusion models in Robotic Control

Paper • 2510.15510 • Published Oct 17, 2025 • 40

upvoted 3 papers 10 months ago

LightMem: Lightweight and Efficient Memory-Augmented Generation

Paper • 2510.18866 • Published Oct 21, 2025 • 116

Every Attention Matters: An Efficient Hybrid Architecture for Long-Context Reasoning

Paper • 2510.19338 • Published Oct 22, 2025 • 116

LiteStage: Latency-aware Layer Skipping for Multi-stage Reasoning

Paper • 2510.14211 • Published Oct 16, 2025 • 9

upvoted a collection 10 months ago

Speculator Models

Collection

29 items • Updated 13 days ago • 24

upvoted a paper 11 months ago

QWHA: Quantization-Aware Walsh-Hadamard Adaptation for Parameter-Efficient Fine-Tuning on Large Language Models

Paper • 2509.17428 • Published Sep 22, 2025 • 9