Attention Sink in Transformers: A Survey on Utilization, Interpretation, and Mitigation
Paper • 2604.10098 • Published • 74
AI & ML interests
Fundamental Al Methods; Perception & World Modeling; Reasoning & Generation; Action & Interaction
Recent Activity
View all activity
Papers
SIN-Bench: Tracing Native Evidence Chains in Long-Context Multimodal Scientific Interleaved Literature
ProFit: Leveraging High-Value Signals in SFT via Probability-Guided Token Selection
taki555
authored a
paper 3 days ago
taki555
authored 4
papers about 2 months ago
SwingArena: Competitive Programming Arena for Long-context GitHub Issue Solving
Paper • 2505.23932 • Published
MMFormalizer: Multimodal Autoformalization in the Wild
Paper • 2601.03017 • Published • 106
BPDQ: Bit-Plane Decomposition Quantization on a Variable Grid for Large Language Models
Paper • 2602.04163 • Published • 10
The Art of Efficient Reasoning: Data, Reward, and Optimization
Paper • 2602.20945 • Published • 7
wanng
updated a
dataset 2 months ago
wanng
published a
dataset 2 months ago
21223wj
updated 2
datasets 2 months ago
TianheWu
authored a
paper 2 months ago
TianheWu
submitted a
paper to Daily Papers 2 months ago
21223wj
authored a
paper 3 months ago
wanng
authored 2
papers 3 months ago
wanng
submitted a
paper to Daily Papers 3 months ago
yang31210999
authored 5
papers 3 months ago
ProFit: Leveraging High-Value Signals in SFT via Probability-Guided Token Selection
Paper • 2601.09195 • Published • 15
PTQTP: Post-Training Quantization to Trit-Planes for Large Language Models
Paper • 2509.16989 • Published • 1
Revisiting Model Interpolation for Efficient Reasoning
Paper • 2510.10977 • Published • 10
Timber: Training-free Instruct Model Refining with Base via Effective Rank
Paper • 2509.23595 • Published • 1
Shadow-FT: Tuning Instruct via Base
Paper • 2505.12716 • Published • 4