phenomenon · first seen 7 Oct 2026
length extrapolation
Model performance on sequences longer than those seen during training.
Mentions per week
PapersCodeDiscussion
Earliest sources
- HySPE: Positional Encoding via Symplectic Dual ShearsPapers · 7 Oct 2026 · Zhongping Ji
- FreeSolo: Linear-time Sparse Attention with Zero-Shot Length ExtrapolationPapers · 12 Oct 2026 · Zekun Shi, Alastair James Arthur Price
- GTFormer: A Syntax-Free Graph Network Alternative to Transformers for 8192$\times$ Length Extrapolation in Linear TimePapers · 27 Oct 2026 · Elisabeth Fink
- RangeSWA: Training-Free Length Extrapolation for Continuous Diffusion Language ModelsPapers · 30 Oct 2026 · Guangfu Guo, Xiaoqian Lu, Linsey Pang, Weiran Yao, Haolin Chen, Long Cheng, Kunp…
- Causal Optimal-Transport Attention: Positional Priors and Length ExtrapolationPapers · 1 Nov 2026 · Ming Lian
- Recursive Sparse Attention: Million-Token Length Extrapolation at Logarithmic CostPapers · 4 Nov 2026 · Ruitian Wang, Xinyu Yang, Minshen Zhang, Hao Zhang
Latest sources
- Recursive Sparse Attention: Million-Token Length Extrapolation at Logarithmic CostPapers · 4 Nov 2026 · Ruitian Wang, Xinyu Yang, Minshen Zhang, Hao Zhang
- Causal Optimal-Transport Attention: Positional Priors and Length ExtrapolationPapers · 1 Nov 2026 · Ming Lian
- RangeSWA: Training-Free Length Extrapolation for Continuous Diffusion Language ModelsPapers · 30 Oct 2026 · Guangfu Guo, Xiaoqian Lu, Linsey Pang, Weiran Yao, Haolin Chen, Long Cheng, Kunp…
- GTFormer: A Syntax-Free Graph Network Alternative to Transformers for 8192$\times$ Length Extrapolation in Linear TimePapers · 27 Oct 2026 · Elisabeth Fink
- FreeSolo: Linear-time Sparse Attention with Zero-Shot Length ExtrapolationPapers · 12 Oct 2026 · Zekun Shi, Alastair James Arthur Price
- HySPE: Positional Encoding via Symplectic Dual ShearsPapers · 7 Oct 2026 · Zhongping Ji