Image de couverture de livre

Frayed RoPE and Long Inputs: A Geometric Perspective

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Wertheimer, Davis, Zhang, Aozhong, Liu, Derrick, Yin, Penghang, Wang, Naigang
Format:	Preprint
Publié:	2026
Sujets:	Machine Learning Computation and Language
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

CoPE: Clipped RoPE as A Scalable Free Lunch for Long Context LLMs
par: Li, Haoran, et autres
Publié: (2026)

Demystifying the Slash Pattern in Attention: The Role of RoPE
par: Cheng, Yuan, et autres
Publié: (2026)

RoPE Distinguishes Neither Positions Nor Tokens in Long Contexts, Provably
par: Du, Yufeng, et autres
Publié: (2026)

DiaBlo: Diagonal Blocks Are Sufficient For Finetuning
par: Gurses, Selcuk, et autres
Publié: (2025)

Rotary Positional Embeddings as Phase Modulation: Theoretical Bounds on the RoPE Base for Long-Context Transformers
par: Liu, Feilong
Publié: (2026)

Fractional Rotation, Full Potential? Investigating Performance and Convergence of Partial RoPE
par: Khan, Mohammad Aflah, et autres
Publié: (2026)

RoPE Attention Can Be Trained in Almost Linear Time
par: Cao, Yang, et autres
Publié: (2024)

Circuit Complexity Bounds for RoPE-based Transformer Architecture
par: Chen, Bo, et autres
Publié: (2024)

Jordan-RoPE: Non-Semisimple Relative Positional Encoding via Complex Jordan Blocks
par: Zhang, Yaobo
Publié: (2026)

Understanding the RoPE Extensions of Long-Context LLMs: An Attention Perspective
par: Zhong, Meizhi, et autres
Publié: (2024)

COMQ: A Backpropagation-Free Algorithm for Post-Training Quantization
par: Zhang, Aozhong, et autres
Publié: (2024)

Theoretical Constraints on the Expressive Power of $\mathsf{RoPE}$-based Tensor Attention Transformers
par: Li, Xiaoyu, et autres
Publié: (2024)

MagR: Weight Magnitude Reduction for Enhancing Post-Training Quantization
par: Zhang, Aozhong, et autres
Publié: (2024)

CLoQ: Enhancing Fine-Tuning of Quantized LLMs via Calibrated LoRA Initialization
par: Deng, Yanxia, et autres
Publié: (2025)

EliteKV: Scalable KV Cache Compression via RoPE Frequency Selection and Joint Low-Rank Projection
par: Zhou, Yuhao, et autres
Publié: (2025)

RoPE-LIME: RoPE-Space Locality + Sparse-K Sampling for Efficient LLM Attribution
par: Picov, Isaac, et autres
Publié: (2026)

Base of RoPE Bounds Context Length
par: Men, Xin, et autres
Publié: (2024)

Scaling Laws of RoPE-based Extrapolation
par: Liu, Xiaoran, et autres
Publié: (2023)

Periodic RoPE for Infinite Context LLMs
par: Huo, Simin
Publié: (2026)

Shuffle the Context: RoPE-Perturbed Self-Distillation for Long-Context Adaptation
par: Li, Zichong, et autres
Publié: (2026)

RoPeSLR: 3D RoPE-driven Sparse-LowRank Attention for Efficient Diffusion Transformers
par: Liu, Yuxi, et autres
Publié: (2026)

LinearARD: Linear-Memory Attention Distillation for RoPE Restoration
par: Yang, Ning, et autres
Publié: (2026)

When Precision Meets Position: BFloat16 Breaks Down RoPE in Long-Context Training
par: Wang, Haonan, et autres
Publié: (2024)

RoPECraft: Training-Free Motion Transfer with Trajectory-Guided RoPE Optimization on Diffusion Transformers
par: Gokmen, Ahmet Berke, et autres
Publié: (2025)

Resonance RoPE: Improving Context Length Generalization of Large Language Models
par: Wang, Suyuchen, et autres
Publié: (2024)

Rethinking RoPE: A Mathematical Blueprint for N-dimensional Positional Embedding
par: Liu, Haiping, et autres
Publié: (2025)

On the token distance modeling ability of higher RoPE attention dimension
par: Hong, Xiangyu, et autres
Publié: (2024)

RAP: KV-Cache Compression via RoPE-Aligned Pruning
par: Xin, Jihao, et autres
Publié: (2026)

Q-ROAR: Outlier-Aware Rescaling for RoPE Position Interpolation in Quantized Long-Context LLMs
par: Qiao, Ye, et autres
Publié: (2025)

RoBERTurk: Adjusting RoBERTa for Turkish
par: Tas, Nuri
Publié: (2024)

Fast RoPE Attention: Combining the Polynomial Method and Fast Fourier Transform
par: Alman, Josh, et autres
Publié: (2025)

RoSA: Enhancing Parameter-Efficient Fine-Tuning via RoPE-aware Selective Adaptation in Large Language Models
par: Pan, Dayan, et autres
Publié: (2025)

Mitigating the Impact of Outlier Channels for Language Model Quantization with Activation Regularization
par: Nrusimha, Aniruddha, et autres
Publié: (2024)

ID-Align: RoPE-Conscious Position Remapping for Dynamic High-Resolution Adaptation in Vision-Language Models
par: Li, Bozhou, et autres
Publié: (2025)

Rethinking RoPE Scaling in Quantized LLM: Theory, Outlier, and Channel-Band Analysis with Weight Rescaling
par: Qiao, Ye, et autres
Publié: (2025)

Does RoBERTa Perform Better than BERT in Continual Learning: An Attention Sink Perspective
par: Bai, Xueying, et autres
Publié: (2024)

Positional versus Symbolic Attention Heads: Learning Dynamics, RoPE Geometry, and Length Generalization
par: Urrutia, Felipe, et autres
Publié: (2026)

ReRoPE: Repurposing RoPE for Relative Camera Control
par: Li, Chunyang, et autres
Publié: (2026)

HoPE: A Novel Positional Encoding Without Long-Term Decay for Enhanced Context Awareness and Extrapolation
par: Chen, Yuhan, et autres
Publié: (2024)

A Decomposition Perspective to Long-context Reasoning for LLMs
par: Xiao, Yanling, et autres
Publié: (2026)