Unverified50% confidenceFactExact time
Current mainstream long-context solutions generally use RoPE combined with extrapolation techniques such as YaRN and LongRoPE.
1
Sources
50%
Confidence
Long-term
Relevance
10/3/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedRoPE encodes relative positions via rotational transformations in complex space, and can extend the effective context length in combination with extrapolation techniques such as YaRN and LongRoPE.77% similarUnverifiedYaRN (Yet another RoPE extensioN) is an extension method for Rotary Position Embedding (RoPE), proposed by Bowen Peng et al. in 2023, which applies differentiated scaling strategies to high-frequency, mid-frequency, and low-frequency components respectively, enabling models to extend their effective context length several times over without retraining or with only minimal fine-tuning.70% similarUnverifiedRoPE (Rotary Position Embedding) enhances long-range dependency capture through relative position representation and has been widely adopted by LLaMA, DeepSeek, and others.62% similarUnverifiedRoPE（旋转位置编码）已成为当前主流LLM（包括Llama、Mistral、Qwen等）的标准选择，兼具实现简洁性和长度外推优势59% similarUnverifiedRoPE rotary positional encoding has been widely adopted by open-source models such as LLaMA and Qwen due to its favorable length extrapolation properties.58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/969980API
curl https://kongchang.com/api/v1/knowledge/claims/969980MCP
get_claim(id=969980)