Research map / Learning theory and algorithms / Matrix and tensor methods
Group Representational Position Encoding
Yifan Zhang, Zixiang Chen, Yifeng Liu, Zhen Qin, Huizhuo Yuan, Kangping Xu, Yang Yuan, Quanquan Gu, Andrew C Yao
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace RoPE with GRAPE-AP positional encoding | Llama 353M · FineWeb-Edu 0-shot avg | average accuracy: 51.73 -> 53.25 | Improved |
| replace RoPE with GRAPE-AP positional encoding | Llama 770M · FineWeb-Edu 0-shot avg | average accuracy: 55.76 -> 56.91 | Improved |
| use multiplicative GRAPE-M (ctx or non-ctx) instead of RoPE | Llama 353M · FineWeb-Edu 0-shot avg | average accuracy: 51.73 -> 51.78 / 51.79 | No clear effect |
| use multiplicative GRAPE-M instead of RoPE | Llama 770M · FineWeb-Edu 0-shot avg | average accuracy: 55.76 -> 54.73 / 54.81 | Hurt |
| add GRAPE-AP vs FoX with KV-shift | Llama 770M · FineWeb-Edu 0-shot avg (with KV-shift) | average accuracy: 57.09 -> 56.86 | Hurt |
| GRAPE-AP in place of RoPE with KV-shift | Llama 353M · FineWeb-Edu 0-shot avg (with KV-shift) | average accuracy: 52.89 -> 53.46 | Improved |
| gate additive slope on both query and key (A-QK) rather than key only (A-K) (ablation) | GRAPE-A · FineWeb-Edu 0-shot avg | average accuracy: 52.43 -> 53.00 | Improved |
Closest papers by meaning
- Equivariance by Contrast: Identifiable Equivariant Embeddings from Unlabeled Finite Group Actions (NeurIPS 2025 · similarity 0.86)
- Algebraic Positional Encodings (NeurIPS 2024 · similarity 0.86)
- ComRoPE: Scalable and Robust Rotary Position Embedding Parameterized by Trainable Commuting Angle Matrices (CVPR 2025 · similarity 0.85)
- Learning Efficient Positional Encodings with Graph Neural Networks (ICLR 2025 · similarity 0.85)
- A Circular Argument: Does RoPE need to be Equivariant for Vision? (NeurIPS 2025 · similarity 0.85)
- On the Stability of Expressive Positional Encodings for Graphs (ICLR 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
