Research map / Label-efficient and robust learning / Recognition and long-tail
Noether Embedding: Efficient Learning of Temporal Regularities
Chi Gao, Zidong Zhou, Luping Shi
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 |
|---|---|---|---|
| use loss constants Cp=1, Cn=0 (ablation) | NE · ICEWS14 | TR detection F1: 0.78 -> 0.82 | Improved |
| reduce Cp from 1 to 0.6 at Cn=0.4 (ablation) | NE · ICEWS14 | TR detection F1: 0.78 -> 0.29 | Hurt |
| set omega_max to 1 instead of 50 (ablation) | NE · GDELT | TR detection F1: 0.56 -> 0.22 | Hurt |
| increase omega_max from 10 to 800 for query (ablation) | NE · GDELT | TR query r: 0.92 -> 0.85 | Hurt |
| use exponential instead of linear frequency distribution (ablation) | NE · GDELT | TR query r: 0.24 -> 0.85 | Improved |
| use exponential instead of linear frequency distribution (ablation) | NE · ICEWS14 | TR query r: 0.81 -> 0.82 | Improved |
| use exponential instead of linear frequency distribution (ablation) | NE · ICEWS18 | TR query r: 0.75 -> 0.82 | Improved |
| normalize event type vectors (ablation) | NE · ICEWS14 | TR detection F1: 0.80 -> 0.82 | Improved |
| increase embedding dimension d from 50 to 400 (ablation) | NE · ICEWS14 | TR detection F1: 0.29 -> 0.82 | Improved |
| increase embedding dimension d from 400 to 800 (ablation) | NE · ICEWS14 | TR detection F1: 0.82 -> 0.83 | No clear effect |
| decode with direct g(τ) instead of indirect g'(τ) | NE · ICEWS14 | TR detection F1: 0.78 -> 0.82 | Improved |
| use NE instead of TeRo | NE · ICEWS18 | TR detection F1: 0.64 -> 0.83 | Improved |
Closest papers by meaning
- NV-Embed: Improved Techniques for Training LLMs as Generalist Embedding Models (ICLR 2025 · similarity 0.86)
- Human-inspired Episodic Memory for Infinite Context LLMs (ICLR 2025 · similarity 0.85)
- Next-Latent Prediction Transformers Learn Compact World Models (NeurIPS 2026 · similarity 0.85)
- EmbedLLM: Learning Compact Representations of Large Language Models (ICLR 2025 · similarity 0.85)
- Transformer Block Coupling and its Correlation with Generalization in LLMs (ICLR 2025 · similarity 0.85)
- Towards Understanding the Universality of Transformers for Next-Token Prediction (ICLR 2025 · 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.
