Research map / Large language models / Language, safety and interpretability
GeoNum: Bridging Numerical Continuity and Language Semantics via Geometric Embedding
Shengkai Jin, Tianyu Chen, Chonghan Gao, Jun Han
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 |
|---|---|---|---|
| add joint discrete-continuous training (ablation) | GeoNum · NumericBench | ACC@0.1: 0.307 -> 0.680 | Improved |
| add polar encoding (ablation) | GeoNum · NumericBench | ACC@0.1: 0.538 -> 0.680 | Improved |
| add trigonometric fractional encoding (ablation) | GeoNum · NumericBench | ACC@0.1: 0.527 -> 0.680 | Improved |
| add positional and enhanced digit encodings (ablation) | GeoNum · NumericBench | ACC@0.1 | No clear effect |
| use GeoNum instead of fine-tuning | LLaMA-3.2-3B · NUPA Float-ADD | ACC@0.1: 0.35 -> 0.52 | Improved |
| use GeoNum instead of fine-tuning | LLaMA-3.2-3B · FERMAT | ACC@0.1: 0.39 -> 0.45 | Improved |
| use GeoNum instead of fine-tuning at coarse threshold | LLaMA-3.2-1B · FERMAT | ACC@5: 0.69 -> 0.66 | Hurt |
| use GeoNum for extrapolation beyond training range | GeoNum · Range generalization [-5K, 5K] | accuracy: 0.164 (FoNE) -> 0.213 | Improved |
| scale LLaMA-3.2 from 1B to 3B | GeoNum · NumericBench | MAE: 0.79 -> 0.63 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FoNE — FoNE: Precise Single-Token Number Embeddings via Fourier Features (ICLR 2026) · better on 8 of 8 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Do Large Language Models Truly Understand Geometric Structures? (ICLR 2025 · similarity 0.87)
- Geometric Latent Reasoning Induces Shorter Generations in LLMs (NeurIPS 2026 · similarity 0.86)
- Geoint-R1: Formalizing Multimodal Geometric Reasoning with Dynamic Auxiliary Constructions (CVPR 2026 · similarity 0.85)
- GeoX: Geometric Problem Solving Through Unified Formalized Vision-Language Pre-training (ICLR 2025 · similarity 0.85)
- Hilbert-Geo: Solving Solid Geometric Problems by Neural-Symbolic Reasoning (CVPR 2026 · similarity 0.85)
- G-LLaVA: Solving Geometric Problem with Multi-Modal Large Language Model (ICLR 2025 · similarity 0.85)
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.
