Research map / Large language models / Agents, code and math
ELASTIC: Numerical Reasoning with Adaptive Symbolic Compiler
Jiaxin Zhang, Yashar Moshfeghi
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 Memory Register for previous sub-program results (ablation) | ELASTIC (RoBERTa-large) · FinQA | program accuracy: 64.78 -> 65.21 | Improved |
| add Memory Register (ablation) | ELASTIC (RoBERTa-large) · MathQA | program accuracy: 82.68 -> 83.00 | Improved |
| add Memory Register for long memory departing distance (M-MDD > 5) (ablation) | ELASTIC · MathQA | program accuracy | Improved |
| separate operator and operand generation (ELASTIC vs FinQANet) with RoBERTa-large | ELASTIC · FinQA test | execution accuracy: 65.05 -> 68.96 | Improved |
| separate operator and operand generation vs FinQANet with RoBERTa-large | ELASTIC · MathQA test | program accuracy: 79.20 -> 83.00 | Improved |
| use RoBERTa-large instead of RoBERTa-base encoder (ablation) | ELASTIC · MathQA test | program accuracy: 82.27 -> 83.00 | Improved |
| use ELASTIC rather than FinQANet on programs with 3 or more steps | ELASTIC · FinQA >=3 steps | program accuracy: 31.10 vs 31.65 (original FinQANet), 23.80 (retrained) | Mixed |
Closest papers by meaning
- Scalable Chain of Thoughts via Elastic Reasoning (ICLR 2026 · similarity 0.86)
- ORACLE: Optimizing Reasoning Abilities of Large Language Models via Constraint-Led Synthetic Data Elicitation (AAAI 2026 · similarity 0.84)
- CRANE: Reasoning with constrained LLM generation (ICML 2025 · similarity 0.84)
- HERMES: Towards Efficient and Verifiable Mathematical Reasoning in LLMs (ICML 2026 · similarity 0.83)
- Exploring Equation as a Better Intermediate Meaning Representation for Numerical Reasoning of Large Language Models (AAAI 2024 · similarity 0.83)
- ARM: Adaptive Reasoning Model (NeurIPS 2025 · similarity 0.83)
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.
