Research map / Large language models / Language, safety and interpretability
Solving Math Word Problems concerning Systems of Equations with GPT-3
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Provide more in-context examples for equation extraction (ablation) | GPT-3 · L30 test set | accuracy: 0.3077 -> 0.6923 | Helped |
| Fine-tune GPT-3 rather than use three-shot prompting | GPT-3 · L20, equation extraction | accuracy: 0.40 -> 0.80 | Helped |
| Increase fine-tuning data from 500 to 1000 examples (ablation) | GPT-3 fine-tuned · L20, equation extraction | accuracy: 0.75 -> 0.80 | Helped |
| Increase fine-tuning data from 100 to 500 examples (ablation) | GPT-3 fine-tuned · L20, equation extraction | accuracy: 0.75 -> 0.75 | No clear effect |
| Include a same-category example in the prompt (ablation) | GPT-3 three-shot equation extraction · L50, Sum and Difference | accuracy: 0.5 -> 0.2 | Hurt |
| Include a same-category example in the prompt (ablation) | GPT-3 three-shot equation extraction · L50, Item and Property | accuracy: 0.7 -> 0.7 | No clear effect |
| Include a same-category example in the prompt (ablation) | GPT-3 three-shot equation extraction · L50, Perimeter of Rectangle | accuracy: 0.2 -> 0.4 | Helped |
| Include a same-category example in the prompt (ablation) | GPT-3 three-shot equation extraction · L50, Motion | accuracy: 0.0 -> 0.4 | Helped |
| Include a same-category example in the prompt (ablation) | GPT-3 three-shot equation extraction · L50, Mixture | accuracy: 0.5 -> 0.7 | Helped |
| Increase in-context examples from four to five (ablation) | GPT-3 equation extraction · L30 test set | accuracy: 0.6923 -> 0.6923 | No clear effect |
| Generate a problem within the same topic | GPT-3 · Within-topic generation, across categories | success rate: 0.33 to 0.93 | Mixed |
| Generate a problem using a different topic within the same category | GPT-3 · Cross-topic generation | success rate: 0/10 to 10/10 | Mixed |
Closest papers by meaning
- Solving Challenging Math Word Problems Using GPT-4 Code Interpreter with Code-based Self-Verification (ICLR 2024 · similarity 0.86)
- Physics of Language Models: Part 2.2, How to Learn From Mistakes on Grade-School Math Problems (ICLR 2025 · similarity 0.85)
- JiuZhang3.0: Efficiently Improving Mathematical Reasoning by Training Small Data Synthesis Models (NeurIPS 2024 · similarity 0.85)
- Advancing Mathematical Reasoning in Language Models: The Impact of Problem-Solving Data, Data Synthesis Methods, and Training Stages (ICLR 2025 · similarity 0.85)
- Physics of Language Models: Part 2.1, Grade-School Math and the Hidden Reasoning Process (ICLR 2025 · similarity 0.84)
- MathCoder2: Better Math Reasoning from Continued Pretraining on Model-translated Mathematical Code (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.