Research map / Large language models / Agents, code and math
Generalizing Math Word Problem Solvers via Solution Diversification
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 |
|---|---|---|---|
| Add solution diversity through a buffer and discriminator | GTS · D-Math23k, fully supervised | answer accuracy (%): 65.7 -> 67.1 | Helped |
| Add solution diversity through a buffer and discriminator | Graph2Tree · D-Math23k, fully supervised | answer accuracy (%): 66.6 -> 68.9 | Helped |
| Add solution diversity through a buffer and discriminator | MWP-BERT · D-Math23k, fully supervised | answer accuracy (%): 69.2 -> 73.3 | Helped |
| Initialize semi-weak buffers with WDA rather than leave them empty (ablation) | D-GTS · D-Math23k, semi-weakly supervised | answer accuracy (%): 68.4 -> 67.1 | Hurt |
| Initialize semi-weak buffers with WDA rather than leave them empty (ablation) | D-GTS · Weak12k, semi-weakly supervised | answer accuracy (%): 59.2 -> 57.9 | Hurt |
| Initialize semi-weak buffers with WDA rather than leave them empty (ablation) | D-MWP-BERT · D-Math23k, semi-weakly supervised | answer accuracy (%): 74.4 -> 73.2 | Hurt |
| Initialize semi-weak buffers with WDA rather than leave them empty (ablation) | D-MWP-BERT · Weak12k, semi-weakly supervised | answer accuracy (%): 70.9 -> 67.5 | Hurt |
| Use WDA to initialize solution buffers | D-GTS · D-Math23k, weakly supervised | answer accuracy (%): +2.1 | Helped |
| Use WDA to initialize solution buffers | D-GTS · Weak12k, weakly supervised | answer accuracy (%): +5.5 | Helped |
| Use WDA to initialize solution buffers | D-MWP-BERT · D-Math23k, weakly supervised | answer accuracy (%): +2.9 | Helped |
| Use WDA to initialize solution buffers | D-MWP-BERT · Weak12k, weakly supervised | answer accuracy (%): +2.3 | Helped |
| Add solution diversity through a buffer and discriminator | GTS · D-Math23k, fully supervised | top-5 answer accuracy (%): 29.0 -> 55.9 | Helped |
Closest papers by meaning
- Template-Driven LLM-Paraphrased Framework for Tabular Math Word Problem Generation (AAAI 2025 · similarity 0.86)
- Token-Supervised Value Models for Enhancing Mathematical Problem-Solving Capabilities of Large Language Models (ICLR 2025 · similarity 0.86)
- MAmmoTH: Building Math Generalist Models through Hybrid Instruction Tuning (ICLR 2024 · similarity 0.85)
- A Compact Model for Mathematics Problem Representations Distilled from BERT (AAAI 2025 · similarity 0.85)
- Augmenting Math Word Problems via Iterative Question Composing (AAAI 2025 · similarity 0.85)
- Advancing Mathematical Reasoning in Language Models: The Impact of Problem-Solving Data, Data Synthesis Methods, and Training Stages (ICLR 2025 · similarity 0.84)
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.