Research map / Large language models / Agents, code and math
Mapping natural-language problems to formal-language solutions using structured neural representations
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 TPR binding in the encoder (ablation) | TP-N2F on MathQA · MathQA test | Operation Accuracy (%): 68.21 -> 71.89 | Helped |
| add TPR unbinding in the decoder (ablation) | TP-N2F on MathQA · MathQA test | Operation Accuracy (%): 68.84 -> 71.89 | Helped |
| add TPR binding in the encoder (ablation) | TP-N2F on AlgoLisp · AlgoLisp full test | Acc (%): 75.31 -> 84.02 | Helped |
| add TPR unbinding in the decoder (ablation) | TP-N2F on AlgoLisp · AlgoLisp full test | Acc (%): 72.28 -> 84.02 | Helped |
| use a parameter-matched LSTM2LSTM+attention baseline (ablation) | TP-N2F comparison on AlgoLisp · AlgoLisp | performance | Hurt |
| increase the Reasoning MLP depth from 1 to 2 or 3 layers (ablation) | TP-N2F · MathQA and AlgoLisp | performance | Hurt |
| use the sum of all token TPRs instead of the last hidden states' tensor product (ablation) | TP-N2F encoder · MathQA and AlgoLisp | performance | Helped |
Closest papers by meaning
- Neural-Symbolic Recursive Machine for Systematic Generalization (ICLR 2024 · similarity 0.86)
- Discrete Dictionary-based Decomposition Layer for Structured Representation Learning (NeurIPS 2024 · similarity 0.86)
- Training Neural Networks as Recognizers of Formal Languages (ICLR 2025 · similarity 0.85)
- Frugal LMs Trained to Invoke Symbolic Solvers Achieve Parameter-Efficient Arithmetic Reasoning (AAAI 2024 · similarity 0.85)
- Mastering Symbolic Operations: Augmenting Language Models with Compiled Neural Networks (ICLR 2024 · similarity 0.85)
- Cross-Dialect Generalization Without Retraining: Benchmarks and Evaluation of Schema-Derived Constrained Decoding for MLIR (NeurIPS 2026 · 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.