Research map / Large language models / Agents, code and math
Neural Module Networks for Reasoning over Text
Nitish Gupta, Kevin Lin, Dan Roth, Sameer Singh, Matt Gardner
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 |
|---|---|---|---|
| Use the NMN with BERT representations instead of MTMSN | DROP (pruned) · DROP (pruned) | F1: 76.5 -> 77.4 | Improved |
| Use the NMN with BERT representations instead of MTMSN | DROP (pruned), Date-Compare · Date-Compare | F1: 85.2 -> 82.6 | Hurt |
| Use the NMN with BERT representations instead of MTMSN | DROP (pruned), Date-Difference · Date-Difference | F1: 72.5 -> 75.4 | Improved |
| Use the NMN with BERT representations instead of MTMSN | DROP (pruned), Number-Compare · Number-Compare | F1: 85.1 -> 92.7 | Improved |
| Use the NMN with BERT representations instead of MTMSN | DROP (pruned), Extract-Number · Extract-Number | F1: 80.7 -> 86.1 | Improved |
| Use the NMN with BERT representations instead of MTMSN | DROP (pruned), Count · Count | F1: 61.6 -> 55.7 | Hurt |
| Use the NMN with BERT representations instead of MTMSN | DROP (pruned), Extract-Argument · Extract-Argument | F1: 66.6 -> 69.7 | Improved |
| Remove the unsupervised auxiliary objective (ablation) | NMN with GRU · DROP (pruned) | F1: 73.1 -> 57.3 | Hurt |
| Remove the unsupervised auxiliary objective (ablation) | NMN with BERT · DROP (pruned) | F1 | Hurt |
| Add intermediate module output supervision (ablation) | NMN · DROP (pruned) | F1 | Improved |
| Add an addition and subtraction module (ablation) | NMN · DROP (full validation) | F1: 65.4 vs. 72.8 | Hurt |
Closest papers by meaning
- Compositional Reasoning with Transformers, RNNs, and Chain of Thought (NeurIPS 2025 · similarity 0.87)
- Neural Reasoning Networks: Efficient Interpretable Neural Networks with Automatic Textual Explanations (AAAI 2025 · similarity 0.86)
- Mastering Symbolic Operations: Augmenting Language Models with Compiled Neural Networks (ICLR 2024 · similarity 0.86)
- Detection-Based Intermediate Supervision for Visual Question Answering (AAAI 2024 · similarity 0.85)
- GENOME: Generative Neuro-Symbolic Visual Reasoning by Growing and Reusing Modules (ICLR 2024 · similarity 0.85)
- CodePlan: Unlocking Reasoning Potential in Large Language Models by Scaling Code-form Planning (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.
