Research map / Large language models / Language, safety and interpretability
Hierarchical Poset Decoding for Compositional Generalization in Language
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 |
|---|---|---|---|
| Replace poset sketch prediction with Seq2Seq-based sketch prediction (ablation) | Hierarchical Poset Decoding · CFQ MCD1 | accuracy: 79.6% -> 74.3% | Hurt |
| Replace poset sketch prediction with Seq2Seq-based sketch prediction (ablation) | Hierarchical Poset Decoding · CFQ MCD2 | accuracy: 59.6% -> 45.7% | Hurt |
| Replace poset sketch prediction with Seq2Seq-based sketch prediction (ablation) | Hierarchical Poset Decoding · CFQ MCD3 | accuracy: 67.8% -> 50.2% | Hurt |
| Replace poset sketch prediction with Seq2Tree-based sketch prediction (ablation) | Hierarchical Poset Decoding · CFQ MCD1 | accuracy: 79.6% -> 75.7% | Hurt |
| Replace poset sketch prediction with Seq2Tree-based sketch prediction (ablation) | Hierarchical Poset Decoding · CFQ MCD2 | accuracy: 59.6% -> 40.9% | Hurt |
| Replace poset sketch prediction with Seq2Tree-based sketch prediction (ablation) | Hierarchical Poset Decoding · CFQ MCD3 | accuracy: 67.8% -> 51.1% | Hurt |
| Add the hierarchical mechanism (ablation) | Basic poset decoding · CFQ MCD1 | accuracy: 21.3% -> 79.6% | Helped |
| Add the hierarchical mechanism (ablation) | Basic poset decoding · CFQ MCD2 | accuracy: 6.4% -> 59.6% | Helped |
| Add the hierarchical mechanism (ablation) | Basic poset decoding · CFQ MCD3 | accuracy: 10.1% -> 67.8% | Helped |
| Simplify SPARQL expressions by merging similar triples | LSTM+Attention · CFQ MCD1 | accuracy: 28.9% -> 42.2% | Helped |
| Simplify SPARQL expressions by merging similar triples | Transformer · CFQ MCD1 | accuracy: 34.9% -> 53.0% | Helped |
Closest papers by meaning
- Consistency of Compositional Generalization Across Multiple Levels (AAAI 2025 · similarity 0.86)
- ConMe: Rethinking Evaluation of Compositional Reasoning for Modern VLMs (NeurIPS 2024 · similarity 0.85)
- Task Generalization with Autoregressive Compositional Structure: Can Learning from $D$ Tasks Generalize to $D^T$ Tasks? (ICML 2025 · similarity 0.84)
- The Consensus Game: Language Model Generation via Equilibrium Search (ICLR 2024 · similarity 0.84)
- The Role of Sparsity for Length Generalization in LLMs (ICML 2025 · similarity 0.84)
- When does compositional structure yield compositional generalization? A kernel theory. (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.