Research map / Probabilistic and scientific modeling / Molecules and proteins
Implicit Argument Prediction as Reading Comprehension
Pengxiang Cheng, Katrin Erk
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 PAR instead of EventComp | implicit argument prediction · On-Short | accuracy: 36.90 -> 58.12 | Improved |
| Use PAR instead of EventComp | implicit argument prediction · On-Long | accuracy: 21.26 -> 51.52 | Improved |
| Add entity salience features | EventComp · On-Short | accuracy: 36.90 -> 46.06 | Improved |
| Add entity salience features | EventComp · On-Long | accuracy: 21.26 -> 31.43 | Improved |
| Add entity salience features (ablation) | PAR · OntoNotes | accuracy | No clear effect |
| Add a second attention hop (ablation) | PAR · On-Short-Multi | accuracy: 48.45 -> 50.54 | Improved |
| Add a second attention hop (ablation) | PAR · On-Long-Multi | accuracy: 39.90 -> 42.69 | Improved |
| Add first-hop attention supervision (ablation) | 2-hop PAR · On-Short-Multi | accuracy: 50.54 -> 50.73 | No clear effect |
| Add first-hop attention supervision (ablation) | 2-hop PAR · On-Long-Multi | accuracy: 42.69 -> 41.72 | Hurt |
| Add a second attention hop (ablation) | PAR · G&C | F1: 44.4 -> 46.2 | Improved |
| Add first-hop attention supervision (ablation) | 2-Hop PAR · G&C | F1: 46.2 -> 48.3 | Improved |
| Remove only one argument at a time rather than all potential implicit arguments | PAR on the multi-argument subsets · On-Short-Multi | accuracy: 48.45 -> 51.49 | Improved |
Closest papers by meaning
- How do Transformers Learn Implicit Reasoning? (NeurIPS 2025 · similarity 0.83)
- Finding the Minimal Parameter Budget for Implicit Reasoning: A Data Complexity Driven Scaling Law for Language Models (ICML 2026 · similarity 0.82)
- On the Bias of Next-Token Predictors Toward Systematically Inefficient Reasoning: A Shortest-Path Case Study (NeurIPS 2025 · similarity 0.82)
- Towards Long-Horizon Interpretability: Efficient and Faithful Multi-Token Attribution for Reasoning LLMs (ICML 2026 · similarity 0.82)
- ImpScore: A Learnable Metric For Quantifying The Implicitness Level of Sentences (ICLR 2025 · similarity 0.82)
- On the Inductive Bias of Stacking Towards Improving Reasoning (NeurIPS 2024 · similarity 0.81)
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.
