Research map / Label-efficient and robust learning / Recognition and long-tail
Understanding Robust Generalization in Learning Regular Languages
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 state sequence auxiliary supervision | LSTM RNN classifier · L_mod-3, L_mod-4, and L_mod-5 OOD tests | accuracy: 66.10 -> 98.68; 64.61 -> 97.58; 65.15 -> 93.00 | Helped |
| Add zero-count auxiliary supervision (ablation) | E2E RNN classifier · Parity OOD test | accuracy: 5-8% | Helped |
| Increase the number of training examples (ablation) | E2E RNN classifier · Parity OOD test | accuracy: stabilizing around 66% at 2800 training examples | No clear effect |
| Use LSTM or GRU cells instead of RNN cells (ablation) | E2E RNN classifier · Parity OOD test | accuracy | Helped |
| Train on longer sequences rather than shorter sequences (ablation) | Parity classifier · Shifted test distributions P1, P2, and Q | accuracy | Helped |
| Add state sequence auxiliary supervision | LSTM RNN classifier · Parity OOD test, shifts 0.0, 0.4, 0.8, and 1.2 | Brier Score: 5e-06 -> 3e-06; 9e-04 -> 1e-04; 0.073 -> 0.008; 0.262 -> 0.051 | Helped |
| Add state sequence auxiliary supervision | LSTM RNN classifier · Parity OOD test, shifts 0.0, 0.4, 0.8, and 1.2 | ECE: 4e-04 -> 3e-04; 0.003 -> 0.002; 0.097 -> 0.035; 0.303 -> 0.125 | Helped |
Closest papers by meaning
- Task Generalization with Autoregressive Compositional Structure: Can Learning from $D$ Tasks Generalize to $D^T$ Tasks? (ICML 2025 · similarity 0.89)
- When does compositional structure yield compositional generalization? A kernel theory. (ICLR 2025 · similarity 0.87)
- A transfer learning framework for weak to strong generalization (ICLR 2025 · similarity 0.87)
- Understanding and Improving Length Generalization in Recurrent Models (ICML 2025 · similarity 0.87)
- Scaling can lead to compositional generalization (NeurIPS 2025 · similarity 0.86)
- Quantifying the Gain in Weak-to-Strong Generalization (NeurIPS 2024 · similarity 0.86)
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.