Research map / Architectures and efficiency / RNNs and normalization
Fact-Aware Sentence Split and Rephrase with Permutation Invariant Training
Yinuo Guo, Tao Ge, Furu Wei
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 |
|---|---|---|---|
| add Fact-aware Sentence Encoding (multi-task fact classification) (ablation) | LSTM seq2seq · WebSplit-v1.0 | BLEU: 32.5 -> 34.2 | Improved |
| add Permutation Invariant Training (ablation) | LSTM seq2seq · WebSplit-v1.0 | BLEU: 32.5 -> 34.7 | Improved |
| add PIT on top of FaSE (ablation) | LSTM + FaSE · WebSplit-v1.0 | BLEU: 34.2 -> 37.3 | Improved |
| add FaSE on top of PIT (ablation) | LSTM + PIT · WebSplit-v1.0 | BLEU: 34.7 -> 37.3 | Improved |
| choose min-loss permutation (PIT) instead of random permutation (ablation) | LSTM seq2seq · WebSplit-v1.0 | BLEU: 25.9 -> 34.7 | Improved |
| choose max-loss permutation instead of min-loss (ablation) | LSTM seq2seq · WebSplit-v1.0 | BLEU: 34.7 -> 23.9 | Hurt |
| add FaSE (ablation) | Transformer pre-trained on WikiSplit · WebSplit-v1.0 | BLEU: 69.7 -> 70.7 | Improved |
| add PIT (ablation) | Transformer pre-trained on WikiSplit · WebSplit-v1.0 | BLEU: 69.7 -> 70.5 | Improved |
| add FaSE on top of PIT (ablation) | Transformer + PIT · WebSplit-v1.0 | BLEU: 70.5 -> 71.0 | Improved |
| pre-train on WikiSplit with Transformer | FaSE+PIT model · WebSplit-v1.0 | BLEU: 37.3 -> 71.0 | Improved |
Closest papers by meaning
- Narrowing the Gap between Supervised and Unsupervised Sentence Representation Learning with Large Language Model (AAAI 2024 · similarity 0.84)
- Perplexed by Perplexity: Perplexity-Based Data Pruning With Small Reference Models (ICLR 2025 · similarity 0.83)
- Prompt Compression with Context-Aware Sentence Encoding for Fast and Improved LLM Inference (AAAI 2025 · similarity 0.83)
- Dataset Decomposition: Faster LLM Training with Variable Sequence Length Curriculum (NeurIPS 2024 · similarity 0.83)
- BaWA: Automatic Optimizing Pruning Metric for Large Language Models with Balanced Weight and Activation (ICML 2025 · similarity 0.82)
- Fewer Truncations Improve Language Modeling (ICML 2024 · similarity 0.82)
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.
