Research map / Vision-language and multimodal / Human and embodied modeling
Hierarchical Encoder with Auxiliary Supervision for Neural Table-to-Text Generation: Learning Better Representation for Tables
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 two-level LSTM encoder with coarse-to-fine attention (ablation) | seq2seq · WIKIBIO | BLEU-4: 43.37 -> 44.42 | Helped |
| use two-level LSTM instead of stacked/bi-directional LSTM with more parameters (ablation) | seq2seq encoder · WIKIBIO | BLEU-4: 44.14 vs 43.25 | Helped |
| replace word-attribute concatenation with auxiliary sequence labeling (ablation) | hs2s · WIKIBIO | BLEU-4: 44.42 -> 44.63 | Helped |
| replace field-gating with sequence labeling (MLP) (ablation) | hs2s · WIKIBIO | BLEU-4: 44.57 -> 44.63 | Helped |
| use CRF instead of MLP as label function (ablation) | hs2s + SL · WIKIBIO | BLEU-4: 44.63 vs 44.54 | No clear effect |
| add text auto-encoder supervision (Eq 7 & 8) (ablation) | seq2seq · WIKIBIO | BLEU-4: 43.37 -> 44.04 | Helped |
| add text auto-encoder supervision (Eq 7 & 8) (ablation) | hs2s + SL-CRF · WIKIBIO | BLEU-4: 44.63 -> 44.84 | Helped |
| add distance minimizing on top of co-training (ablation) | hs2s + concat · WIKIBIO | BLEU-4: 44.64 -> 44.68 | Helped |
| add multi-labeling supervision (Eq 10) (ablation) | seq2seq · WIKIBIO | BLEU-4: 43.37 -> 43.69 | Helped |
| add multi-labeling supervision (Eq 10) (ablation) | hs2s + SL-CRF · WIKIBIO | BLEU-4: 44.63 -> 45.01 | Helped |
| add auto-encoder supervision gives more gain to weaker encoder (ablation) | seq2seq vs hierarchical · WIKIBIO | BLEU-4: +0.67 vs +0.26 | Mixed |
| shuffle attribute order in train and test | s2s · WIKIBIO shuffled | BLEU-4: 43.37 -> 41.80 | Hurt |
Closest papers by meaning
- Hybrid Autoencoders for Tabular Data: Leveraging Model-Based Augmentation in Low-Label Settings (NeurIPS 2025 · similarity 0.84)
- UniTabE: A Universal Pretraining Protocol for Tabular Foundation Model in Data Science (ICLR 2024 · similarity 0.84)
- Making Pre-trained Language Models Great on Tabular Prediction (ICLR 2024 · similarity 0.84)
- HeGTa: Leveraging Heterogeneous Graph-enhanced Large Language Models for Few-shot Complex Table Understanding (AAAI 2025 · similarity 0.84)
- TRivia: Self-supervised Fine-tuning of Vision-Language Models for Table Recognition (CVPR 2026 · similarity 0.84)
- TRL-Bench: Standardizing Cross-Paradigm Representation-Level Evaluation of Tabular Encoders (NeurIPS 2026 · 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.