AI Paper Atlas

Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning

Meta Multi-Task Learning for Sequence Modeling

Junkun Chen, Xipeng Qiu, Pengfei Liu, Xuanjing Huang

AAAI 2018 · Label-efficient and robust learning › Zero-/few-shot and meta-learning · approach: Meta-augmentation and meta-objectives · 88 citations (OpenAlex)

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 changedWhereMeasuredEffect
Share a Meta-LSTM that generates task-specific, position-dependent composition parameters instead of using feature-level shared-private layersMulti-task sequence modeling · Classification AVGaccuracy: +2.2% to best competitor Multi-task modelImproved
Use Meta-LSTM-generated dynamic composition parameters instead of a standard LSTMSingle-task text classification · 16 classification datasetsaccuracy: 3.2%Improved
Transfer the trained shared Meta-LSTM and freeze its parameters on a new taskMeta-MTL · Classification AVGaccuracy: 3.1%Improved
Vary the meta-vector size z from 20 to 60 (ablation)Meta-MTL · 16 classification datasetsaverage accuracy: less than 0.8%No clear effect
Use Meta-LSTM-MTL+CRF instead of LSTM-SSP-MTL+CRFSequence tagging · CoNLL 2000F1: 94.32 -> 95.11Improved
Use Meta-LSTM-MTL+CRF instead of LSTM-SSP-MTL+CRFSequence tagging · CoNLL 2003F1: 90.38 -> 90.72Improved
Use Meta-LSTM-MTL+CRF instead of LSTM-SSP-MTL+CRFSequence tagging · WSJaccuracy: 97.23 -> 97.45Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy