Research map / Graph learning / GNN architectures
Learning Multi-Task Communication with Message Passing for Sequence Learning
Pengfei Liu, Jie Fu, Yue Dong, Xipeng Qiu, Jackie Chi Kit Cheung
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 learned direct inter-task message passing (DC-MTL) over shared-private sharing | SP-MTL · 16 text classification tasks | error rate (%): 13.9 -> 12.5 | Improved |
| replace shared-private layers with mailbox virtual node and selective retrieval (IC-MTL) | SP-MTL · 16 text classification tasks | error rate (%): 13.9 -> 12.7 | Improved |
| learn multi-task communication with message passing instead of fully shared LSTM | FS-MTL · 16 text classification tasks | error rate (%): 15.3 -> 12.5 | Improved |
| transfer frozen shared layer from IC-MTL with attention retrieval instead of SP-MTL shared layer | SP-MTL · 16 text tasks leave-one-task-out transfer | error rate (%): 14.3 -> 13.1 | Improved |
| multi-task transfer (IC-MTL) vs single-task training | single-task LSTMs · 16 text tasks leave-one-task-out transfer | error rate (%): 18.0 -> 13.1 | Improved |
| use IC-MTL instead of best baseline FS-MTL | FS-MTL · CoNLL 2003 NER | F1: 90.94 -> 91.47 | Improved |
| use IC-MTL instead of FS-MTL | FS-MTL · CoNLL 2000 chunking | F1: 95.41 -> 95.61 | Improved |
| use IC-MTL instead of FS-MTL | FS-MTL · PTB POS tagging | accuracy: 97.55 -> 97.69 | Improved |
| multi-task training with some baselines (FS-MTL, MT-CNN) vs single task on individual review tasks | FS-MTL · Apparel, Video, Magazines (text classification) | error rate (%): Apparel 15.7 -> 16.3 | Mixed |
Closest papers by meaning
- MTI-Net: Multi-Scale Task Interaction Networks for Multi-Task Learning (ECCV 2020 · similarity 0.88)
- Task Switching Network for Multi-Task Learning (ICCV 2021 · similarity 0.88)
- Conditionally Adaptive Multi-Task Learning: Improving Transfer Learning in NLP Using Fewer Parameters & Less Data (ICLR 2021 · similarity 0.87)
- Towards Graph Foundation Models: Learning Generalities Across Graphs via Task-Trees (ICML 2025 · similarity 0.86)
- Towards Dynamic Message Passing on Graphs (NeurIPS 2024 · similarity 0.85)
- Can Graph Learning Improve Planning in LLM-based Agents? (NeurIPS 2024 · similarity 0.85)
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.
