Research map / Graph learning / Knowledge and temporal graphs
Language-Conditioned Graph Networks for Relational Reasoning
Ronghang Hu, Anna Rohrbach, Trevor Darrell, Kate Saenko
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 LCGN contextualized features to single-hop classifier | single-hop · GQA val (detection features) | accuracy: 62.0% -> 63.9% | Improved |
| add LCGN to single-hop on GQA test-dev | single-hop · GQA test-dev | accuracy: 53.8% -> 55.8% | Improved |
| add LCGN to single-hop on grid features | single-hop · GQA val (grid features) | accuracy: 55.0% -> 55.3% | Improved |
| add LCGN to single-hop with GT objects and attributes | single-hop · GQA val (GT objects) | accuracy: 87.0% -> 90.2% | Improved |
| add LCGN to single-hop on CLEVR | single-hop · CLEVR test | accuracy: 72.6% -> 97.9% | Improved |
| add multiple message passing rounds (T=4 instead of T=1) (ablation) | single-hop + LCGN · CLEVR val | accuracy: 94.0% -> 97.9% | Improved |
| increase T from 4 to 5 (ablation) | single-hop + LCGN · CLEVR val | accuracy: 97.9% -> 96.9% | Hurt |
| add text conditioning to message passing (ablation) | single-hop + LCGN · CLEVR val | accuracy: 78.6% -> 97.9% | Improved |
| add dynamic per-iteration connection weights instead of static (ablation) | single-hop + LCGN · CLEVR val | accuracy: 96.5% -> 97.9% | Improved |
| add LCGN to GroundeR | GroundeR · CLEVR-Ref+ | accuracy: 61.7% -> 74.8% | Improved |
| add text conditioning to message passing (ablation) | GroundeR + LCGN · CLEVR-Ref+ | accuracy: 65.0% -> 74.8% | Improved |
| add dynamic connection weights instead of static (ablation) | GroundeR + LCGN · CLEVR-Ref+ | accuracy: 71.4% -> 74.8% | Improved |
Closest papers by meaning
- Context-Aware Graph Inference With Knowledge Distillation for Visual Dialog (TPAMI 2021 · similarity 0.89)
- A simple neural network module for relational reasoning (NeurIPS 2017 · similarity 0.88)
- HAIR: Hierarchical Visual-Semantic Relational Reasoning for Video Question Answering (ICCV 2021 · similarity 0.88)
- Systematic Relational Reasoning With Epistemic Graph Neural Networks (ICLR 2025 · similarity 0.86)
- Towards Foundation Models for Knowledge Graph Reasoning (ICLR 2024 · similarity 0.86)
- Talk like a Graph: Encoding Graphs for Large Language Models (ICLR 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.
