Research map / Large language models / Agents, code and math
Goal-Driven Sequential Data Abstraction
Umar Riaz Muhammad, Yongxin Yang, Timothy M. Hospedales, Tao Xiang, Yi-Zhe Song
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 |
|---|---|---|---|
| select strokes holistically with GDSA instead of keep/remove decisions in original order (DSA) | GDSA · QuickDraw, budget 25%, RNN | category recognition accuracy (%): 38.36 -> 50.50 | Improved |
| select strokes holistically with GDSA instead of DSA | GDSA · QuickDraw, budget 25%, Sketch-a-Net | category recognition accuracy (%): 65.05 -> 71.92 | Improved |
| select strokes holistically with GDSA instead of DSA | GDSA · QuickDraw, budget 50%, RNN | category recognition accuracy (%): 67.89 -> 71.75 | Improved |
| select strokes with GDSA instead of first strokes in human order | GDSA · QuickDraw, budget 25%, RNN | category recognition accuracy (%): 36.66 -> 50.50 | Improved |
| select segments with GDSA instead of DSA | GDSA · TVSum, budget 1 segment | category recognition accuracy (%): 42.0 -> 68.0 | Improved |
| select segments with GDSA instead of DQSN | GDSA · TVSum, budget 25% | category recognition accuracy (%): 64.0 -> 74.0 | Improved |
| select segments with GDSA instead of DQSN | GDSA · TVSum, budget 50% | category recognition accuracy (%): 72.0 -> 76.0 | Improved |
| select sentences with GDSA instead of DSA (HAN classifier) | GDSA · Amazon reviews, budget 1 sentence | sentiment accuracy (%): 72.42 -> 83.77 | Improved |
| select sentences with GDSA instead of DQSN (HAN classifier) | GDSA · Amazon reviews, budget 25% | sentiment accuracy (%): 80.25 -> 86.08 | Improved |
| select sentences with GDSA instead of random sentences (RNN classifier) | GDSA · Amazon reviews, budget 25% | sentiment accuracy (%): 66.44 -> 73.39 | Improved |
| select sentences randomly instead of in original order | GDSA · Amazon reviews, budget 25%, HAN | sentiment accuracy (%): 76.06 -> 77.14 | No clear effect |
| replace the 2-state keep/remove action space with a pool-based picking policy trained by RL on a synthetic task | GDSA · synthetic 3x3 images, budget 2 | recognition accuracy: 50%-70% -> ~90% | Improved |
Closest papers by meaning
- Reconciling Spatial and Temporal Abstractions for Goal Representation (ICLR 2024 · similarity 0.85)
- Learning Planning Abstractions from Language (ICLR 2024 · similarity 0.84)
- Autonomous Option Invention for Continual Hierarchical Reinforcement Learning and Planning (AAAI 2025 · similarity 0.84)
- Goal-Oriented Skill Abstraction for Offline Multi-Task Reinforcement Learning (ICML 2025 · similarity 0.83)
- Compositional Automata Embeddings for Goal-Conditioned Reinforcement Learning (NeurIPS 2024 · similarity 0.83)
- Learning with Language-Guided State Abstractions (ICLR 2024 · similarity 0.83)
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.
