Research map / Probabilistic and scientific modeling / Dynamical systems
Forward Modeling for Partial Observation Strategy Games - A StarCraft Defogger
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 |
|---|---|---|---|
| Use a Convolutional-LSTM encoder rather than the best heuristic baseline | StarCraft defogging model · STARDATA test, g=64, s=15s | op_u F1: 0.53 -> 0.62 | Helped |
| Use a Convolutional-LSTM encoder rather than the best heuristic baseline | StarCraft defogging model · STARDATA test, g=64, s=15s | hid_u F1: 0.47 -> 0.56 | Helped |
| Use a Convolutional-LSTM encoder rather than the best heuristic baseline | StarCraft defogging model · STARDATA test, g=64, s=15s | Huber loss: 28.97 -> 10.40 | Helped |
| Use a Convolutional-LSTM encoder rather than a ConvNet encoder (ablation) | StarCraft defogging model · STARDATA test, g=64, s=15s | op_u F1: 0.53 -> 0.62 | Helped |
| Use a Convolutional-LSTM encoder rather than a ConvNet encoder (ablation) | StarCraft defogging model · STARDATA test, g=64, s=15s | hid_u F1: 0.51 -> 0.56 | Helped |
| Use a Convolutional-LSTM encoder rather than a ConvNet encoder (ablation) | StarCraft defogging model · STARDATA test, g=32, s=15s | op_u F1: 0.48 -> 0.51 | Helped |
| Use a Convolutional-LSTM encoder rather than a ConvNet encoder (ablation) | StarCraft defogging model · STARDATA test, g=32, s=15s | Huber loss: 0.488 -> 0.430 | Helped |
| Use defogger predictions for Build Actions only | Rule-based StarCraft bot · Full-game bot, all matchups | win rate (%): 61 -> 66 | Helped |
| Use defogger predictions for Tactics only | Rule-based StarCraft bot · Full-game bot, all matchups | win rate (%): 61 -> 61 | No clear effect |
| Use 30-second defog predictions for both Tactics and Build Actions | Rule-based StarCraft bot · Full-game bot, all matchups | win rate (%): 61 -> 49 | Hurt |
| Use defogger predictions for both Tactics and Build Actions | Rule-based StarCraft bot · Full-game bot, all matchups | win rate (%): Defog t+0s: 59; Defog t+5s: 61; Defog t+30s: 49 | Mixed |
| Replace the LSTM with convolutions over time and no RNN (ablation) | StarCraft defogging model · Preliminary proxy-task experiments | performance | Hurt |
Closest papers by meaning
- StarCraft Motion: A Dataset for Agent Simulation in Adversarial and Partially Observable Scenarios (NeurIPS 2026 · similarity 0.91)
- DefogGAN: Predicting Hidden Information in the StarCraft Fog of War with Generative Adversarial Nets (AAAI 2020 · similarity 0.91)
- Chess-World-Model: A 10M-Game Benchmark for Exact State Tracking from Chess Move Sequences (NeurIPS 2026 · similarity 0.85)
- Diffusion Models Are Real-Time Game Engines (ICLR 2025 · similarity 0.85)
- Large Language Models Play StarCraft II:Benchmarks and A Chain of Summarization Approach (NeurIPS 2024 · similarity 0.84)
- LLM-PySC2: Starcraft II learning environment for Large Language Models (NeurIPS 2025 · 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.