| # References | |
| ## Core Papers | |
| 1. **I-JEPA** (arXiv:2301.08243) — Self-supervised learning with JEPA | |
| 2. **Image World Models** (arXiv:2403.00504) — Action-conditioned prediction in latent space | |
| 3. **DreamerV3** (arXiv:2301.04104) — RSSM world model with discrete latents | |
| 4. **HRM** (arXiv:2506.21734) — Hierarchical reasoning with two-timescale processing | |
| 5. **TTT for ARC** (arXiv:2411.07279) — Test-time training achieving 53% on ARC-AGI-1 | |
| ## ARC Benchmark | |
| 6. **ARC-AGI-3** (arXiv:2603.24621) — Competition specification | |
| 7. **ARC-AGI-2** (arXiv:2505.11831) | |
| 8. **ARC Prize 2025** (arXiv:2601.10904) — SOTA overview | |
| 9. **ARCLE** (arXiv:2407.20806) — RL environment | |
| 10. **ARCTraj** (arXiv:2511.11079) — Human trajectories | |
| 11. **GPAR** (arXiv:2401.07426) — Planning for ARC | |
| ## Competition | |
| - [ARC-AGI-3 on Kaggle](https://www.kaggle.com/competitions/arc-prize-2026-arc-agi-3) | |
| - Prize pool: $850,000 | |
| - Deadline: November 2, 2026 | |