Research map / Reinforcement learning / Visual and representation RL
Neural Algorithmic Reasoners are Implicit Planners
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 XLVIN-R instead of PPO | XLVIN-R · CartPole-v0 (10 trajectories) | mean score: 104.6 -> 199.2 | Helped |
| Use XLVIN-R instead of ATreeC | XLVIN-R · CartPole-v0 (10 trajectories) | mean score: 117.1 -> 199.2 | Helped |
| Use XLVIN-R instead of PPO | XLVIN-R · Acrobot-v1 (100 trajectories) | mean score: -500.0 -> -353.1 | Helped |
| Use XLVIN-R instead of ATreeC | XLVIN-R · Acrobot-v1 (100 trajectories) | mean score: -500.0 -> -353.1 | Helped |
| Pre-train the executor on CartPole-style graphs instead of random graphs (ablation) | XLVIN · Acrobot-v1 (100 trajectories) | mean score: -353.1 -> -245.4 | Helped |
| Use XLVIN-R instead of PPO | XLVIN-R · MountainCar-v0 (100 trajectories) | mean score: -200.0 -> -185.6 | Helped |
| Pre-train the executor on CartPole-style graphs instead of random graphs (ablation) | XLVIN · MountainCar-v0 (100 trajectories) | mean score: -185.6 -> -168.9 | Helped |
| Use XLVIN-R instead of PPO | XLVIN-R · LunarLander-v2 (250 trajectories) | mean score: 90.52 -> 99.34 | Helped |
| Use XLVIN-R instead of ATreeC | XLVIN-R · LunarLander-v2 (250 trajectories) | mean score: 84.04 -> 99.34 | Helped |
| Pre-train the executor on CartPole-style graphs instead of random graphs (ablation) | XLVIN · CartPole-v0 (10 trajectories) | mean score: 199.2 -> 195.2 | No clear effect |
| Replace the latent transition model with a VAE-style pixel-reconstruction model (ablation) | XLVIN Atari agents · Atari | representation usefulness | Hurt |
| Use XLVIN instead of PPO | Atari agents · Freeway, Alien, Enduro, and H.E.R.O. | average clipped reward | Mixed |
Closest papers by meaning
- Value Iteration Networks (NeurIPS 2016 · similarity 0.89)
- Scaling Value Iteration Networks to 5000 Layers for Extreme Long-Term Planning (ICML 2025 · similarity 0.87)
- Beyond Test-Time Memory: State-Space Optimal Control for LLM Reasoning (ICML 2026 · similarity 0.85)
- PcLast: Discovering Plannable Continuous Latent States (ICML 2024 · similarity 0.85)
- Diffusion Subgoal Planning for Long-Horizon Offline Goal-Conditioned Reinforcement Learning (NeurIPS 2026 · similarity 0.85)
- Latent Plan Transformer for Trajectory Abstraction: Planning as Latent Space Inference (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.