Research map / Reinforcement learning / Visual and representation RL
NIL: No-data Imitation Learning
Mert Albaba, Chenhao Li, Markos Diomataris, Omid Taheri, Andreas Krause, Michael J. Black
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 regularization term to the reward (ablation) | NIL · Unitree H1 walking | environment reward: 382.4 -> 396.1 | Improved |
| add IoU mask reward (ablation) | NIL · Unitree H1 walking | environment reward: 381.4 -> 396.1 | Improved |
| add video similarity reward (ablation) | NIL · Unitree H1 walking | environment reward: 387.3 -> 396.1 | Improved |
| use only IoU reward instead of all components (ablation) | NIL · Unitree H1 walking | environment reward: 396.1 -> 328.4 | Hurt |
| use only video similarity instead of all components (ablation) | NIL · Unitree H1 walking | environment reward: 396.1 -> 369.6 | Hurt |
| use Kling v1.6 instead of v1.0 for reference video generation (ablation) | NIL · Unitree H1 walking | environment reward | No clear effect |
| use more visually plausible video diffusion model (lower LPIPS to mocap video) (ablation) | NIL · Unitree H1 walking | environment reward | Improved |
| use NIL instead of AMP on Talos | AMP · Talos walking | environment reward: 231.1 -> 352.8 | Improved |
| use NIL instead of AMP on Unitree G1 | AMP · Unitree G1 walking | environment reward: 393.4 -> 356.9 | Hurt |
| use NIL with generated video instead of RL with engineered rewards | BRO RL baseline · H1 loco-manipulation (sit, highbar, balance) | success rate / normalized reward | No clear effect |
Closest papers by meaning
- What's the Move? Hybrid Imitation Learning via Salient Points (ICLR 2025 · similarity 0.89)
- DynaMo: In-Domain Dynamics Pretraining for Visuo-Motor Control (NeurIPS 2024 · similarity 0.88)
- Imitation from Observations with Trajectory-Level Generative Embeddings (NeurIPS 2026 · similarity 0.88)
- Robots Pre-train Robots: Manipulation-Centric Robotic Representation from Large-Scale Robot Datasets (ICLR 2025 · similarity 0.88)
- VLMimic: Vision Language Models are Visual Imitation Learner for Fine-grained Actions (NeurIPS 2024 · similarity 0.88)
- AnySkill: Learning Open-Vocabulary Physical Skill for Interactive Agents (CVPR 2024 · similarity 0.88)
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.
