Research map / Vision-language and multimodal / Robot action and navigation
Reward Learning From Narrated Demonstrations
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 |
|---|---|---|---|
| Fine-tune reward detectors with narrated-video hard negatives instead of using the pretrained detector | Visual reward detector · Spatial relations, average | classification accuracy: 0.51 -> 0.92 | Helped |
| Use narrated-video hard negatives instead of randomly sampled static-image negatives (ablation) | Visual reward detector · Spatial relations, average | classification accuracy: 0.50 -> 0.92 | Helped |
| Use object-factorized policy input instead of RGB frames (ablation) | Pick-and-place policy · Place objects inside containers, seen objects | accuracy: 0.71 -> 0.96 | Helped |
| Use object-factorized policy input instead of RGB frames (ablation) | Pick-and-place policy · Place objects inside containers, unseen objects | accuracy: 0.27 -> 0.78 | Helped |
| Use learned detector rewards instead of oracle rewards (ablation) | ObjectPolicyNet · Place objects inside containers, seen objects | accuracy: 0.96 -> 0.88 | Hurt |
| Use learned detector rewards instead of oracle rewards (ablation) | ObjectPolicyNet · Place objects inside containers, unseen objects | accuracy: 0.78 -> 0.50 | Hurt |
| Use learned detector rewards instead of oracle rewards (ablation) | RGBPolicyNet · Place objects inside containers, seen objects | accuracy: 0.71 -> 0.71 | No clear effect |
| Use learned detector rewards instead of oracle rewards (ablation) | RGBPolicyNet · Place objects inside containers, unseen objects | accuracy: 0.27 -> 0.40 | Mixed |
| Use continuous distance-based reward shaping alongside binary rewards rather than binary rewards alone (ablation) | Pick-and-place policy learning · Place objects inside containers | policy training success | Helped |
Closest papers by meaning
- GoalLadder: Incremental Goal Discovery with Vision-Language Models (NeurIPS 2025 · similarity 0.90)
- Subtask-Aware Visual Reward Learning from Segmented Demonstrations (ICLR 2025 · similarity 0.89)
- Ground-Compose-Reinforce: Grounding Language in Agentic Behaviours using Limited Data (NeurIPS 2025 · similarity 0.89)
- RLZero: Direct Policy Inference from Language Without In-Domain Supervision (NeurIPS 2025 · similarity 0.88)
- ELEMENTAL: Interactive Learning from Demonstrations and Vision-Language Models for Reward Design in Robotics (ICML 2025 · similarity 0.88)
- SOLE-R1: Video-Language Reasoning as the Sole Reward for On-Robot Reinforcement Learning (NeurIPS 2026 · 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.