Research map / Reinforcement learning / Visual and representation RL
Breadcrumbs to the Goal: Goal-Conditioned Exploration from Human-in-the-Loop Feedback
Marcel Torne Villasevil, Max Balsells I Pamies, Zihan Wang, Samedh Desai, Tao Chen, Pulkit Agrawal, Abhishek Gupta
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 |
|---|---|---|---|
| learn a parametric goal selector instead of directly using human-selected goals (DDL) (ablation) | HuGE · simulated goal-reaching benchmarks | human annotations needed: 50% less human annotations | Improved |
| pretrain with 5 noisy trajectories (BC + Ours) | HuGE · simulated goal-reaching benchmarks | timesteps to success | Improved |
| guide exploration with learned goal selector instead of indiscriminate exploration (LEXA) | HuGE · simulated goal-reaching benchmarks | success | Improved |
| use human feedback to bias exploration instead of learning a reward function (Human Preferences) | HuGE · simulated goal-reaching benchmarks | success | Improved |
| add large amounts of noise to human annotations (ablation) | HuGE · simulated goal-reaching benchmarks | convergence to optimal solution | No clear effect |
| query for human labels more frequently (ablation) | HuGE · simulated goal-reaching benchmarks | labels needed vs timesteps to success | Mixed |
| use real crowdsourced human feedback instead of synthetic labels | HuGE · kitchen environment | success | Improved |
| use dense reward PPO instead of HuGE | PPO · simulated goal-reaching benchmarks | success | Hurt |
Closest papers by meaning
- GUIDE: Real-Time Human-Shaped Agents (NeurIPS 2024 · similarity 0.90)
- DUO: Diverse, Uncertain, On-Policy Query Generation and Selection for Reinforcement Learning from Human Feedback (AAAI 2025 · similarity 0.90)
- Behavioral Exploration: Learning to Explore via In-Context Adaptation (ICML 2025 · similarity 0.89)
- Leveraging Sub-Optimal Data for Human-in-the-Loop Reinforcement Learning (ICLR 2025 · similarity 0.88)
- Decoupling Exploration and Policy Optimization: Uncertainty Guided Tree Search for Hard Exploration (NeurIPS 2026 · similarity 0.88)
- Efficient Language-instructed Skill Acquisition via Reward-Policy Co-Evolution (AAAI 2025 · 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.
