Research map / Reinforcement learning / RL from feedback
Polychromic Objectives for Reinforcement Learning
Jubayer Ibn Hamid, Ifdita Hasan Orney, Ellen Xu, Chelsea Finn, Dorsa Sadigh
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 |
|---|---|---|---|
| replace PPO with polychromic PPO | PPO · BabyAI/MiniGrid Goto | success rate (%): 46.2 -> 80.2 | Improved |
| replace PPO with polychromic PPO | PPO · BabyAI/MiniGrid Pickup | success rate (%): 33.4 -> 63.2 | Improved |
| replace PPO with polychromic PPO | PPO · BabyAI/MiniGrid Bosslevel | success rate (%): 38.8 -> 45.2 | Improved |
| replace PPO with polychromic PPO | PPO · BabyAI/MiniGrid Synthseq | success rate (%): REINFORCE 45.4 vs Poly-PPO 47.0 | Mixed |
| add UCB bonus to PPO | PPO · BabyAI/MiniGrid | success rate (%): Pickup 33.4 -> 27.8; Goto 46.2 -> 47.4 | Mixed |
| add UCB bonus to Poly-PPO | Poly-PPO · BabyAI/MiniGrid | success rate (%): Pickup 63.2 -> 65.6; Goto 80.2 -> 76.2 | Mixed |
| replace PPO with polychromic PPO under large initial-state perturbations | PPO · BabyAI/MiniGrid Goto, perturbed | pass@1 (%): 21.1 -> 60.6 | Improved |
| replace PPO with polychromic PPO under large initial-state perturbations | PPO · BabyAI/MiniGrid Four Rooms, perturbed | pass@1 (%): 15.3 -> 88.7 | Improved |
| use PPO instead of the pretrained policy for Four Rooms perturbed starts | PPO · BabyAI/MiniGrid Four Rooms, perturbed | pass@1 (%): 65.0 -> 15.3 | Hurt |
| apply PPO / REINFORCE fine-tuning | PPO and REINFORCE · BabyAI pass@k at large k | pass rate | Hurt |
| use polychromic PPO | Poly-PPO · Algorithmic Creativity triangle discovery | creativity and diversity | Improved |
| use polychromic PPO instead of PPO | PPO · Algorithmic Creativity triangle discovery | validity | Hurt |
Closest papers by meaning
- Preference Conditioned Multi-Objective Reinforcement Learning: Decomposed, Diversity-Driven Policy Optimization (NeurIPS 2026 · similarity 0.88)
- DGPO: Discovering Multiple Strategies with Diversity-Guided Policy Optimization (AAAI 2024 · similarity 0.88)
- Policy Filtration for RLHF to Mitigate Noise in Reward Models (ICML 2025 · similarity 0.88)
- Beyond the Dirac Delta: Mitigating Diversity Collapse in Reinforcement Fine-Tuning for Image Generation (NeurIPS 2026 · similarity 0.87)
- On Generalization Across Environments In Multi-Objective Reinforcement Learning (ICLR 2025 · similarity 0.87)
- Proximal Policy Gradient Arborescence for Quality Diversity Reinforcement Learning (ICLR 2024 · similarity 0.87)
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.
