Research map / Reinforcement learning / Policy gradient and actor-critic
Knothe-Rosenblatt Quantile Regression for Risk-sensitive Multi-objective Reinforcement Learning
Gwangpyo Yoo, Woo Kyung Kim, Honguk Woo
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 |
|---|---|---|---|
| use KR-IQN instead of strongest baseline | KR-IQN · MO-Gymnasium (all envs) | average of EU/HV metrics: 29.77% average improvement | Improved |
| add shuffling of objective order when training the critic (ablation) | KR-IQN(RS) · Hopper (d=3) | EURisk: -34.3% EURisk and -27.2% HVrisk when removed | Improved |
| use transformer without positional encoding (ablation) | KR-IQN(RS) · Hopper (d=3) | EURisk: -66.9% EURisk and -58.3% HVrisk when PE added | Improved |
| add projective TQC to marginal TQC (ablation) | KR-IQN(RS) · Hopper (d=3) | EURisk: marginal alone -46.0% EURisk | Improved |
| use projective TQC alone (ablation) | KR-IQN(RS) · Hopper (d=3) | EURisk: -56.9% EURisk vs removing all TQC | Hurt |
| train with CVaR50% risk-sensitive critic instead of neutral | KR-IQN · Finance test | earning (%): 18.09 -> 25.40 | Improved |
| use KR map instead of marginal quantiles | KR-IQN vs MarginalIQN · Finance test | earning (%): 15.73 -> 25.40 (CVaR50%) | Improved |
| use risk-sensitive training (RS) | KR-IQN · MO-Gymnasium Hopper, Walker2d, Ant | EURisk: 934.89 -> 1086.62 (Hopper) | Improved |
| use KR-IQN | KR-IQN · Finance (MO-Gymnasium table) | HV: DPMORL 19.56 vs KR-IQN(RS) 11.87 | Hurt |
| use KR-IQN in Cheetah | KR-IQN · Cheetah | EU: 2097.21 -> 2172.71 | No clear effect |
Closest papers by meaning
- Risk-Conditioned Reinforcement Learning: A Generalized Approach for Adapting to Varying Risk Measures (AAAI 2024 · similarity 0.88)
- Quantile Geometry Regularization for Distributional Reinforcement Learning (NeurIPS 2026 · similarity 0.88)
- Tilted Quantile Gradient Updates for Quantile-Constrained Reinforcement Learning (AAAI 2025 · similarity 0.87)
- Provable Risk-Sensitive Distributional Reinforcement Learning with General Function Approximation (ICML 2024 · similarity 0.87)
- AETDICE: Unified Framework and Offline Optimization for Nonlinear Multi-Objective RL (NeurIPS 2026 · similarity 0.86)
- Beyond CVaR: Leveraging Static Spectral Risk Measures for Enhanced Decision-Making in Distributional Reinforcement Learning (ICML 2025 · similarity 0.86)
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.
