Research map / Reinforcement learning / MDPs and planning
The Impact of Task Underspecification in Evaluating Deep Reinforcement Learning
Vindula Jayawardana, Catherine H Tang, Sirui Li, Dajiang Suo, Cathy Wu
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 |
|---|---|---|---|
| Evaluate traffic signal controllers across an MDP family instead of a single intersection | Traffic signal control methods · Traffic signal control | relative method ranking: IPPO: 0.84 -> 1.7 | Mixed |
| Evaluate methods across a task's MDP family instead of its standard point MDP | DRL methods · Cartpole | relative method ranking: PPO 61%; TRPO 56% | Mixed |
| Use k-means clustering to select representative point MDPs | MDP-family performance estimation · Traffic signal control and control-task families | approximation accuracy | Improved |
| Use random sampling with replacements rather than representative clustering | MDP-family performance estimation · Control-task families | approximation variance | Hurt |
| Use random sampling without replacements rather than representative clustering | MDP-family performance estimation · Traffic signal control | approximation accuracy | Mixed |
| Use random sampling without replacements rather than sampling with replacements | MDP-family performance estimation · Cartpole, Pendulum, and Half-cheetah | approximation accuracy | Improved |
| Evaluate methods on a simplified point MDP rather than other family instances | DRL methods · Quad, Pendulum, and Swimmer | task performance | Mixed |
| Evaluate on one selected point MDP rather than a family of MDPs | DRL methods · Underspecified tasks | reliability of relative rankings | Hurt |
Closest papers by meaning
- Sample Efficient Myopic Exploration Through Multitask Reinforcement Learning with Diverse Tasks (ICLR 2024 · similarity 0.85)
- Scaling Offline Evaluation of Reinforcement Learning Agents through Abstraction (AAAI 2024 · similarity 0.85)
- SPOT: Scalable Policy Optimization with Trees for Markov Decision Processes (NeurIPS 2025 · similarity 0.85)
- Deep Reinforcement Learning for Robotics: A Survey of Real-World Successes (AAAI 2025 · similarity 0.85)
- Analyzing Generalization in Policy Networks: A Case Study with the Double-Integrator System (AAAI 2024 · similarity 0.85)
- Provable Benefits of Multi-task RL under Non-Markovian Decision Making Processes (ICLR 2024 · similarity 0.85)
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.
