Deep Reinforcement Learning from Human Preferences
GPTKB entity
Statements (47)
| Predicate | Object |
|---|---|
| gptkbp:instanceOf |
Research Paper
|
| gptkbp:addresses |
Sample Efficiency
|
| gptkbp:aimsTo |
Align AI behavior with human values
|
| gptkbp:appliesTo |
Autonomous_Systems
|
| gptkbp:author |
gptkb:Paul_Christiano
|
| gptkbp:contributedTo |
AI Safety
|
| gptkbp:designedBy |
New Training Algorithms
|
| gptkbp:develops |
Reward_Models
|
| gptkbp:discusses |
Ethical Implications
|
| gptkbp:exhibits |
Improved Performance
|
| gptkbp:focusesOn |
Reinforcement Learning
Human Feedback |
| https://www.w3.org/2000/01/rdf-schema#label |
Deep Reinforcement Learning from Human Preferences
|
| gptkbp:includes |
Experimental Results
|
| gptkbp:influences |
Future AI Development
|
| gptkbp:involves |
Human_Evaluators
|
| gptkbp:isChallengedBy |
Scalability Issues
Human Bias Interpretability Issues Generalization Problems Data_Quality_Concerns |
| gptkbp:isCitedBy |
Numerous Subsequent Studies
|
| gptkbp:isEvaluatedBy |
Benchmark Tests
|
| gptkbp:isExploredIn |
Education
Entertainment Finance Healthcare Marketing Robotics Security Transportation Game Playing Various Domains Social Good |
| gptkbp:isPartOf |
AI_Research_Community
|
| gptkbp:isRelatedTo |
Artificial Intelligence
Machine Learning Inverse Reinforcement Learning Behavioral Cloning Human-Computer_Interaction |
| gptkbp:isSupportedBy |
Empirical Evidence
Real-World Applications Simulation Environments Theoretical Frameworks |
| gptkbp:publishedIn |
NeurIPS 2017
|
| gptkbp:relatedTo |
Traditional Reinforcement Learning
|
| gptkbp:uses |
Preference-based Learning
|