An isotonic label-cleaning step removes score-versus-preference conflicts in robot reward training data, letting a 4B VLM reward model approach GPT-5-mini.
Eureka: Human-level reward design via coding large language models
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.AI 1years
2026 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
TrustRoboReward: Preference-Ordered Isotonic Score Editing for Multi-Paradigm Robot Reward Models
An isotonic label-cleaning step removes score-versus-preference conflicts in robot reward training data, letting a 4B VLM reward model approach GPT-5-mini.