IR3: Contrastive Inverse Reinforcement Learning for Interpretable Detection and Mitigation of Reward Hacking

Mohammad Beigi, Ming Jin, Junshan Zhang, Jiaxin Zhang, Qifan Wang, Lifu HuangarXiv 2026 · DOI

BibTeX
@misc{beigi2026ir,
  author = {Mohammad Beigi and Ming Jin and Junshan Zhang and Jiaxin Zhang and Qifan Wang and Lifu Huang},
  title = {{IR3: Contrastive Inverse Reinforcement Learning for Interpretable Detection and Mitigation of Reward Hacking}},
  year = {2026},
  eprint = {2602.19416},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2602.19416},
}

Loading related content...

Figure