Paper Detail

Extending MONA in Camera Dropbox: Reproduction, Learned Approval, and Design Implications for Reward-Hacking Mitigation

Nathan Heath

arxiv Score 7.8

Published 2026-03-31 · First seen 2026-04-01

General AI

Abstract

Myopic Optimization with Non-myopic Approval (MONA) mitigates multi-step reward hacking by restricting the agent's planning horizon while supplying far-sighted approval as a training signal~\cite{farquhar2025mona}. The original paper identifies a critical open question: how the method of constructing approval -- particularly the degree to which approval depends on achieved outcomes -- affects whether MONA's safety guarantees hold. We present a reproduction-first extension of the public MONA Camera Dropbox environment that (i)~repackages the released codebase as a standard Python project with scripted PPO training, (ii)~confirms the published contrast between ordinary RL (91.5\% reward-hacking rate) and oracle MONA (0.0\% hacking rate) using the released reference arrays, and (iii)~introduces a modular learned-approval suite spanning oracle, noisy, misspecified, learned, and calibrated approval mechanisms. In reduced-budget pilot sweeps across approval methods, horizons, dataset sizes, and calibration strategies, the best calibrated learned-overseer run achieves zero observed reward hacking but substantially lower intended-behavior rates than oracle MONA (11.9\% vs.\ 99.9\%), consistent with under-optimization rather than re-emergent hacking. These results operationalize the MONA paper's approval-spectrum conjecture as a runnable experimental object and suggest that the central engineering challenge shifts from proving MONA's concept to building learned approval models that preserve sufficient foresight without reopening reward-hacking channels. Code, configurations, and reproduction commands are publicly available. https://github.com/codernate92/mona-camera-dropbox-repro

Workflow Status

Review status
pending
Role
unreviewed
Read priority
soon
Vote
Not set.
Saved
no
Collections
Not filed yet.
Next action
Not filled yet.

Reading Brief

No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.

Why It Surfaced

No ranking explanation is available yet.

Tags

No tags.

BibTeX

@article{heath2026extending,
  title = {Extending MONA in Camera Dropbox: Reproduction, Learned Approval, and Design Implications for Reward-Hacking Mitigation},
  author = {Nathan Heath},
  year = {2026},
  abstract = {Myopic Optimization with Non-myopic Approval (MONA) mitigates multi-step reward hacking by restricting the agent's planning horizon while supplying far-sighted approval as a training signal\textasciitilde{}\textbackslash{}cite\{farquhar2025mona\}. The original paper identifies a critical open question: how the method of constructing approval -- particularly the degree to which approval depends on achieved outcomes -- affects whether MONA's safety guarantees hold. We present a reproduction-first extension of the public MONA Came},
  url = {https://arxiv.org/abs/2603.29993},
  keywords = {cs.AI},
  eprint = {2603.29993},
  archiveprefix = {arXiv},
}

Metadata

{}