Risk Evaluation · Method

Beyond g-mean²: complementary metrics for monitorability

NUWA research note · independent review pending

First published: 2026-10-20Source checked: 2026-10-25

Original source ↗Cite this record ↓Deception →

NUWA Research Note

Research question Design and assumptions Source findings NUWA interpretation Limits and contrary evidence Research and decision relevance Institutions and roles Original resources

Model-assisted reading · Selected sections of the original full text or official article; a complete section-by-section audit remains pending.

At a glance

NUWA interpretation: discrimination, comparable-instance ordering and post-control risk answer different questions.

Research question

How can monitorability comparisons separate instance selection from deployment gating?

Design and assumptions

The article adds pair-matched comparisons on shared instances and a post-control rejection-gating simulation, plus lower/upper intervention bounds.

Source findings

Historical pair-matched rankings generally agreed with dataset-level rankings; disagreements highlighted selection-sensitive or near-tied cases.

Retry policies can restore some risk removed by a monitor gate.

NUWA interpretation

NUWA interpretation: discrimination, comparable-instance ordering and post-control risk answer different questions.

Limits and contrary evidence

Gating estimates depend on retry and deployment assumptions; uncertain causal attribution requires both intervention bounds.

Research and decision relevance

Model recommendation: show deployed-bad and any-deployment rates together rather than optimizing a single monitorability number.

Precise source locations

Beyond g-mean²: complementary metrics for monitorability · Why one monitorability number is not enoughSelected original resource https://alignment.openai.com/beyond-gmean-squared/; sha256:6c845aa01364667d193dc47aafa490ce24ff4185abb839e9c70109e2258a7ea7; archived source capture; exact capture time not independently established
Beyond g-mean²: complementary metrics for monitorability · Metric 1: compare models on shared instancesSelected original resource https://alignment.openai.com/beyond-gmean-squared/; sha256:6c845aa01364667d193dc47aafa490ce24ff4185abb839e9c70109e2258a7ea7; archived source capture; exact capture time not independently established
Beyond g-mean²: complementary metrics for monitorability · Historical system card resultsSelected original resource https://alignment.openai.com/beyond-gmean-squared/; sha256:6c845aa01364667d193dc47aafa490ce24ff4185abb839e9c70109e2258a7ea7; archived source capture; exact capture time not independently established
Beyond g-mean²: complementary metrics for monitorability · Metric 2: simulate the monitor as a rejection gateSelected original resource https://alignment.openai.com/beyond-gmean-squared/; sha256:6c845aa01364667d193dc47aafa490ce24ff4185abb839e9c70109e2258a7ea7; archived source capture; exact capture time not independently established
Beyond g-mean²: complementary metrics for monitorability · Gating reduces risk; retries give some risk backSelected original resource https://alignment.openai.com/beyond-gmean-squared/; sha256:6c845aa01364667d193dc47aafa490ce24ff4185abb839e9c70109e2258a7ea7; archived source capture; exact capture time not independently established

Source checking, extraction, editorial approval and replication are recorded separately. Methodology →

Institutions and roles

Published by OpenAI

Authors (3)

Melody Guan · Yuxiang Xie · Bowen Baker

Related work

Extends · Beyond g-mean²: complementary metrics for monitorability

Monitoring Monitorability

Revisits the original monitorability metric for complementary decision objectives.

Cite and reuse

Cite the original publication for its findings. Cite this NUWA record separately for its explanation and retained snapshot.

BibTeX ↓RIS ↓Record + source metadata · JSON ↓Reading coverage · JSON Suggest a correction

Original publication · BibTeX

@misc{monitorability-metrics,
  title = {Beyond g-mean²: complementary metrics for monitorability},
  url = {https://alignment.openai.com/beyond-gmean-squared/},
  author = {Melody Guan and Yuxiang Xie and Bowen Baker},
  year = {2026},
  date = {2026-10-20},
  version = {Source capture 2026-10-25},
  note = {Source capture 2026-10-25}
}

NUWA record · BibTeX

@misc{monitorability-metrics-nuwa,
  title = {Beyond g-mean²: complementary metrics for monitorability — NUWA research record},
  url = {https://whitzard.tech/en/nuwa/whitzard-index/work/monitorability-metrics/},
  version = {2026-10-29.1},
  organization = {NUWA Frontier AI Safety Lab},
  howpublished = {Retained note edition: https://whitzard.tech/nuwa/whitzard-index/data/record-editions/2026-10-29.1--monitorability-metrics.json},
  note = {NUWA research note. Source-reading scope and independent review status are stated on the record. Snapshot 2026-10-29.1. Retained note edition: https://whitzard.tech/nuwa/whitzard-index/data/record-editions/2026-10-29.1--monitorability-metrics.json.}
}

Evidence snapshot 2026-10-29.1 · Index updates · RSS

References

[1] Melody Guan, Yuxiang Xie, et al.. Beyond g-mean²: complementary metrics for monitorability. 2026-10-20 ; Official publication masthead and article byline; Research background and discussion in the original publication; Why one monitorability number is not enough