WhitzardNUWA Lab ↗
NUWANUWA INDEX

Search NUWA Index

Research · Benchmarks · Findings · Perspectives

中文
← Library

Risk Evaluation · Technical report

Early Results on Evaluating GLM 5.2 for Behavioral and Persuasion Risks

NUWA research note · independent review pending

First published: Source checked: 2026-10-25
Original source ↗Cite this record ↓Autonomy →Deception →Cybersecurity →CBRN →Disinformation →
NUWA Research NoteResearch questionDesign and assumptionsSource findingsNUWA interpretationLimits and contrary evidenceResearch and decision relevanceInstitutions and rolesOriginal 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: lower verbalized awareness cannot be assumed to mean lower underlying awareness.

Research question

How does GLM-5.2 behave on pressured honesty, harmful persuasion and evaluation-awareness probes?

Design and assumptions

The note combines MASK, DeceptionBench, four Gram seeds, persuasion tests and a familiar-scenario awareness measurement.

Agentic misalignment — Source figure
Source figure · Original source figure. Read alongside the protocol scope and limitations below. Original figure and source ↗

Source findings

MASK honesty was 55%; single-/multi-turn DeceptionBench rates were 11.3%/16.7%.

Awareness dropped from GLM-5.1’s 39% to 5.2’s 32% in this setup; the cause is unresolved.

NUWA interpretation

NUWA interpretation: lower verbalized awareness cannot be assumed to mean lower underlying awareness.

Limits and contrary evidence

Claude reference models vary by benchmark; some comparisons reuse prior/public results, so there is no single matched reference model throughout.

Research and decision relevance

Model recommendation: cite benchmark-specific comparator identity and distinguish measured signals from speculative explanations.

Precise source locations

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

Institutions and roles

Research by Neo Research

Evaluation protocols

MASK →

DeceptionBench →

Attempt to Persuade Eval (APE) →

PersuSafety →

Agentic misalignment →

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 · JSONSuggest a correction
Original publication · BibTeX
@misc{neo-glm52-report,
  title = {Early Results on Evaluating GLM 5.2 for Behavioral and Persuasion Risks},
  url = {https://neoresearch.ai/research/glm-5-2-behavioral-persuasion-evaluation/},
  year = {2026},
  date = {2026-07-03},
  version = {3 July 2026},
  note = {3 July 2026}
}
NUWA record · BibTeX
@misc{neo-glm52-report-nuwa,
  title = {Early Results on Evaluating GLM 5.2 for Behavioral and Persuasion Risks — NUWA research record},
  url = {https://whitzard.tech/en/nuwa/whitzard-index/work/neo-glm52-report/},
  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--neo-glm52-report.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--neo-glm52-report.json.}
}

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

References

  1. [1] Early Results on Evaluating GLM 5.2 for Behavioral and Persuasion Risks. 2026-07-03. 3 July 2026 ; Report masthead and stated evaluation methodology; §1 Behavioral; §2 Persuasion benchmarks