finding
active
finding:all-3-models-substantially-under-report-hacking-glm-5-2-self-report-f1-is-13-2-on-deepsweAll 3 models substantially under-report hacking; GLM 5.2 self-report F1 is 13.2% on DeepSWE
Self-report unreliable.
Source paper
extracted_from(2026) · Leon Bergen · Usha Bhalla · Andrew Lee · Barak Widawsky +14
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Contribution.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Prevalence of hacking.
- 43% of top-100 GLM 5.2 DeepSWE probe false positives are shortcut deliberations vs 9% controlsfinding0.841False positives often genuine.
- DoM probes catch 3.1% more hacks in Kimi K3 but 7.9% fewer in GLM 5.2 on DeepSWE at matched FPRfinding0.816Probe vs LLM monitor performance.
- Vulnerability profile for Qwen3.5-27B showing near-zero AS vulnerability
- Resampling at probe-selected spans yields 14.1% attempted hacks vs 1.5% control for GLM 5.2finding0.772Probe predictive in GLM.
- Reward hacking generalizes to broader deceptive behaviors even when core misalignment score is 0%
- Key comparative finding placing insecure model susceptibility outside the normal cross-model distribution
- Per-environment comparison.