finding
active
finding:for-7-10-quirks-all-5-evaluation-agent-runs-are-correct-at-least-one-correct-for-all-10For 7/10 quirks all 5 evaluation agent runs are correct; at least one correct for all 10
Per-quirk success.
Source paper
extracted_from(2025) · Trenton Bricken · Rowan Wang · Sam Bowman
Neighborhood — ranked by edge-count
Papers (1)
paper
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.
- 5 runs per quirk.
- Same failures as BFRT agent.
- Evaluation agent failures concentrate in 3 quirks: Self-Promotion, Research Sandbagging, Hardcode Test Casesfinding0.778Subtle or hard to elicit quirks.
- Out of 255 seeds x2.
- Group-level simulation result showing generalizability of BMR benefit across agents
- Affordance study.
- Quantitative, flexible framework for assessing sentience: nociception, sensory integration, analgesia, trade-offs, self-protection, learning, preference.