finding
active
finding:paired-auroc-uplift-between-0-012-and-0-004-with-all-95-cis-straddling-zero-across-tool-call-and-token-position-sweepsPaired AUROC uplift between -0.012 and +0.004 with all 95% CIs straddling zero across tool-call and token-position sweeps
Robustness to hyperparameters on Gemma.
Source paper
extracted_from(2026) · Adam Karvonen · Euan Ong · Subhash Kantamneni · Samuel Marks
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- No activation tool (AO, NLA, SAE) improves on transcript-only baseline for Gemma-3-27B-IT targetextendsMain negative result; baseline AUROC ~81 with tools within ~1 point.
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.
- Significant downlift for Gemini.
- Alternative phrase comparison.
- EVEE demonstrates strong generalization to indels without explicit training, indicating learned mechanistic principles.
- Steered loss can identify whether a dataset is likely to lead to misalignment
- Logit monitor reaches AUROC 0.79 from a single rollout separating Fortress from WildChat on Kimi K2.5finding0.771Key result.
- Filters do not drive the null.
- Shows that explicit instructions delay the emergence of truth directions in arithmetic tasks.
- Signal before any rollout.