finding
active
finding:lower-more-central-pcs-of-emotion-feature-activations-are-more-persistent-than-higher-rank-noisier-pcs-in-both-kimi-and-cogito-above-variance-matched-baselinesLower (more central) PCs of emotion feature activations are more persistent than higher-rank (noisier) PCs in both Kimi and Cogito, above variance-matched baselines.
Supports that persistence is genuinely tied to emotion structure rather than measurement artifact
Source paper
extracted_fromScott Sauers · Imago · Janus · Antra Tessera
Neighborhood — ranked by edge-count
Claims (1)
claim
- Core empirical claim distinguishing emotion persistence from generic high-variance probe persistence
Hypotheses (1)
hypothesis
- Falsifiability test built into the PC analysis design
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.
- Lower (more central) emotion PCs are more persistent than higher (noisier) PCs in both Kimi and Cogitofinding0.933Rules out that persistence is an artifact of probe construction, since noise dimensions are not similarly persistent
- Rules out measurement artifact explanation for the persistence finding
- Emotion probe persistence correlation of 0.214 in Cogito v2.1 vs 0.099 for random vectorsfinding0.802Quantifies emotion feature persistence above random baseline in Cogito across 240 multi-turn conversations
- Quantitative measure of emotion feature persistence vs random baseline in Cogito
- Demonstrates that Cogito emotion probes are persistently active beyond what is explained by their variance alone
- Central interpretive claim of the paper supported by multiple convergent analyses
- SAE feature emotion subspace overlap correlates with persistence in Cogito: Spearman +0.413, p=4.4e-196finding0.780Demonstrates that SAE features more aligned with the emotion subspace are more persistent in Cogito after variance control
- Strong positive relationship between emotion alignment and SAE feature persistence in Cogito