finding
active
finding:sae-feature-emotion-subspace-overlap-correlates-with-persistence-in-cogito-spearman-0-413-p-4-4e-196SAE feature emotion subspace overlap correlates with persistence in Cogito: Spearman +0.413, p=4.4e-196
Demonstrates that SAE features more aligned with the emotion subspace are more persistent in Cogito after variance control
Source paper
extracted_fromScott Sauers · Imago · Janus · Antra Tessera
Neighborhood — ranked by edge-count
Claims (1)
claim
- Central interpretive claim of the paper supported by multiple convergent analyses
Methods (2)
method
- Fraction of an SAE feature's length lying inside the 171-dimensional subspace spanned by emotion probes, computed via SVD orthogonalization
- P(feature fires at t+100 | fired at t) minus P(feature fires at t+100 | did not fire at t), used because SAE features are binary unlike probe activations
Findings (1)
finding
- Strong positive relationship between emotion alignment and SAE feature persistence in Cogito
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.
- Emotion probe persistence correlation of 0.214 in Cogito v2.1 vs 0.099 for random vectorsfinding0.820Quantifies emotion feature persistence above random baseline in Cogito across 240 multi-turn conversations
- Quantitative measure of emotion feature persistence vs random baseline in Cogito
- Explains why variance correction is needed to see the self-evaluation–persistence relationship
- Shows that model self-report of emotion predicts long-range feature persistence
- Agentic self-evaluation emotionality correlates with SAE feature persistence: rho=+0.124, p=0.0001finding0.814Shows that features Kimi rates as more emotional via self-steering are more persistent, independent of probe construction
- Surprising finding that the two evaluation methods diverge in their relationship with persistence
- Shows self-evaluated emotionality is negatively confounded by variance, requiring variance control to reveal the true signal
- Novel finding that agentic self-evaluation of emotionality correlates with feature persistence
Restated by (1)
cosine ≥ 0.90Other entities that say roughly the same thing. May be merge candidates or independent restatements across papers.