concept
active
concept:kv-state-persistenceKV State Persistence
The key-value cache from steered tokens is retained during no-steering continuation, allowing causal effect of steering to propagate
Neighborhood — ranked by edge-count
Methods (1)
method
- Applies a 5-token steering pulse to each emotion probe and measures persistence of causal effect via contrast z-score over 200 subsequent tokens
Concepts (1)
concept
- The causal steering experiment persists KV state over steered tokens so downstream effects can be observed without continued steering
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.
- The property of emotion features maintaining elevated activation well beyond the local token context that triggered them
- Janus's claim about KV caching as an introspection mechanism.
- Caching of key-value pairs to avoid recomputation; also provides a mechanism for introspection of earlier computations.
- The phenomenon that emotion feature activations remain elevated above baseline beyond local token bursts, measurable as long-range correlation
- Emotion feature persistence above and beyond the persistence expected from high variance explained alone, computed by subtracting median variance-matched probe persistence
- Core logical puzzle: if an agent does not change, it dies; if it changes, the self ceases to exist. Applies to all scales from organelles to evolutionary lineages.
- Baseline persistence of any probe direction arising from the autoregressive nature of LLMs, not specific to emotion content
- Core logical paradox: if a species fails to change it dies; if it changes, it ceases to exist. Same applies to individuals.