framework
active
framework:turn-averaged-sae

Turn-Averaged SAE

Paper's main proposal: SAE trained on the mean activation across a turn.

Neighborhood — ranked by edge-count

Methods (1)

method

Concepts (1)

concept

Frameworks (4)

framework
  • Method using large language models (Claude) to generate and test explanations of features at scale
  • Hybrid SAE jointly training turn-averaged inner and per-token outer partitions.
  • Hybrid with separate turn-averaged and residual per-token SAEs.
  • Temporal SAEs
    contradicts
    Contrastive loss approach for smooth features across adjacent tokens; contrasted with turn averaging.

Datasets (2)

dataset
  • Chat dataset (n=18,777 sampled) used to measure how much persona space PCs explain overall activation variance and to calibrate steering norms
  • Cross-family model used for replication of wellbeing concept analysis; shows strong introspection (ρ=0.49, R²=0.76)

Related by similarity (8)

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.