concept
active
concept:locally-scoped-emotion-representationsLocally scoped emotion representations
Vectors track operative emotion at a token position, not persistent state.
Neighborhood — ranked by edge-count
Papers (1)
paper
Concepts (2)
concept
- Emotion vectorsextendsLinear representations of emotion concepts extracted from activations.
- Early-middle layers encode present connotations; middle-late encode upcoming-token emotion.
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.
- Main conclusion about the temporal dynamics of emotion features
- Characterizes the temporal dynamics of emotion feature activation in LLMs
- Prior assumption that high-level variables align with disjoint groups of neurons in standard basis; contrasted with distributed representations.
- Interpretation of Part 2.
- Core summary of the paper's main empirical conclusion about emotion feature dynamics
- Not Assistant-specific.
- Interpretive hypothesis offered to explain why emotion features are more persistent
- Theoretical distinction between representing token-context pairs as individual features (local) vs combining independent features (compositional)