claim
active
claim:g20b-s-post-training-committed-more-strongly-toward-and-against-particular-dispositions-than-q8b-leaving-fewer-traits-in-the-steerable-middleG20B's post-training committed more strongly toward and against particular dispositions than Q8B, leaving fewer traits in the steerable middle
Explains the distributional difference in generic domain: G20B has more mass at both extremes (natural and intractable)
Source paper
extracted_from(2026) · Winston Zeng · Ali Emami · J H Choi
Neighborhood — ranked by edge-count
Papers (1)
paper
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.
- G20B has 7 steerable, 7 natural, 4 intractable generic traits, with more mass at extremes than Q8Bfinding0.856G20B's post-training commits more strongly toward and against particular dispositions, leaving fewer in the steerable middle
- Both models independently converge on the same six clinician traits as natural defaults
- Q8B shows a broad steerable surface in the generic domain, like an uncommitted canvas
- Steering preferentially exposes exaggerated styles in G20B as well
- The costly error direction (labeling steerable as natural) almost never occurs
- Key finding showing agentic behavior is encoded as a default operating mode in both models
- Central interpretive claim and motivation for future work
- Validates G20B as a local judge for exploratory mapping