claim
active
claim:prioritizing-interpretability-during-training-may-beat-post-hoc-interpretationPrioritizing interpretability during training may beat post hoc interpretation
Intrinsic interpretability section.
Source paper
extracted_from(2025) · Lee Sharkey · Bilal I. Chughtai · Joshua Batson · Jack W. Lindsey +25
Neighborhood — ranked by edge-count
Papers (1)
paper
Methods (1)
method
- Intrinsically interpretable models using concepts in training.
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.
- Closing thesis.
- Finding that base models have high false positives and no net positive performance.
- Base pretrained models show high false positive rates and achieve no net task performance on concept injection detection; post-training essential for introspection.
- Developmental analogy used to explain sample efficiency under high ρd conditions
- Authors' interpretive endorsement of PSM view, backed by transfer experiments
- Extension of the thesis to deployed LLM inference via in-context learning
- Second falsifiable prediction linking objective function structure to valence profile
- Limitation statement.