claim
active
claim:probe-output-can-be-interpreted-as-a-posterior-belief-uncertainty-over-a-conceptProbe output can be interpreted as a posterior belief (uncertainty) over a concept.
Framework premise.
Source paper
extracted_from(2026) · Aaditya Vikram Prasad · Connor Watts · Jack Merullo · Dhruvil Gala +3
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- Reward probes attend to entity and relation tokens; Failed Fix shows higher attention on punctuationassociated_withAttention map observation.
Claims (1)
claim
- Central thesis.
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.
- Key interpretive claim from Case Study II distinguishing probe accuracy from causal relevance
- Interpretation of resampling.
- Probe predictions are systematically miscalibrated until isotonic regression corrects biasfinding0.782Calibration necessity.
- Caveat on probe interpretation; does not negate the introspection result but affects interpretation of the target variable
- Motivation for causal evaluation over purely behavioural probing accuracy
- Conceptual framing: integrates mechanistic interpretability tools with alignment-focused data curation.
- Key methodological insight: introspection enables a new probe validation criterion beyond conventional separation metrics
- Empirical finding contrasting difficult questions with easy ones, supporting genuine reasoning on hard tasks