Contra-Positive Coherence

Measure of whether LMs follow contrapositive logic: if A→B is asserted true and B is false, then A should be false.

Neighborhood — ranked by edge-count

paper

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.

Bilateral Coherenceconcept0.820
Percentage of time a model's assessment of the veracity of base and entailed properties match, given the entailing statement is true.
Coherenceconcept0.813
A property that makes a segment of space stand out as a center; determined by symmetry, connectedness, convexity, etc.
Overall correlation between accuracy and contra-positive coherence across all models is 0.41 on Leap-of-Thought.finding0.781
Quantitative result showing weaker relationship between accuracy and contra-positive coherence.
Coherence Scoremethod0.771
GPT-4.1-mini-rated 0-100 score measuring response coherence; used to detect side effects of steering
long-range coherenceconcept0.763
Ability to maintain structural consistency over extended sequences
Coherence Maximizationconcept0.761
The functional role consciousness plays: minimizing constraint violations between simultaneously active partial models of reality
Perceived coherence of patterns is an objective measure, not idiosyncratic or subjective—people agree on relative coherence regardless of experimental taskfinding0.761
Finding that relative coherence rankings remain constant across different people and across different cognitive processing tasks (description, memorization, tachistoscopic recognition), establishing coherence as an objective feature of cognitive processing
Coherency Scoremethod0.757
GPT-4.1-mini based score (0-100) measuring clarity, absence of hallucinations, and lack of confusion in generated text