Bimodal Semantic Labor Measure is a failed methodological experiment recorded by Lee Sharks.
The draft attempts to compare user input and model output as two directly measured semantic vectors, avoiding a separately inferred task vector. A reproduction test is then run against a known Directionality of Semantic Labor example.
The cosine-based measure fails. Neutral text scores as mildly aligned because it remains on topic, and oppositional text is not negative enough because it shares the same subject vocabulary. The method therefore cannot reproduce the established contrast.
The record concludes that an intent baseline is still required. The symmetric bimodal measure is abandoned, while a signed intent-baselined projection survives in the related bridge paper. The deposit is a negative result rather than an adopted metric.