hypothesis
active
hypothesis:post-trained-models-repurpose-user-token-activations-to-represent-the-assistant-s-thoughtsPost-trained models repurpose user token activations to represent the Assistant's thoughts.
Tested by J-lens comparison.
Source paper
extracted_fromAuthors
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- Assistant reactions.
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 mechanistic claim about the developmental origin of the Assistant persona
- Core interpretive claim providing mechanistic explanation for early persona formation
- Hidden goal model.
- Central interpretive claim and motivation for future work
- Implanted workspace.
- Core definitional quote for performative chain-of-thought
- Future work hypothesis about expanding SOO to use conversational role tags as self/other referents
- Characterizes what the Assistant persona resembles in terms of human archetypes