claim
active
claim:models-can-modulate-workspace-contents-when-instructed-but-control-is-imperfect-and-sensitive-to-phrasingModels can modulate workspace contents when instructed, but control is imperfect and sensitive to phrasing.
Directed modulation.
Source paper
extracted_fromAuthors
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (2)
finding
- Suppression imperfect.
- J-space privileged.
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 consequence: GPT's power comes from simulating something contingent.
- Central thesis.
- The workspace may be verbalizable because the model's output space and input space are verbal.claim0.796Speculative explanation.
- Today, as a step towards the control of complex dynamic systems, models are being used ubiquitously.quote0.795Opening sentence of the abstract, stating the prevalence of modeling.
- Antra's functional claim about the performance impact of validating model self-reports.
- Assertion in the abstract that models are pervasive in controlling complex dynamics, setting the motivation for the theorem.
- Key jailbreak insight.
- Tested in D.1.