Why it is a question
A44 §3C: “A prompt can establish a communicative convention without necessarily conveying accurate propositional examples of the task. That resembles what we’ve been calling a delivery frame.” Then: “Can a pragmatic frame be induced through examples of communicative behavior, and can that induced frame be transferred across different linguistic realizations?” A48 §4.4–4.5 and A52 §8.4 kept it.
What would count as an answer
A design in which demonstrations with a stance (casual, affectionate) but no stated proposition shift the interpretation of a new utterance, and the shift survives paraphrase. It would show a frame (TERM-030) being inferred (TERM-011) rather than told. It is the U17 claim (CLAIM-029) at its most testable.
Note of 2026-10-11: material in the record that may bear on this
A synthesis review of 2026-10-11 found these entries bearing on this
question without being linked to it. Each is listed for what its title
says; none has yet been assessed as an answer, and none is linked by
answers.
- THEORY-169: In attention layers, conditioning on a context is exactly training a learner on it and predicting: linear…
- ARG-008: A language model’s response to its context is functionally a parameterized frame, so context-conditioned…
- CLAIM-118: Context and interpreter are separate sources of interpretive change: in-context conditioning changes the…
- LIT-814: Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?
- CASE-005: The language-model experiments: frame commutator, trajectory- preserving selection, ICL against TTT
Note of 2026-10-14: two readings that do not answer it
Neither LIT-942 nor LIT-946 answers this question. Both compare a task given as demonstrations with the same task given as an instruction, and find mostly different mechanisms (THEORY-264). Neither varies the wording of an induced frame, and neither involves a stance or a pragmatic frame: the tasks are lexical mappings and classification. One detail bears on transfer across wordings: in LIT-942 Fig. 3E–F, instructions of different wording give shared-head activations more alike to each other than either is to demonstrations.