Answers are grounded in this paper and cite their sources.
Addendum to the preregistrations of Studies 1 to 3 on error-informed coding instructions for large language model coding of teacher talk moves. It fixes further runs requested by an external review, before any of them was made: repeated runs of existing conditions to measure run-to-run variation, a single general instruction in place of the learned rules, rules written by the optimizer without human codes, a hand-written prompt with examples, and an optimization that protects recall on rare moves. No hypothesis is confirmatory. The readings of the estimates are fixed in advance. The record holds the addendum, the instrument texts with their identifiers, and the scripts for the runs and the analysis. The examples in one instrument are excerpts from the TalkMoves dataset (Suresh et al., 2022), which is licensed under CC BY-NC-SA 4.0. The scripts are also available under AGPL-3.0-or-later.
No takes yet. Share an insight, caveat, or question.
Simon Filler (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: