Fetching from the wire…
01
02
03
04
05
06
07
08
Research2026-08-08 · source-backed
TutorMoments replays real decision points from 462 de-identified one-on-one math tutoring transcripts with 1,500+ teacher-marked moments and thousands of annotations from 27 educators. Across seven models, default assistant behavior was to over-scaffold, making problems easier instead of letting students reason. Performance improved substantially once the prompt explicitly named the scaffolding-versus-rigor trade-off, though variance stayed wide. Transferable well beyond education: if your task requires the model to hold back, you have to say so explicitly.
Each link below shares sources, entities, or timing with this story.
AI2 released OlmoEarth / Shared entity: AI2 / Same source domain / Earlier coverage
Linked by a graph relationship (AI2 released OlmoEarth); both cover AI2; reported by the same outlet (huggingface.co).
Shippy built by AI2 / Shared entity: AI2 / Same source domain / Earlier coverage
Linked by a graph relationship (Shippy built by AI2); both cover AI2; reported by the same outlet (huggingface.co).
Shared entity: Performance / Shared topic / Earlier coverage
Both cover Performance; overlapping topics (behavior, model); earlier Performance coverage from 2026-06-19.
Same source domain / Shared topic / Tension
Reported by the same outlet (huggingface.co); overlapping topics (correct, model); pushes against this story (but).
Shared entity: Performance / Earlier coverage / Tension
Both cover Performance; earlier Performance coverage from 2026-08-02; pushes against this story (but).
Both cover Performance; earlier Performance coverage from 2026-07-21; pushes against this story (against).
Same source domain / Shared topic
Reported by the same outlet (huggingface.co); overlapping topics (assistant, behavior, correct).
Shared entity: Performance / Earlier coverage / Tension
Both cover Performance; earlier Performance coverage from 2026-04-08; pushes against this story (vs).