Trending storyDeveloping
Ai2 launches explorer for comparing model and human tutor responses
3 articles1 sourcesince Oct 12Last article 1h ago ·
Article timeline
The articles in this story. Times are ET.
Oct 12
Ai2@allen_aiOfficialAi2 launches a tutor moments explorer comparing model and human tutor responsesAIAi2 announces a new explorer that compares responses from AI models and human tutors. The tool also shows model tutoring performance alongside response time and cost.
- Ai2@allen_aiOfficialAi2 builds TutorMoments, a teacher-annotated benchmark for math tutoring replies
AIAi2 says experienced teachers marked moments in math tutoring transcripts where tutors should offer support or ask more of a student. Models were then asked to continue those conversations and their replies were scored against the teachers' judgments.
- Ai2@allen_aiOfficialAi2 tests frontier models on TutorMoments, a tutoring eval
AIAi2 tested frontier models, including Astra and Opus 5.5, on TutorMoments, an eval of whether LLMs help students solve problems themselves or do too much for them. Models generally scored better with detailed prompts than when told only to "tutor well."
