Unverified50% confidenceFactExact time
The research team analyzed real usage data from a Socratic LLM chemistry tutor, comprising 20,462 student turns from 1,260 real learning sessions, and identified 6,630 impasse turns.
1
Sources
50%
Confidence
Long-term
Relevance
10/1/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe researchers eliminated the dependency on SenseNova at inference time by training a distilled student network with the architecture 5120→512→512→5120 (RMS normalization + SiLU activation) to reconstruct the projected teacher representations from H3's L49 conditions.69% similarUnverifiedThe research team developed a prototype system and conducted an eight-week teaching experiment with 186 non-English-major undergraduates, with the control group using static digital textbooks.65% similarUnverifiedBLOOM项目由数百名研究者通过Jean Zay超算集群协作完成，训练了1760亿参数的多语言模型65% similarUnverifiedA study published on arXiv systematically revealed the behavioral patterns of a Socratic AI tutor when students fall into persistent impasses, as well as the differential effects of various intervention strategies, by analyzing more than 20,000 real teacher–student dialogues.63% similarUnverifiedThe Machine Learning course launched by Andrew Ng on Coursera in 2011 has accumulated over 5 million enrollments and uses Octave/MATLAB as the teaching tool.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/963970API
curl https://kongchang.com/api/v1/knowledge/claims/963970MCP
get_claim(id=963970)