Local LLMs + RAG + BKT for adaptive ITS
Key Questions
How are local LLMs combined with RAG and BKT in adaptive tutoring?
The approach integrates local LLMs, retrieval-augmented generation (RAG), and Bayesian Knowledge Tracing (BKT) to create personalized tutoring plans and detect learning gaps. It leverages datasets such as StudyChat, Convolearn, and K12-KGraph for curriculum-aligned support.
What role does AI play in supporting refugees with disabilities?
AI tutors and chatbots deliver language and job skills training in refugee camps while improving accessibility for learners with disabilities. This fosters inclusive education through adaptive and scaffolded interactions.
What is Prashnadyuti and how is it used?
Prashnadyuti is an automatic multilingual QA generation system designed to create high-quality questions across languages. It supports efficient assessment and learning content creation for diverse educational settings.
What does AlpsBench measure in LLM personalization research?
AlpsBench evaluates LLM personalization using real-dialogue data and identifies gaps in current models and memory-centric methods. It includes benchmarking of frontier LLMs for adaptive tutoring scenarios.
Why focus on structured pedagogy rather than just answer quality in AI tutors?
Effective AI tutors use structured pedagogy to identify gaps, build student profiles, and incorporate dialog acts with skill modeling. This goes beyond response accuracy to include cognitive scaffolds and comprehensive tutoring plans.
StudyChat/Convolearn/CoMTA/CIMA datasets, AlpsBench gaps, K12-KGraph, UPenn/UCSD evals. New AI-structured pedagogy gap detection + tutoring plans, Prashnadyuti multilingual QA. Thesis: ed-RAG/KT/RL/cog, benchmarks, scaffolds. Inclusive AI tutors for refugees/disabilities.