Does agentic reasoning (plan → retrieve → reason → verify) beat direct prompting and standard RAG on scientific QA? SPIQA, three 4-bit 3–4B models, one 6 GB laptop GPU. CSE495B.1, North South University.
nlp question-answering quantization retrieval-augmented-generation llm-agents langgraph spiqa scientific-qa
-
Updated
Sep 22, 2026 - Python