AutothinkRAG: Complexity-Aware Control of Retrieval-Augmented Reasoning for Image-Text Interaction

This silver document aggregates recent arXiv papers (Mar 9, 2026) on retrieval-augmented reasoning, multimodal LLMs, RL-based optimization for agents and retrieval, synthetic video-data tooling, fake-news detection tooling, and an algorithmic audit of TikTok’s advertising/profile targeting. Highlights include: AutoThinkRAG (complexity-aware query routing and VLM-to-LLM functional decoupling for image-text DocQA); CBR-to-SQL (case-based retrieval for text-to-SQL in healthcare EHRs); a sensitivity-aware retrieval-augmented intent-clarification framework that defines attack models and retrieval‑f