ClueWeaver Dual-Agent Framework for Long-Narrative Reasoning in Compact LLMs

ClueWeaver uses a Finder agent to retrieve critical passages and an Interpreter agent to derive answers with paragraph-ID citations. Both agents utilize reward-guided reinforcement learning to improve evidence retention and reasoning accuracy in locally deployable, small-scale models.