Tag
Meta 的 IdeaScientist 论文将科研构思拆解为缺口发现、创新与报告撰写三个角色并分别用强化学习训练，基于 Qwen3.6-27B 在 2.77M 理念语料库上检索，超越最强开源 autoresearch 基线 14.0%，并优于 Claude Code SDK 与 Codex SDK 方案最多 5.9%。
This paper introduces AI Night-Scientist, a framework using reinforcement learning to enhance creative reasoning in large language models for scientific ideation, significantly improving idea diversity and impact.
IdeaTrail is a dataset of multi-turn process trajectories for scientific ideation, synthesizing research processes from evidence gathering to proposal construction using a Generator–Advisor loop to ensure grounding.
Proposes Agentic-Ideation, a framework for efficient synthesis of agentic trajectories to train LLMs for scientific ideation, achieving over 10x improvement in sample efficiency and outperforming existing workflow-based baselines.
This paper introduces RQ-Bench, a benchmark to evaluate LLMs' ability to assess the novelty of scientific research questions. It finds that LLM judges consistently rate generated questions as more novel than human experts do, raising concerns about the reliability of using LLMs for scientific novelty evaluation.
This paper systematically evaluates human creativity tests for LLMs and finds they fail to predict scientific ideation. It introduces the DRAT, a new test that combines convergent and divergent thinking to reliably predict scientific ideation ability in language models.