突破科学发现中假设生成的计算瓶颈，实现高效可扩展训练。
MOOSE-Star: Unlocking Tractable Training for Scientific Discovery by Breaking the Complexity Barrier
- 将复杂推理分解为子任务，通过分层搜索实现对数级检索。
- 在10万+论文数据集上训练，性能随数据和算力持续提升。
- 适合需要大规模科学假设生成的研究者与自动化发现系统。
尽管大语言模型在科学发现中展现出潜力，但现有研究多集中于推理或反馈驱动训练，未探索直接建模生成推理过程 $P( ext{hypothesis}| ext{background})$（即 $P(h|b)$）的可能性。我们证明，由于从海量知识库中检索并组合灵感存在组合复杂性（$O(N^k)$），直接训练 $P(h|b)$ 在数学上不可行。为此，我们提出 MOOSE-Star 框架，通过（1）基于发现概率方程分解子任务进行训练，（2）采用动机引导的分层搜索实现对数级检索并剪枝无关空间，（3）使用有界组合增强对检索噪声的鲁棒性，将复杂度从指数级降至对数级（$O(/log N)$）。为支持该方法，我们发布 TOMATO-Star，一个包含 108,717 篇分解论文的数据集（耗时 38,400 GPU 小时）。实验证明，MOOSE-Star 能随训练数据与推理预算持续扩展，而暴力采样则遭遇复杂性瓶颈。
原文摘要 · Abstract (English)
While large language models (LLMs) show promise in scientific discovery, existing research focuses on inference or feedback-driven training, leaving the direct modeling of the generative reasoning process, $P(\text{hypothesis}|\text{background})$ ($P(h|b)$), unexplored. We demonstrate that directly training $P(h|b)$ is mathematically intractable due to the combinatorial complexity ($O(N^k)$) inherent in retrieving and composing inspirations from a vast knowledge base. To break this barrier, we introduce MOOSE-Star, a unified framework that enables tractable and scalable training of $P(h|b)$, while supporting more scalable inference. In the best case, MOOSE-Star reduces complexity from exponential to logarithmic ($O(\log N)$) by (1) training on decomposed subtasks derived from the probabilistic equation of discovery, (2) employing motivation-guided hierarchical search to enable logarithmic retrieval and prune irrelevant subspaces, and (3) utilizing bounded composition for robustness against retrieval noise. To facilitate this, we release TOMATO-Star, a dataset of 108,717 decomposed papers (38,400 GPU hours) for training. Empirically, MOOSE-Star scales continuously with training data and inference budget, whereas direct brute-force sampling hits a complexity wall.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。