WildSci: Advancing Scientific Reasoning from In-the-Wild Literature
This addresses the problem of scalable scientific reasoning for researchers in fields like medicine and materials science, though it is incremental as it builds on existing methods with new data.
The authors tackled the limited progress in LLM reasoning for scientific domains by introducing WildSci, a dataset of domain-specific science questions synthesized from peer-reviewed literature, and applied reinforcement learning to finetune models, resulting in improved performance on scientific benchmarks.
Recent progress in large language model (LLM) reasoning has focused on domains like mathematics and coding, where abundant high-quality data and objective evaluation metrics are readily available. In contrast, progress in LLM reasoning models remains limited in scientific domains such as medicine and materials science due to limited dataset coverage and the inherent complexity of open-ended scientific questions. To address these challenges, we introduce WildSci, a new dataset of domain-specific science questions automatically synthesized from peer-reviewed literature, covering 9 scientific disciplines and 26 subdomains. By framing complex scientific reasoning tasks in a multiple-choice format, we enable scalable training with well-defined reward signals. We further apply reinforcement learning to finetune models on these data and analyze the resulting training dynamics, including domain-specific performance changes, response behaviors, and generalization trends. Experiments on a suite of scientific benchmarks demonstrate the effectiveness of our dataset and approach. We release WildSci to enable scalable and sustainable research in scientific reasoning, available at https://huggingface.co/datasets/JustinTX/WildSci.