A recent paper on arXiv (arXiv:2608.06931v1) introduces Science Edge Evaluation (SEE), a multimodal benchmark for evaluating the capability of large language models (LLMs) in scientific discovery. The benchmark includes expert-curated questions grounded in peer-reviewed literature and experimental practice in chemistry, biology, and materials science. The evaluation results show that even the best-performing model among 19 multimodal large language models (MLLMs) reaches only 48.7% accuracy. Moreover, general-purpose models outperform science-specialized models on average. In the visual-agent evaluation, the use of tools increases the best accuracy to 52.7%. However, the key challenge is whether models can manage tool-derived information within the boundaries of the original experimental evidence to make reliable scientific reasoning. These findings reveal that current MLLMs still cannot reliably make justified and evidence-bounded inferences from experimental results, which is an essential capability in real scientific discovery. Bridging this gap requires MLLMs to transition from explaining established scientific concepts to deriving novel and evidence-based insights from experimental data. Blogger's Review: The SEE evaluation results reveal the limitations of large language models in scientific discovery, highlighting the challenge of making reliable inferences in experimental data, and emphasizing the need for further research and improvement to make MLLMs truly support scientific progress.