Large language models have enabled automated agents to tackle complex scientific research tasks, yet most existing approaches struggle to deliver scientifically rigorous and verifiable outputs. In this work, we propose a novel agent framework, TrustSciAgent, which introduces a unified evidence–reasoning–validation pipeline explicitly governed by newly formulated scientific research trustworthiness principles. TrustSciAgent structurally organizes the entire research process into pre-research, in-research, and post-research phases, ensuring that each stage strictly adheres to these principles. This design compels the agent to generate transparent, logically sound reasoning chains and deliver auditable scientific conclusions. Comprehensive experiments across four scientific domains and three representative language models demonstrate that TrustSciAgent consistently improves both the structural completeness and the correctness of reasoning outputs, outperforming standard LLM-based agents. Our results provide strong evidence that embedding domain-agnostic trustworthiness principles into the agent workflow is critical for enabling credible, generalizable, and verifiable automated scientific research.

Recent advances in large language models have created unprecedented opportunities for scientific research and innovation, while also raising critical challenges related to academic integrity, intellectual property protection, and scientific data governance. In this paper, we present a comprehensive review of watermarking techniques for LLMs, systematically categorizing current approaches into post-processing watermarking and generative embedding watermarking methods, and analyzing their respective advantages, limitations, and technical characteristics. Beyond this survey, we further explore the application prospects of LLM watermarking in scientific scenarios, focusing on three key areas: the detection of academic misconduct, the protection of intellectual property, and the provenance tracing and attribution of synthesized scientific data. For each scenario, we propose corresponding conceptual frameworks and practical design recommendations, aiming to guide future research and promote the trustworthy, transparent, and reproducible application of LLM watermarking technologies in scientific research.