Doubao Large Model 2.1 Pro has recently launched a new 0915 version upgrade. This update focuses on real-world production scenarios, systematically reducing AI hallucinations, enhancing evidence traceability, and improving the reliability of multi-task processing, offering enterprise users and developers a more productive AI experience.

When dealing with complex business scenarios, the issue of "seriously talking nonsense" in large models has always been an industry pain point. The new version strengthens the evidence traceability mechanism, prioritizing information from authoritative sources, and combining time judgment and cross-data verification, making the Agent more fact-based when performing complex tasks. For example, in financial research and analysis scenarios that heavily rely on data analysis, the model can autonomously break down macro or micro research needs, coordinate hundreds of sub-Agents to simultaneously retrieve hundreds of web pages. Throughout this process, it not only compares conventional news releases but also introduces multi-dimensional external data such as satellite images and ship tracks to verify the findings, ensuring that each research conclusion can be traced back to the corresponding original evidence.

In addition to being more persuasive in text and analytical tasks, the multimodal programming and code engineering understanding capabilities have also evolved synchronously. When dealing with large code repositories, the new version can clearly clarify complex cross-file dependencies, accurately locate the root cause of bugs, and provide solutions. Meanwhile, thanks to its powerful visual language model capabilities, developers just need to provide operation recordings or a few hand-drawn sketches, and the system can convert them into executable web pages, game logic, or 3D scene code.

In terms of visual and multimedia understanding, the new version demonstrates outstanding long video processing capabilities. It can watch the entire long video content, extract key information across frames, and strictly compare it with standard operating procedures (SOP) to identify issues in on-site operations, perfectly adapting to professional fields such as engineering drawing interpretation, teaching experiment analysis, and product quality inspection.