March 3 News.iFLYTEKAnnounced todayspark X1 upgrades and releases a series of new products such as the Starfire All-in-One.
Test collection source: Chinese test collection from 2023/2024 examination questions / simulation questions / competition questions of each academic level, English test collection from AIME 2024/2025 and MATH 500 competition collection.
in,Deep Reasoning Large ModelsThe Starfire X1 has been upgraded.Math capabilities fully benchmarked against DeepSeek R1 and OpenAI o1.
KU Xunfei said that the upgrade has improved the math question answering effect across the board, especially the remarkable performance in dealing with competition-level puzzles. At the same time, StarFire X1 has also shown significant advantages in tasks such as correction of primary and secondary school math homework, tutoring, and topic recommendation.
In the upgraded Starfire X1 support, Xunfei Xiaomedicine, Starfire Teacher Assistant, AI Judge Assistant and other applications and products are also upgraded, IT home with the official introduction is as follows:
CyberXiao Medical debuts Starfire Medical Grand Model X1
For this upgrade, Cyberoam Medical debuts the Starfire Medical Grand Model X1, which is called "Significantly reduced medical hallucination problems", the ability to explain evidence-based processes step-by-step when answering complex questions improves the logical correctness, professionalism, and interpretability of reasoning in medically complex scenarios.It significantly outperforms GPT-4o and DeepSeek R1 in reasoning tasks such as diagnosis recommendation, health counseling, and test report interpretation..
At present, Starfire medical model X1 is formally applied to the AI health assistant "Xunfei Xiaomedicine", and the integration with personal health records is more in-depth, and in the user's health consultation, it can provide in-depth insight into the user's health status and think about the potential needs, solving the pain point of the past AI medical consulting, "suggesting broad answers not practical", and giving more personalized medical advice. Practical" pain point in the past AI medical consultation, the medical advice given is more personalized, more specific and practical.
For example, when a user inquires about medication for a disease, Xiaomedic is able to point out the correct principles of medication for that disease based on evidence, and at the same time give more appropriate and detailed treatment advice and precautions, taking into account the individual's previous health condition.
Starfire Teacher's Assistant
Comprehensive upgrade based on CyberLink Starfire X1The Starfire Teacher's Assistant for lesson planning has also been further upgraded to give clearer and more detailed AI teaching ideas..
- In the preparation of lessonsCompared with the previous direct linear process of generating instructional design, the Starfire Teacher's Assistant is now able to combine the teacher's intention, clarify the key points of teaching and the objectives to be achieved, and explain them step by step during the process of generating instructional design, which makes it easy for the teacher to understand the logic of the generated content and inspire creative inspiration.
- In the course of the lessonThe presentation of the chain of thought is explicit, allowing teachers and students to observe the cognitive thinking construction process in real time, making classroom interaction more intuitive and easier to carry out inquiry-based classroom.
These competencies have been applied in a number of subjects, including math, language arts, and history.
Upgraded Starfire Teacher's Aide.Visualize the thinking of AI tools in teaching and learning, from students' questions on basic issues to advanced exploration of complex issues.that are more conducive to instructional innovation and the development of higher-order thinking skills in students.
AI Assistant to the Judge
This upgrade, KDDI will starfire long thinking chain depth reasoning technology, laws and regulations and judicial interpretations and other professional knowledge, legal experts thinking chain depth integration.It is claimed that it not only solves the problem of knowledge illusion that exists in the traditional big model, but also improves the legal reasoning ability required in complex scenarios such as evidence review, sentencing assistance, and summarization of adjudication opinions.
For example, in the summary scene of the decision point of view, summarize the facts of the case according to the file materials, analyze and compare the facts with the matching laws, regulations and cases one by one, and visualize the process, and output the decision point of view with legal basis to provide a basis for decision-making in handling the case.
At present, Starfire's large model capability has been through the case analysis, evidence review, trial outline auxiliary generation, adjudication assistance and other aspects.
