Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

1. The graphs are not entirely clear: the names of the x and y axes. Why the number of groups differs
2. There is no explanation for these five rules (17-18 p.)
3. There is no explanation of how other researchers interpret and measure the value of Adaptive learning. There is no evidence that adaptive learning was performed.
4. There are no citations in the text of tables and figures.
5. Augmentation and validation flows should be more precisely described in Figures 2-6. What are the input/output/preconditions between processes?

Author Response

Comments 1: 

The graphs are not entirely clear: the names of the x and y axes. Why the number of groups differs

Response 1: 

Thank you for pointing this out. We agree with this comment. Therefore, we have modified the related diagrams based on the feedback(Figures 6,7,8,9). Regarding the question of the different number of groups, we divide the different test sets based on the different purposes of the experiments. More details about the tested groups in the experiment can be found in Appendix A. Additional discussion on the selection principles of knowledge domains in the experiment can be found on lines 500-502 and lines 543-549.

Comments 2:

There is no explanation for these five rules (17-18 p.)

Response 2:

Thank you for pointing this out. We agree with this comment. Therefore, we have added the corresponding explanation for each of rules. This change can be found on pages 19-20 from lines 660-709.

Comments 3:

There is no explanation of how other researchers interpret and measure the value of Adaptive learning. There is no evidence that adaptive learning was performed.

Response 3:

We appreciate the reviewer’s observation and agree that the discussion of adaptive learning needed further elaboration. In the revised manuscript, we have added a dedicated description to related work section to clarify how adaptive learning is commonly defined, interpreted, and evaluated in existing literature. This change can be found in section 2.4 (lines 227-245).

Additionally, we acknowledge that our original manuscript did not sufficiently demonstrate how adaptive learning was improved in our experiments. We have now explicitly described how our approach facilitated adaptive learning in section 4.2(lines 660-709), where we detail how the LLM was optimized by using customized knowledge update and personal profiles to improve its performance during the learning process. We also include performance evaluation before and after the integration of new knowledge, serving as evidence of adaptive behavior(lines 726-748). We hope these clarifications sufficiently address the reviewer’s concerns.

Comments 4:

There are no citations in the text of tables and figures.

Response 4:

Thank you for pointing this out. We have accordingly modified these issues in our revised manuscript.

Comments 5:

Augmentation and validation flows should be more precisely described in Figures 2-6. What are the input/output/preconditions between processes?

Response 5:

We thank the reviewer for this insightful comment. In the revised manuscript, we have improved the clarity and precision of Figures 2 through 6 to better illustrate the augmentation and validation workflows. We added the dedicated description explaining the input/output /preconditions for each corresponding step as suggested. This change can be found on lines 319-322, 328-331,371-377 and 453-456.

Reviewer 2 Report

Comments and Suggestions for Authors

The document is written in a way that clearly shows how the framework was built. It justifies in a proper manner the incremental approach that led to the integration of the elements outlined in Fig. 1, which perfectly summarizes the framework.

There are some improvements that should be made:

• In the related work section, it seems like more quantitative results of the referenced works should be included. This would set a type of benchmark to compare the results obtained. What is currently included in this section contributes to understand the proposal, but are mostly conclusions and learned lessons from those other works.
• Please review figures 2 and 3 (especially figure 2). There seems to be more concepts and relationships in the paragraphs describing the processes depicted in those figures than what is actually shown in the figures.
• In the results section, please include axis titles and units in the graphs. 
• I would recommend being more specific about the knowledge domains. The first test makes reference to 9 domains, the second to 31, and the third to 11. However, these domains are never mentioned (though is understood they are STEAM related) until Table 1, but looking at the results in this table, the order in which they are presented is not the same as in figures 8 and 9. In addition, further discussion on the domains with better/worst performance would greatly enhance the document.
• In the references section, references 11 and 12 seem to be the same document. Please also include additional data for reference 61.
• Is the thext on line 486 a section title?

 

Author Response

Comments 1:

In the related work section, it seems like more quantitative results of the referenced works should be included. This would set a type of benchmark to compare the results obtained. What is currently included in this section contributes to understand the proposal, but are mostly conclusions and learned lessons from those other works.

Response 1:

Thank you for this valuable suggestion. We agree that including more quantitative results from prior studies would provide a clearer benchmark and strengthen the contextual foundation for our work. In response, we have revised the related work section to incorporate specific evaluation metrics, experimental settings, and corresponding discussion from key referenced studies. This change can be found in sections 2.3(lines 178-192) and 2.4(lines 227-245). We believe this addition improves the utility and clarity of the manuscript, and we thank the reviewer again for the helpful feedback.

Comments 2:

Please review figures 2 and 3 (especially figure 2). There seems to be more concepts and relationships in the paragraphs describing the processes depicted in those figures than what is actually shown in the figures.

Response 2:

Thank you for pointing this out. The diagrams represent a summary of general examples. In the discussion section of the article, we include more details and possible complex situations in the technical implementation. For different specific situations, such as the segmentation of large documents and the switching of different document formats, the system will assign different underlying agents to solve them. To keep it simple, we skipped these specific sub-steps or internal operations of agents in the diagrams of the overall introduction.

Comments 3:

In the results section, please include axis titles and units in the graphs. 

Response 3:

Thank you for pointing this out. We have accordingly revised the related diagrams in our revised manuscript. (Figure 6,7,8,9,10)

Comments 4:

I would recommend being more specific about the knowledge domains. The first test makes reference to 9 domains, the second to 31, and the third to 11. However, these domains are never mentioned (though is understood they are STEAM related) until Table 1, but looking at the results in this table, the order in which they are presented is not the same as in figures 8 and 9. In addition, further discussion on the domains with better/worst performance would greatly enhance the document.

Response 4:

We appreciate the reviewer’s suggestion to clarify and discuss the knowledge domains in greater detail. To address this:

1. We added more details about the tested domains in the experiment to Appendix A.

2. We added the additional discussion on the selection principles of knowledge domains in the experiment and the possible effects of knowledge domains and LLM performance. It also includes the limitations of the current research and the related future work. This change can be found on lines 543-549 and lines 582-603.

The domains discussed in section 4.1.1 are presented in descending order of absolute performance improvement, as measured by F1 score, following knowledge enhancement.

Comments 5:

In the references section, references 11 and 12 seem to be the same document. Please also include additional data for reference 61.

Response 5:

Thank you for pointing this out. We have accordingly modified these issues in our revised manuscript.

Comments 6:

Is the thext on line 486 a section title?

Response 6:

Thank you for pointing this out. We have accordingly changed the title to the subsection title in our revised manuscript. (on line 526)