Visical Results of LLM Dialogue Synthisis and Sufarization Few of Dialogue

In this project, we propose to strengthening data to the MRDs within the llms to improve the Shot Dialogue shooting work. Unlike previous methods that require external information, we intensify the numbers of the LLM discussion and subarization of power, allowing them to accompany one another during training and to improve full-time training. Dialogue Synthesis strength developed with good guidance by finding points available from summarizing. The strength of summarizing is developed by additional summers of summary summaries produced by the ability to integrate the discussion. By installing the proposed Mahds Mechanism, we update the Internal LLM information in the data format, and use it to add real training data. The empirical results indicate that our way is to improve discussion summarizing, achieving 1.5% increase in Rouge Scores and improvement of 0.3% in the Bert scores in a few shots of shooting. In addition, our way detects highest scores in human examination, exceeds both front-trained models and good foundations only organized only summarizing.