We introduce Taser (translation testing with Systatic Assistance and consultation), the metric using larger consulting models (LRMS) by testing automatic quality. Taser includes clear consultation skills for LRMS to conduct a formal program, which has steps by step of translation quality. We examine the taser in WMT24 metric allocated to work in both industries based on the reference and indicators, indicating the performance of the situation. In system testing, taser reaches more soft accuracy in both directions based on reference and reference settings, which exceeds all the Metrics. At section level, taser keeps competitive performance with our unique characters – different different as the improved metric between all the patterns. Our exams indicate that organizational encouraging templates produces high results with LRMS in comparison with open-based opening methods. We examine O3, a large model of thinking from Opena, through various efforts, which gives understanding to the relationship between the intensified consultation and assessment quality. The clear discussion process in LRMS offers interpreting and visibility, addressing the key number of metrics is quickly automatic. Our results indicate that the largest consultation models show the improvement that can be measured in quality examination, including advanced accuracy with obvious tests for each test test.
- 30 University of California, Berkeley
- ** Work done while in Apple
