Figure 6
The top row is labeled “(a) Comparison of different L L Ms” and consists of three vertical bar graphs arranged from left to right. All three graphs in this row share a horizontal axis with four categories, labeled from left to right as follows: “DeepSeek-V 3.2”, “Qwen 3-Maximum”, “K i M i-K 2”, and “G L M-4.6”. Left graph: The vertical axis is labeled “Answer Correctness ” and ranges from 0.60 to 0.85. in increments of 0.05 units. The data is as follows: DeepSeek-V 3.2: 0.77. Qwen 3-Maximum: 0.75. K i M i-K 2: 0.72. G L M-4.6: 0.73. Middle graph: The vertical axis is labeled “Answer Relevancy” and ranges from 0.60 to 1.00 in increments of 0.05 units. The data is as follows: DeepSeek-V 3.2: 0.87. Qwen 3-Maximum: 0.85. K i M i-K 2: 0.84. G L M-4.6: 0.82. Right graph: The vertical axis is labeled “time(seconds)” and ranges from 0 to 20 units in increments of 5 units. The data is as follows: DeepSeek-V 3.2: 3.5. Qwen 3-Maximum: 3.2. K i M i-K 2: 20.0. G L M-4.6: 2.8. The bottom row is labeled “(b) Comparison of DeepSeek-R 1 Models with Different Parameter Scales” and consists of three vertical bar graphs arranged from left to right. All three graphs in this row share a horizontal axis with five categories, labeled from left to right as follows: “671 B”, “32 B”, “14 B”, “7 B”, and “1.5 B”. Left graph: The vertical axis is labeled “Answer Correctness” and ranges from 0.40 to 0.80 in increments of 0.05 units. The data is as follows: 671 B: 0.73. 32 B: 0.70. 14 B: 0.66. 7 B: 0.58. 1.5 B: 0.54. Middle graph: The vertical axis is labeled “Answer Relevancy” and ranges from 0.40 to 0.90 in increments of 0.05 units. The data is as follows: 671 B: 0.83. 32 B: 0.80. 14 B: 0.69. 7 B: 0.62. 1.5 B: 0.58. Right graph: The vertical axis is labeled “time(seconds)” and ranges from 0 to 10 units in increments of 2 units. The data is as follows: 671 B: 9.6. 32 B: 5.8. 14 B: 5.7. 7 B: 3.2. 1.5 B: 1.3. Note: All numerical data values are approximated.LLM comparative study