[Bug]: DeepSeek Models are producing Inaccurate generated Answers

[shailajafissionqa]

FloTorch Version

V2.0.3

Bug Description

DeepSeek Models are producing Inaccurate Answers and resulting the Poor Evaluation Metrics

Steps to Reproduce

1. Login to FloTorch.ai
2. Click on Create Project
3. Enter the details of the Project, select the Region
4. Select the Knowledge base type from the drop down
5. Upload the Knowledge Base
6. Upload the Ground Truth json file and click on Next button
7. Select the Chucking, Embedding Model, Vector Dimension and Indexing Algorithm from the drop down and click on Next button
8. Select the N shot Prompts, and upload the N shot prompt file
9. Select the KNN value from the KNN dropdown
10. Select any of the Deepseek model's from Inferencing Model dropdown
11. Select the Inferencing Model Temperature and click on Next
12. Select the Evaluation Service as Ragas, Embedding model and Inferencing Model from the dropdown and click on Submit button
13. Select the Valid experiments and click on Run button
14. Click on Confirm button

Expected Result : Generated Answers should be accurate and should match with the Ground Truth Answers.
Actual Result : Generated Answers are inaccurate and not matching with the Ground Truth Answers.

Browser and Version

Chrome version : 133.0.6943.54

Runtime Version

Version : 3.12

Error Messages

Additional Context

No response