DBRX-Base @databricks also gets to the top 1 on URIAL Bench, which tests Base LLMs (only…
- User
DBRX-Base @databricks also gets to the top 1 on URIAL Bench, which tests Base LLMs (only pre-trained) on the MT-bench with URIAL prompts (three-shot instruction-following IOs). Check out the full results here:
Check grammar errors
- Assistant
The message is generally clear and correctly conveys the information about DBRX-Base's performance on the URIAL Bench. However, there are a few minor adjustments that can enhance clarity and readability. Here’s a revised version for consideration:
"DBRX-Base from @databricks has achieved the top position in the URIAL Bench rankings. This benchmark tests base-level Large Language Models (LLMs) in their pre-trained state on the MT-bench, using URIAL prompts which involve three-shot instruction-following input-output tasks. For complete results, check out the full details here:"
Note: Since the message ends with an invitation to check out the full results but doesn't provide a link or specific direction on where to find these results, it would be beneficial to include a link or specify where or how to access them for completeness.
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.