crown / artificium

Here is the rewritten text:

AGI-0 Labs’ Reflection Model (Created by agi0labs)

This model was created over 22 days ago using a similar approach to mattshumer/Reflection-Llama-3.1-70B (see commit date).

Due to limited resources, I couldn’t run benchmarks, so I set the project aside. However, after the release of Reflection-Llama-3.1-70B, I decided to publish it anyway since provides very similar results (examples of output below).

This research lab is run by one person, and the model was trained for free using a Colab T4 GPU. If you’d like to support with donations or GPUs, feel free to contact me in the community section.

Also, if you’re able to run benchmarks or other tests, please share the results.

Will release the dataset soon.

X: https://x.com/agi0labs

Consider sharing this model if you find it interesting.

Examples (model 8bit quantized in these examples): Test 1: Prompt: How many r in raspberry, what is the number of letters in the capital of italy, which number is bigger 9.9099 or 9.99

Response:

…

(Note: The rest of the response remains the same)