initial noebook draft wip

[zainhas]

Note

Adds a new LLM judge optimization notebook and a DPO training dataset.

• Evals:
  • Add notebook Evals/Optimizing_LLM_Judges.ipynb for LLM judge optimization experiments.
  • Add DPO training dataset Evals/judge_dpo_data/rewardbench2_dpo_train.jsonl.

^{Written by Cursor Bugbot for commit 0e33a16. This will update automatically on new commits. Configure here.}

[VProv]

Cell that starts as

# pip install together
import json
import os
from together import Together

client = Together()

# Create dataset comparing two model responses
compare_data = [
    {
        "prompt": "Explain photosynthesis",
        "response_a": "Photosynthesis is how plants make food using sunlight.",
        "response_b": "Photosynthesis is the process by which plants convert light energy into chemical energy, using chlorophyll to transform CO2 and water into glucose and oxygen."
    },
]

seems to be redundant