Dataset Overview

Dataset Introduction

The dataset was automatically created for evaluating the model Danielbrdz/Barcenas‑Tiny‑1.1b‑DPO on the Open LLM Leaderboard.

Structure

• Number of Configurations: 63, each corresponding to a specific evaluation task.
• Creation Source: Generated from a single run. Each run yields a split named with the run timestamp; the "train" split always points to the most recent results.
• Additional Configuration: The "results" config stores aggregated results of all runs for computing and displaying metrics on the Open LLM Leaderboard.

Loading Example

from datasets import load_dataset

data = load_dataset(
    "open-llm-leaderboard/details_Danielbrdz__Barcenas-Tiny-1.1b-DPO",
    "harness_winogrande_5",
    split="train"
)

Latest Results

The latest run on 2024‑01‑20T20:17:56.012496 produced the following scores:

{
    "all": {"acc": 0.2555, "acc_stderr": 0.0307, "acc_norm": 0.2564, "acc_norm_stderr": 0.0315, ...},
    "harness|arc:challenge|25": {"acc": 0.3464, "acc_stderr": 0.0139, "acc_norm": 0.3626, "acc_norm_stderr": 0.0140},
    "harness|hellaswag|10": {"acc": 0.4566, "acc_stderr": 0.0050, "acc_norm": 0.6120, "acc_norm_stderr": 0.0049},
    ...
}