roc-hci/Turing-Bench

Dataset: Turing Test Judge Benchmark

From a paired dialogue (Human-Human vs Human-AI), the task is to predict which dialogue is human-human

What this dataset is

A collection of paired dialogue examples. Each example contains two full dialogue transcripts (A and B):

• One transcript is human-human (two humans talking).
• The other transcript is human-AI (a human talking to an AI).

Task

Binary classification (A/B): Given dialogueA and dialogueB, predict which one is the human-human dialogue.

• Input: dialogueA (string), dialogueB (string)
• Label: who_is_human ∈ {"A", "B"}

Data format (recommended)

Each row (one example) includes:

• id (int): pair ID number
• dialogueA (string): transcript A
• dialogueB (string): transcript B

Leaderboard

To submit predictions, clone the repository, and use file run_eval.py using your own model under the predict() function. Then, submit predictions here: https://huggingface.co/spaces/roc-hci/Turing-Bench-Leaderboard