Skip to main content

Table 1 Codes of the 45 responses from 15 chats, each generated three times using ChatGPT models 3.5 and 4

From: Evaluating the authenticity of ChatGPT responses: a study on text-matching capabilities

 

First Response

Regeneration 1

Regeneration 2

Chat Process 1 (model GPT 3.5)

ChatGPT 1–1

ChatGPT 1–2

ChatGPT 1–3

Chat Process 2 (model GPT 3.5)

ChatGPT 2–1

ChatGPT 2–2

ChatGPT 2–3

Chat Process 3 (model GPT 3.5)

ChatGPT 3–1

ChatGPT 3–2

ChatGPT 3–3

Chat Process 4 (model GPT 3.5)

ChatGPT 4–1

ChatGPT 4–2

ChatGPT 4–3

Chat Process 5 (model GPT 3.5)

ChatGPT 5–1

ChatGPT 5–2

ChatGPT 5–3

Chat Process 6 (model GPT 3.5)

ChatGPT 6–1

ChatGPT 6–2

ChatGPT 6–3

Chat Process 7 (model GPT 3.5)

ChatGPT 7–1

ChatGPT 7–2

ChatGPT 7–3

Chat Process 8 (model GPT 3.5)

ChatGPT 8–1

ChatGPT 8–2

ChatGPT 8–3

Chat Process 9 (model GPT 3.5)

ChatGPT 9–1

ChatGPT 9–2

ChatGPT 9–3

Chat Process 10 (model GPT 3.5)

ChatGPT 10–1

ChatGPT 10–2

ChatGPT 10–3

Chat Process 11 (model GPT 4)

ChatGPT 11–1

ChatGPT 11–2

ChatGPT 11–3

Chat Process 12 (model GPT 4)

ChatGPT 12–1

ChatGPT 12–2

ChatGPT 12–3

Chat Process 13 (model GPT 4)

ChatGPT 13–1

ChatGPT 13–2

ChatGPT 13–3

Chat Process 14 (model GPT 4)

ChatGPT 14–1

ChatGPT 14–2

ChatGPT 14–3

Chat Process 15 (model GPT 4)

ChatGPT 15–1

ChatGPT 15–2

ChatGPT 15–3