id string | original_id string | mr3_source string | language string | actual_score string | gpt-oss-prompt_en_prompt_en_thinking string | gpt-oss-120b-en_prompt_en_thinking-reasoning string | gpt-oss-120b-en_prompt_en_thinking-response string | gpt-oss-prompt_tgt_prompt_en_thinking string | gpt-oss-120b-tgt_prompt_en_thinking-reasoning string | gpt-oss-120b-tgt_prompt_en_thinking-response string | gpt-oss-prompt_tgt_prompt_tgt_thinking string | gpt-oss-120b-tgt_prompt_tgt_thinking-reasoning string | gpt-oss-120b-tgt_prompt_tgt_thinking-response string |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
arena-human-preference-c4b9710c-8d64-4bee-a0b0-94637ae4cc65 | c4b9710c-8d64-4bee-a0b0-94637ae4cc65 | arena-human-preference | en | Assistant A | <|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.
Knowledge cutoff: 2024-06
Current date: 2025-09-04
Reasoning: high
# Valid channels: analysis, commentary, final. Channel must be included for every message.<|end|><|start|>developer<|message|># Instruction
You are an impartial j... | We need to decide which response (Assistant A vs B) is better for the user prompt: "Compare Tormenta20 with DnD5e". The user wants a comparison. Both assistants provided information, but we need to evaluate based on correctness, helpfulness, relevance, conciseness, creativity, coverage.
We must systematically evaluate... | {
"explanation": "Assistant A offers a thorough, well‑structured comparison that accurately covers philosophy, core mechanics, character creation, magic, combat, and setting integration, even providing a summary table and recommendations. Its information is correct and highly relevant to the user’s request. Assistant... | <|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.
Knowledge cutoff: 2024-06
Current date: 2025-09-04
Reasoning: high
# Valid channels: analysis, commentary, final. Channel must be included for every message.<|end|><|start|>developer<|message|># Instruction
You are an impartial j... | We need to decide which response (Assistant A vs B) is better for the user prompt: "Compare Tormenta20 with DnD5e". The user wants a comparison. Assistant A provides a long, structured, detailed comparison covering many aspects: high-level philosophy, core mechanic, character creation, magic system, combat and power le... | {
"explanation": "Assistant A gives a thorough, well‑structured comparison covering philosophy, core mechanics, character creation, magic, combat, power curve, setting integration, a summary table, and recommendations. Its information is accurate and directly answers the prompt. Assistant B is much shorter, omits maj... | <|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.
Knowledge cutoff: 2024-06
Current date: 2025-08-28
Reasoning: high
# Valid channels: analysis, commentary, final. Channel must be included for every message.
# Valid language: English. English must be the only language used in every me... | Okay, I have to think explicitly and provide my answer in English. I will carefully examine all provided information and evaluate it according to the given rubric, then respond in the required format. Let's restate the task.
The user is giving a prompt: "Compare Tormenta20 with DnD5e", then showing two assistant respo... | {
"explanation": "Both answers are factually accurate, but Assistant A offers a much richer, structured comparison that directly addresses the user's request. It covers philosophy, core mechanics, character creation, magic, combat, setting integration, and even provides a summary table and recommendations, making it ... |
arena-human-preference-40f981ca-cca6-4e2a-a48c-d10d51884efe | 40f981ca-cca6-4e2a-a48c-d10d51884efe | arena-human-preference | en | Assistant A | <|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.
Knowledge cutoff: 2024-06
Current date: 2025-09-04
Reasoning: high
# Valid channels: analysis, commentary, final. Channel must be included for every message.<|end|><|start|>developer<|message|># Instruction
You are an impartial j... | We need to decide which response is better according to the rubric. The question: "What are Tricky the Clown's forms?" This refers to the character Tricky the Clown from Madness Combat and the FNF (Friday Night Funkin') mod. Assistant A gave a well-researched thorough answer covering both sources, describing four forms... | {
"explanation": "Assistant A accurately identifies Tricky the Clown’s various forms from both Madness Combat and the Friday Night Funkin' Tricky mod, providing correct details, clear organization, and a concise summary. Assistant B misinterprets the character, offers unrelated speculation about other clowns, and ask... | <|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.
Knowledge cutoff: 2024-06
Current date: 2025-09-04
Reasoning: high
# Valid channels: analysis, commentary, final. Channel must be included for every message.<|end|><|start|>developer<|message|># Instruction
You are an impartial j... | We need to decide which response is better according to the rubric. The question: "What are Tricky the Clown's forms?" This refers to the character Tricky the Clown from Madness Combat and the FNF (Friday Night Funkin') mod. Assistant A gave a well-researched thorough answer covering both sources, describing four forms... | {
"explanation": "Assistant A correctly identifies and describes the specific forms of Tricky the Clown from Madness Combat and the Friday Night Funkin' mod, providing accurate details and relevant context. Assistant B mistakenly claims the character is not known, offers unrelated alternatives, and fails to answer th... | <|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.
Knowledge cutoff: 2024-06
Current date: 2025-08-28
Reasoning: high
# Valid channels: analysis, commentary, final. Channel must be included for every message.
# Valid language: English. English must be the only language used in every me... | Okay, I have to think explicitly and provide my answer in English. I will carefully examine all provided information and evaluate it according to the given rubric, then respond in the required format. Let's restate the task.
The user is asking: "What are Tricky the Clown's forms?" The user has presumably asked the ass... | {
"explanation": "Assistant A directly lists Tricky the Clown's known forms from Madness Combat and the Friday Night Funkin' mod, providing accurate descriptions and relevant details. Assistant B misidentifies the character, offers unrelated speculation, and fails to answer the question. Therefore A is correct, helpf... |
arena-human-preference-769564e7-4ec0-43ef-906a-a62f5f959a1a | 769564e7-4ec0-43ef-906a-a62f5f959a1a | arena-human-preference | zh | Assistant A | <|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.
Knowledge cutoff: 2024-06
Current date: 2025-09-04
Reasoning: high
# Valid channels: analysis, commentary, final. Channel must be included for every message.<|end|><|start|>developer<|message|># Instruction
You are an impartial j... | We need to decide which response is better according to the evaluation rubric. Let's go through each factor.
First, we need to understand the user's prompt:
"学生的答题卡上的涂卡如何识别成文字的。。" (which means: "How to recognize the filled bubbles on the student's answer sheet into text?").
So the user wants to know how the filled ... | {
"explanation": "Assistant A provides a fully accurate description of the OMR process, correctly distinguishing it from OCR and covering every stage from answer‑sheet design to data output, including challenges and extensions. Assistant B mislabels the technology as OCR and offers a less detailed, somewhat incomplet... | <|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.
Knowledge cutoff: 2024-06
Current date: 2025-09-04
Reasoning: high
# Valid channels: analysis, commentary, final. Channel must be included for every message.<|end|><|start|>developer<|message|># 指令
你是一位公正的评审，负责评估哪一个助手的回答对给定的用户提问更... | We need to decide which response (Assistant A vs B) is better based on criteria: correctness/accuracy, usefulness (does it directly answer user request and follow instructions), relevance, conciseness, creativity/novelty, coverage.
User question: "学生的答题卡上的涂卡如何识别成文字的。。" Rough translation: "How to recognize the student'... | {
"explanation": "Assistant A 正确地将涂卡识别描述为光学标记识别(OMR)，并详细说明了答题卡设计、图像获取、预处理、区域定位、特征提取、阈值判断以及数据输出等关键步骤，还讨论了技术挑战、OMR 与 OCR 的区别以及实际应用，信息完整且准确。Assistant B 将识别方式误称为 OCR，存在概念错误，且内容相对简略，未提供对 OMR 与 OCR 的区分或更深层次的技术细节。整体来看，A 在正确性、覆盖面和有用性上明显优于 B，虽然稍显冗长，但仍然结构清晰、信息丰富。",
"score": "Assistant A"
} | <|start|>system<|message|>你是 ChatGPT，一个由 OpenAI 训练的大型语言模型。
知识最新更新时间：2024-06
当前日期：2025-08-28
推理水平：高
# 合法通道：analysis、commentary、final。每条消息必须包含通道信息。
# 合法语言：中文。每条消息只能使用中文。<|end|><|start|>developer<|message|># 指令
你是一位公正的评审，负责评估哪一个助手的回答对给定的用户提问更好。请按照以下因素的顺序系统地比较这两个回答：
- 正确性与准确性：识别任何错误、不准确或具有误导性的信息。
- 有用性：判断回答是否直接回应了用户的请... | 好的，我需要明确地思考，并用中文给出我的答案。我会仔细审视所有提供的信息，按照给定的评分标准进行评估，然后以要求的格式作答。以下是思考过程.
## 任务概述
我们被赋予以下任务：
- 作为评审，比较两位助理的回答（A 和 B），判断哪一个更好。
- 基于给定的因素顺序：**正确性与准确性 → 有用性 → 相关性 → 简明性 → 创造力与新颖性 → 覆盖面**，系统地进行比较。
- 然后给出简短理由，并依据评分标准输出结论，选择 **Assistant A、Assistant B、Tie、Both Bad** 四个之一。
- 必须返回 JSON 对象，属性 explanation 为解释文字，score 为裁决结果，两个字段都是... | {
"explanation": "Assistant A 对涂卡识别的原理描述完整且准确，使用了正确的术语 OMR，涵盖了答题卡设计、图像采集、预处理、定位、特征提取、数据输出以及技术挑战等细节，信息覆盖面广。Assistant B 虽然结构简洁，但把涂卡识别称为 OCR，这是概念上的错误，可能导致误导。按照先考虑正确性再看有用性、覆盖面的评估顺序，Assistant A 的回答整体质量更高。",
"score": "Assistant A"
} |
arena-human-preference-e67468c9-69b7-4b84-9ec2-27879128c4c6 | e67468c9-69b7-4b84-9ec2-27879128c4c6 | arena-human-preference | en | Assistant B | "<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.\nKnowledge cut(...TRUNCATED) | "We need to decide which response is better according to the rubric. The question: The user asked: \(...TRUNCATED) | "{\n \"explanation\": \"Both answers are factually correct and stay on topic, but Assistant B pro(...TRUNCATED) | "<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.\nKnowledge cut(...TRUNCATED) | "We need to decide which response (Assistant A vs B) is better based on criteria: Correctness & Accu(...TRUNCATED) | "{\n \"explanation\": \"Both answers are factually correct and stay on topic. Assistant B, however,(...TRUNCATED) | "<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.\nKnowledge cut(...TRUNCATED) | "Okay, I have to think explicitly and provide my answer in English. I will carefully examine all pro(...TRUNCATED) | "{\n \"explanation\": \"Both answers are factually correct, but Assistant B provides more thorough (...TRUNCATED) |
arena-human-preference-821f691b-d4a9-4803-a315-a4fa31bc71a2 | 821f691b-d4a9-4803-a315-a4fa31bc71a2 | arena-human-preference | ru | Assistant B | "<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.\nKnowledge cut(...TRUNCATED) | "We need to decide which response (Assistant A vs B) is better for the user prompt: \"Сделай (...TRUNCATED) | "{\n \"explanation\": \"Both answers are factually correct, but Assistant B provides a far more com(...TRUNCATED) | "<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.\nKnowledge cut(...TRUNCATED) | "We need to decide: compare Assistant A and Assistant B responses. The question: \"Сделай до(...TRUNCATED) | "{\n \"explanation\": \"Оба ответа содержат в целом правильную и(...TRUNCATED) | "<|start|>system<|message|>Вы — ChatGPT, большая языковая модель, обуч(...TRUNCATED) | "Хорошо, мне нужно думать вслух и дать свой ответ на рус(...TRUNCATED) | "{\n \"explanation\": \"Оба доклада содержат правильную информа(...TRUNCATED) |
arena-human-preference-b0333c20-7fad-45b6-9581-bb2784fdd23b | b0333c20-7fad-45b6-9581-bb2784fdd23b | arena-human-preference | en | Assistant A | "<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.\nKnowledge cut(...TRUNCATED) | "We need to decide which response is better according to the evaluation rubric: correct, helpful, re(...TRUNCATED) | "{\n \"explanation\": \"Both responses correctly show how to use the `-c` option to run Python code(...TRUNCATED) | "<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.\nKnowledge cut(...TRUNCATED) | "We need to decide which response is better according to the evaluation rubric. Let's go through eac(...TRUNCATED) | "{\n \"explanation\": \"Both answers correctly show that you can run Python code in one line with t(...TRUNCATED) | "<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.\nKnowledge cut(...TRUNCATED) | "Okay, I have to think explicitly and provide my answer in English. I will carefully examine all pro(...TRUNCATED) | "{\n \"explanation\": \"Both responses correctly state that Python code can be executed in one line(...TRUNCATED) |
arena-human-preference-8c86f9ea-8b8d-4ef0-8def-94352adf603e | 8c86f9ea-8b8d-4ef0-8def-94352adf603e | arena-human-preference | es | Assistant B | "<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.\nKnowledge cut(...TRUNCATED) | "We need to decide which response is better according to the rubric. The question: \"Cuál es el mej(...TRUNCATED) | "{\n \"explanation\": \"Both answers are factually correct, but Assistant B is more helpful and c(...TRUNCATED) | "<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.\nKnowledge cut(...TRUNCATED) | "We need to decide: compare Assistant A and Assistant B responses. The question: \"Cuál es el mejor(...TRUNCATED) | "{\n \"explanation\": \"Ambas respuestas son correctas y precisas en la información básica sobre (...TRUNCATED) | "<|start|>system<|message|>Eres ChatGPT, un gran modelo de lenguaje entrenado por OpenAI.\nFecha de (...TRUNCATED) | "De acuerdo, tengo que pensar explícitamente y dar mi respuesta en español. Examinaré cuidadosame(...TRUNCATED) | "{\n \"explanation\": \"Ambos asistentes son correctos en la información básica, pero el Asistent(...TRUNCATED) |
arena-human-preference-6930734a-20cd-421e-bf1a-4db1c3b6ef2d | 6930734a-20cd-421e-bf1a-4db1c3b6ef2d | arena-human-preference | en | Assistant B | "<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.\nKnowledge cut(...TRUNCATED) | "We need to decide which response is better according to the evaluation rubric: correct, helpful, re(...TRUNCATED) | "{\n \"explanation\": \"Assistant B is overall superior. It provides accurate, verifiable product i(...TRUNCATED) | "<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.\nKnowledge cut(...TRUNCATED) | "We need to decide which response is better according to the evaluation rubric. Let's go through eac(...TRUNCATED) | "{\n \"explanation\": \"Assistant B provides a realistic list of actual Nike products with correct (...TRUNCATED) | "<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.\nKnowledge cut(...TRUNCATED) | "Okay, I have to think explicitly and provide my answer in English. I will carefully examine all pro(...TRUNCATED) | "{\n \"explanation\": \"Assistant B provides factual product information, accurate price ranges, an(...TRUNCATED) |
arena-human-preference-e93352c0-7259-4afb-b0ae-d077fca864cb | e93352c0-7259-4afb-b0ae-d077fca864cb | arena-human-preference | en | Assistant B | "<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.\nKnowledge cut(...TRUNCATED) | "We need to decide which response is better according to the rubric. The question: \"Mit AC-Car-Tune(...TRUNCATED) | "{\n \"explanation\": \"Assistant B provides a more accurate answer. It correctly identifies the ae(...TRUNCATED) | "<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.\nKnowledge cut(...TRUNCATED) | "We need to decide which response is better according to the evaluation rubric: correct, helpful, re(...TRUNCATED) | "{\n \"explanation\": \"Assistant B correctly focuses on the files that actually control aerodynami(...TRUNCATED) | "<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.\nKnowledge cut(...TRUNCATED) | "Okay, I have to think explicitly and provide my answer in English. I will carefully examine all pro(...TRUNCATED) | "{\n \"explanation\": \"Assistant B gives a more accurate and relevant answer for increasing the ca(...TRUNCATED) |
arena-human-preference-6736b182-71b2-46f6-8380-d1d01b1b30db | 6736b182-71b2-46f6-8380-d1d01b1b30db | arena-human-preference | en | Assistant B | "<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.\nKnowledge cut(...TRUNCATED) | "We need to decide which response is better according to the rubric. The question: \"What does the f(...TRUNCATED) | "{\n \"explanation\": \"Both answers correctly explain the function's steps, but Assistant B is mor(...TRUNCATED) | "<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.\nKnowledge cut(...TRUNCATED) | "We need to decide which response is better according to the rubric. The question: \"What does the f(...TRUNCATED) | "{\n \"explanation\": \"Both answers explain the function’s behavior, but Assistant B is more acc(...TRUNCATED) | "<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.\nKnowledge cut(...TRUNCATED) | "Okay, I have to think explicitly and provide my answer in English. I will carefully examine all pro(...TRUNCATED) | "{\n \"explanation\": \"Both assistants describe the function correctly, but Assistant B is more pr(...TRUNCATED) |
End of preview. Expand in Data Studio
README.md exists but content is empty.
- Downloads last month
- 5