Skip to content
TheChatArchiveThe Chat Archive
Metallama-2-70b-chat· MetaHugging Face Arena Human Preference 55K

Given the following question, identify all potential concepts (nouns) and relationships…

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with gpt-3.5-turbo-0613’s answer.
  1. User

    Given the following question, identify all potential concepts (nouns) and relationships (verbs), and entities (individuals), referenced in the question. If none are found, the empty set is acceptable. Proper nouns such as names are considered entities. All concepts, relationships, and entities must be present in the given question. Do not include any terms from the following examples unless that term also occurs in the question.

    Return the results in JSON using the following format: { concepts: , relationships: , entities: }

    Some example outputs:

    Question: Which presidents were born in California? Output: { concepts: ["president"], relationships: ["born in"], entities: ["California"] }

    Question: List cars manufactured by Ford with manual transmission and moonroofs or sunroofs. Output: { concepts: ["car", "manual transmission", "moonroof", "sunroof"], relationships: ["manufactured by"], entities: ["Ford"] }

    Question: Find carnivore mammals living in mountains and preying on rodents. Output: { concepts: ["carnivore", "mammal", "mountain", "rodent"], relationships: ["live in", "prey on"], entities: [] }

    Question: Who won most swimming medals in olympics in Japan? Output: { concepts: ["swimming", "medal", "olympics"], relationships: ["won"], entities: ["Japan"] }

    Question: Show me houses with a lord who is female Output:

  2. Meta Assistant

    { "concepts": ["house", "lord"], "relationships": ["has a"], "entities": ["female"] }

    The question is asking for houses that have a female lord. The concepts are "house" and "lord", the relationship is "has a", and the entity is "female".

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.