Skip to content
TheChatArchiveThe Chat Archive
Geminigemini-2.5-pro-preview-05-06· GoogleHugging Face LMArena Human Preference 140K

can you always see the model name after you choose which answer is better

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with Claudeclaude-opus-4-20250514-thinking-16k’s answer.
  1. User

    can you always see the model name after you choose which answer is better

  2. Gemini Assistant

    No, you generally do not see the model names immediately after you choose which answer is better in a single voting round.

    Here's why and how it works:

    1. Blind Comparison: The core principle of LM Arena is blind pairwise comparison. This means when you are presented with "Model A" and "Model B," you are not supposed to know their actual identities. This prevents your biases about specific models (e.g., "GPT-4 is always better" or "I dislike model X") from influencing your vote. You are purely judging the quality of the two responses.

    2. Protecting Anonymity (Especially for Unreleased Models): If the model names were revealed after each vote, it would:

      • Compromise the anonymity of unreleased or experimental models that developers are testing.
      • Potentially influence your subsequent votes in the same session, even if unintentionally.
    3. What you see:

      • During the voting: "Model A" and "Model B."
      • After voting: You typically just move on to the next pair of anonymous models or your session ends.
    4. When Identities are Revealed:

      • Leaderboard: The identities of publicly known models (or private models whose developers have agreed to a reveal) are shown on the main LM Arena leaderboard. This leaderboard is based on the aggregate Elo ratings derived from thousands of votes over time.
      • Not After Individual Votes: The system doesn't tell you "You just picked Claude 3 Opus over Gemini Pro."

    The goal is to gather unbiased human preference data. Revealing model identities immediately after each vote would undermine this primary goal. You contribute to the overall Elo scores, which then inform the public leaderboard where names are visible.

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