---
title: "GPT, Gemini, and Claude All Pick the Same Right Answers"
subtitle: "A new study of 270 puzzles with multiple correct answers finds AI models cluster tightly together and diverge sharply from how humans pick solutions."
url: "https://alphasignal.ai/news/gpt-gemini-and-claude-all-pick-the-same-right-answers"
author: "AlphaSignal Newsroom"
author_url: "https://alphasignal.ai/editorial-team"
company: "AlphaSignal"
company_url: "https://alphasignal.ai/company/alphasignal"
published: "2026-09-28T04:03:09.000Z"
updated: "2026-09-30T12:08:36.622Z"
type: "Paper"
topics:
  - "Reasoning"
  - "Benchmarks"
subtopics:
  - "Test Time Compute"
  - "Math Reasoning"
upvotes: 111
source_url: "https://arxiv.org/abs/2609.34258"
comments: 0
truncated: true
---

# GPT, Gemini, and Claude All Pick the Same Right Answers

> A new study of 270 puzzles with multiple correct answers finds AI models cluster tightly together and diverge sharply from how humans pick solutions.

# AI Models Converge on the Same Valid Answers

A [new study](https://arxiv.org/abs/2609.34258) by Zihao Wang, Francesco Insulla, and Andrea Montanari finds that frontier AI models cluster around the same valid solutions to multi-answer reasoning puzzles, while people distribute their choices more broadly. Across 270 puzzles, models from different vendors resembled one another more than they resembled humans.

## Accuracy conceals solution preferences

A solver can achieve perfect accuracy while repeatedly choosing only a small subset of the available correct answers. Sampling a model many times, or aggregating answers from many people, produces a distribution that reveals which solutions dominate, which receive little attention, and how broadly each group explores the valid answer space.

That distribution matters for code suggestions, synthetic data, creative generation, policy drafts, red-teaming, and other tasks with several defensible outputs. Standard accuracy scores record whether an answer is correct. They leave shared preferences and concentrated output patterns unmeasured.

## Five puzzle families, 270 controlled tests

The benchmark contains 270 puzzles drawn from five families:

* Arithmetic
* Maze
* Rooks
* Minesweeper
* Sudoku

![Examples of arithmetic, maze, rooks, Minesweeper, and Sudoku puzzles](https://imagedelivery.net/kokADdxuw1FCXyR2p8D2CQ/ef23e10a-a1e9-4346-c348-b5a418654e00/public) 

Examples from the five puzzle families used in the study.

Each puzzle has between three and eight valid solutions. The low difficulty keeps failure rates down, allowing the analysis to focus on preferences among correct answers. The researchers collected responses from human participants and repeated samples from three frontier model families represented by GPT, Gemini, and Claude systems.

_Preview only. Full text requires an AlphaSignal Pro account._

## Sources

- [Investigating Human-AI Discrepancies via Multiple-Solution Problems](https://arxiv.org/abs/2609.34258)
- [HAI Discrepancies project site](https://hai-discrepancies.github.io/)
- [HAI Discrepancies GitHub repository](https://github.com/franinsu/haidiscrepancies)