PuzzleWorld: 667 Real-World Puzzlehunt Problems for AI Reasoning

667 real-world puzzlehunt-style problems curated from Puzzled Pint's Creative Commons archives between 2010 and 2025. The dataset, created by author hzli1202 and last updated on June 10, 2025, is designed as a benchmark to evaluate open-ended, multimodal reasoning in AI models.

Use Cases

• Benchmarking AI models on open-ended reasoning based on ambiguous, multimodal clues.
• Evaluating multimodal inference capabilities based on puzzles combining text, visual, and structured inputs.
• Training models for creative problem-solving based on puzzles with no explicitly stated instructions.
• Researching human-like reasoning chains based on multi-step puzzle solutions.

Strengths

• Contains 667 distinct, real-world problems, providing substantial evaluation material.
• Sourced from a specific, documented archive (Puzzled Pint) with a 15-year time range (2010-2025).
• Designed for a specific, advanced AI evaluation task (open-ended, multimodal reasoning).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known, but dataset size, file formats, and sample data are unknown.
• Data may reflect the thematic and stylistic bias inherent to its single source (Puzzled Pint).

Provenance

Source

Puzzled Pint's Creative Commons–licensed archives.

Collection Method

Curated from the source archives.

Time Range

2010–2025

Freshness

Last updated 2025-06-10 00:56:37; freshness should be verified.

Geography

null