VGRP-Bench/VGRP-Bench

VGPBench: Benchmarking Large Vision Language Models and Fine-Tuning Techniques on Visual Grid Puzzles

EPFL; National and Kapodistrian University of Athens; Meta GenAI; University College London; University of Oxford

Yufan Ren, Konstantinos Tertikas, Shalini Maiti, Junlin Han, Sabine Süsstrunk, Filippos Kokkinos

@article{ren2025vgrp,
  title={VGRP-Bench: Visual Grid Reasoning Puzzle Benchmark for Large Vision-Language Models},
  author={Ren, Yufan and Tertikas, Konstantinos and Maiti, Shalini and Han, Junlin and Zhang, Tong and S{\"u}sstrunk, Sabine and Kokkinos, Filippos},
  journal={arXiv preprint arXiv:2503.23064},
  year={2025}
}

Large Vision-Language Models (LVLMs) struggle with puzzles, which require precise perception, rule comprehension, and logical reasoning. Assessing and enhancing their performance in this domain is crucial, as it reflects their ability to engage in structured reasoning — an essential skill for real-world problem-solving. However, existing benchmarks primarily evaluate pre-trained models without additional training or fine-tuning, often lack a dedicated focus on reasoning, and fail to establish a systematic evaluation framework. To address these limitations, we introduce VGRP-Bench, a Visual Grid Reasoning Puzzle Benchmark featuring 20 diverse puzzles. VGRP-Bench spans multiple difficulty levels, and includes extensive experiments not only on existing chat LVLMs (e.g., GPT-4o), but also on reasoning LVLMs (e.g., Gemini-Thinking). Our results reveal that even the state-of-the-art LVLMs struggle with these puzzles, highlighting fundamental limitations in their puzzle-solving capabilities. Most importantly, through systematic experiments, we identify and analyze key factors influencing LVLMs’ puzzle-solving performance, including the number of clues, grid size, and rule complexity. Furthermore, we explore two Supervised Fine-Tuning (SFT) strategies that can be used in post-training: SFT on solutions (S-SFT) and SFT on synthetic reasoning processes (R-SFT). While both methods significantly improve performance on trained puzzles, they exhibit limited generalization to unseen ones. We will release VGRP-Bench to facilitate further research on LVLMs for complex, real-world problem-solving.

Current supported puzzles (20 in total and 100 puzzles for each):

1. Aquarium (aquarium)
2. Battle-Ship (battleship)
3. Binairo (binairo)
4. Colored-Sudoku (coloredsudoku)
5. Field-Explorer (fieldexplore)
6. Futoshiki (futoshiki)
7. Hitori (hitori)
8. Jigsaw-Sudoku (jigsawsudoku)
9. Kakurasu (kakurasu)
10. Kakuro (kakuro)
11. Killer-Sudoku (killersudoku)
12. Light-Up (lightup)
13. Nonogram (nonogram)
14. Odd-Even-Sudoku (oddevensudoku)
15. Renzoku (renzoku)
16. Skyscraper (skyscraper)
17. Star-Battle (startbattle)
18. Sudoku (sudoku)
19. Thermometers (thermometers)
20. Trees-and-Tents (treesandtents)

Run evaluations for your LVLM

This dataset contains the following entries: image, initialization, prompt, and sample_answer, where sample_answer is not used during evaluation, due to the multi-solution nature, but just serves as a reference.

Please refer to this GitHub repo for detailed instructions.

(Alternatively) Evaluate your model on our benchmark within the VLMEvalKit

We integrated our benchmark to the VLMEvalKit as in this fork. One could use a script like below to run

OPENAI_API_KEY={Your OpenAI API key} python run.py --data VGRPBench --model ChatGPT4o --verbose --api-nproc 16