2026 · Preprint
GameReplica: A Benchmark for Black-Box Visual Game Replication by Vision-Language Agents
Evaluates whether vision-language agents can infer hidden game rules through screenshots and active interaction, then build runnable single-file HTML replicas across 125 tasks from 25 web games, each with five difficulty levels.
Scope and limitations
The tasks cover rule-driven web puzzles and board games, not native or commercial game reconstruction. Rule, implementation, and visual scores use LLM judges and do not prove behavioral equivalence. The official site provides demos; a benchmark-code repository was not verified.
Why it fits this collection
Directly evaluates active black-box specification recovery and executable game reconstruction, filling the bibliography’s game-replication gap.
Publication and review details
- Venue
- arXiv
- Status
- Preprint
- Publication / posting date
- September 15, 2026
- Date source
- arXiv v1 submission history
- First arXiv submission
- September 15, 2026
- arXiv
- 2609.22308
- Reviewed arXiv version
- v1
- DOI
- 10.48550/arXiv.2609.22308
The official project page marks the work as under review. arXiv reports the v1 submission on 2026-09-15 and displays its arXiv DOI as pending registration. The project link is a showcase with demos, not a verified release of the full evaluation framework.
Sources reviewed
Source review only; experiments were not rerun for this collection. Cite the original authors using the linked paper or BibTeX.