vTikZ, a 100-task benchmark for visual code customization, shows LLMs solve at most 28 percent of TikZ editing scenarios even with best-of-five sampling.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SE 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
LLM Code Customization with Visual Results: A Benchmark on TikZ
vTikZ, a 100-task benchmark for visual code customization, shows LLMs solve at most 28 percent of TikZ editing scenarios even with best-of-five sampling.