Search papers, labs, and topics across Lattice.
University of Texas at Dallas
2
0
3
Reasoning-tuned LLMs mirror human struggles with obfuscated code, revealing critical insights into model design and comprehension limits.
Code obfuscation doesn't always make things harder for humans: certain renaming techniques in Python can actually *improve* program comprehension compared to the original code.
