@misc{indiciae93edde161cf6,
  title = {Left-Right Symmetry Breaking in CLIP-style Vision-Language Models Trained on Synthetic Spatial-Relation Data},
  author = {Takaki Yamamoto and Chihiro Noguchi and Toshihiro Tanizawa},
  year = {2026},
  url = {https://arxiv.org/abs/2601.12809},
  note = {Source identifier: 2601.12809}
}