TY  - RPRT
TI  - Bidirectional Preference Synthesis: Learning Prompt-Conditioned Preferences from Boundary Failures
AU  - Junbo Wang
AU  - Lidong Lu
AU  - Zhuoqun Li
AU  - Guiping Jiang
AU  - Xiangyu Wu
AU  - Tinghai Zhang
AU  - Tong Lu
PY  - 2026
UR  - https://arxiv.org/abs/2610.04328
ID  - 2610.04328
ER  -