Accepted papers - [NeurIPS’26-a] (SELF) on why RLVR reduces reasoning coverage of LLMs, [NeurIPS’26-b] on how primacy bias affects learnability in RLVR training, [NeurIPS’26-c] on controlling the hidden critique mechanism in LRMs, and [NeurIPS’26-d] (ChemBack) on molecular graph backdoors under chemistry-aware admission.