Language models can produce plausible short proofs, but may still be unreliable on long-horizon research problems, where progress depends on a sequence of un…
机构:Google
来源:arXiv 2609.15983 | AI4Papers 论文推荐平台