When Many Answers Are Valid, Voting Fails: Symbolic Verification for Best-of-K Causal Reasoning in LLMs

  • 类型：arxiv
  • 标识：2608.03506
  • 链接：https://arxiv.org/abs/2608.03506
  • 主分类：engineering
  • 形态：method
  • 被引：0
  • 被引来源：Semantic Scholar + OpenAlex
  • S2被引：0
  • OpenAlex被引：0
  • 影响力被引：0
  • TLDR：This work introduces CALVER (Causal Axiom-Level VERification), a training-free symbolic verifier that scores structured traces against Pearl's causal criteria, including -separation, backdoor adjustment, and intervention, and selects the highest-scoring candidate without consulting a reference answer.
  • OpenAlex ID：W7172513099
  • OpenAlex DOI：10.48550/arxiv.2608.03506
  • DOI：10.48550/arxiv.2608.03506
  • DOI来源：OpenAlex
  • 开放获取：green
  • 开放获取链接：https://doi.org/10.48550/arxiv.2608.03506
  • OpenAlex更新：2026-08-29
  • 待LLM分类：否
  • 标题中文：当多个答案均有效时，投票失效：面向 LLM Best-of-K 因果推理的符号验证
  • TLDR中文：提出 CALVER (Causal Axiom-Level VERification)，一种无需训练的对称验证器，根据 Pearl 的因果准则（包括 d-分离、backdoor 调整与干预）对结构化 trace 评分，并在不参考标准答案的情况下选择得分最高的候选。
  • 来源文件：
  • /inbox/tom/_candidates/2026-08-06-agent-rag-longcontext-candidates.json
  • [S2 enrich]
  • [OpenAlex backfill]