arXiv cs.CL· Tsuyoshi Okita·· 14 小时前AI 评分45
CIKA：用 LLM 作为概念掌握模拟器，通过干预式时间序列因果发现提升数学推理
Mathematical Reasoning via Intervention-Based Time-Series Causal Discovery Using LLMs as Concept Mastery Simulators
AI 导读
CIKA 框架把 LLM 本身当作干预模拟器，通过提示词将概念状态设为“已掌握”，以正确率变化估计因果效应，并形式化为干预能力探针（ICP）。
正文 · 原文
Abstract:Recent methods for improving LLM mathematical reasoning, whether through MCTS-based test-time search or causal graph-guided knowledge injection, cannot identify which concepts causally contribute to a correct answer, as the observed association may be spurious, driven by confounders such as problem difficulty.
We propose CIKA (Causal Intervention for Knowledge Activation), a framework that uses the LLM itself as an interventional simulator: a prompt sets the concept state to ``mastered'' and the correctness change estimates the causal effect. We formalize this quantity as an Interventional Capability Probe (ICP), which diagnoses whether the LLM can use a given concept -- distinct from merely possessing knowledge. Because the intervention exogenously sets the concept state independently of problem difficulty, ICP separates confounding that observational methods cannot.
On 67 screened problems, the ICP of the top-ranked concept (+0.219) is significantly larger than that of the negative control (+0.039; paired $t$-test, $p < 10^{-6}$, Cohen's $d = 0.86$), confirming that the probe discriminates causally relevant concepts from irrelevant ones. Analysis of 601 Omni-MATH problems further shows that solved problems have 6.1$\times$ higher ATE than unsolved ones (0.338 vs. 0.055), confirming that ICP is predictive of problem-solving success. With a 7B-parameter LLM whose weights are entirely frozen, CIKA achieves 69.7\% on the contamination-free Omni-MATH-Rule benchmark and 64.0\% overall, compared to 60.5\% for o1-mini, and 97.2\% on GSM8K, 46--50\% on AIME 2024--2026, and 46.2\% on MathArena. The Causal Knowledge Activation component contributes 33.8\% of correct answers on problems where the base model alone fails, demonstrating that the LLM already possessed but had not activated the requisite knowledge.
| Comments: | 17 pages, 0 figures. v2: Revised as a technical report. Clarified that the problem-decomposition design leads to 110+ LLM calls per problem, and added a discussion of call cost and independent sampling. Corrected the bibliography |
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2605.07600 [cs.LG] |
| (or arXiv:2605.07600v2 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2605.07600 arXiv-issued DOI via DataCite |
Submission history
From: Tsuyoshi Okita [view email]
[v1]
Fri, 8 May 2026 11:19:01 UTC (28 KB)
[v2]
Sun, 25 Oct 2026 17:27:38 UTC (23 KB)
来源：arXiv cs.CL · arxiv.org