Causal Concept Graphs in LLM Latent Space for Stepwise Reasoning
This addresses the challenge of interpretability in AI reasoning for researchers and practitioners, though it is incremental as it builds on existing sparse autoencoder and graph learning methods.
The paper tackled the problem of understanding how concepts interact during multi-step reasoning in language models by proposing Causal Concept Graphs (CCG), which achieved a Causal Fidelity Score of 5.654±0.625, outperforming baselines like ROME-style tracing (3.382±0.233) with statistical significance (p<0.0001).
Sparse autoencoders can localize where concepts live in language models, but not how they interact during multi-step reasoning. We propose Causal Concept Graphs (CCG): a directed acyclic graph over sparse, interpretable latent features, where edges capture learned causal dependencies between concepts. We combine task-conditioned sparse autoencoders for concept discovery with DAGMA-style differentiable structure learning for graph recovery and introduce the Causal Fidelity Score (CFS) to evaluate whether graph-guided interventions induce larger downstream effects than random ones. On ARC-Challenge, StrategyQA, and LogiQA with GPT-2 Medium, across five seeds ($n{=}15$ paired runs), CCG achieves $\CFS=5.654\pm0.625$, outperforming ROME-style tracing ($3.382\pm0.233$), SAE-only ranking ($2.479\pm0.196$), and a random baseline ($1.032\pm0.034$), with $p<0.0001$ after Bonferroni correction. Learned graphs are sparse (5-6\% edge density), domain-specific, and stable across seeds.