The paper introduces Probabilistic Causal Impact (PCI), a framework that blends actual causality (AC) with Pearl’s probability of necessity and sufficiency to provide tractable, causally grounded explanations. PCI reframes explainability as an estimation problem on a probabilistic causal model, enabling efficient approximation via Monte Carlo sampling. The authors evaluate PCI on synthetic and real-world data, demonstrating consistency with AC, scalability, and applicability to complex continuous systems and large-scale causal machine learning models.
By Rafal Urbaniak, Sam Witty, Daniel Waxman, Andy Zane, Poorva Garg, Emily Bunnapradist, Sankaran Vaidyanathan, Jack Feser, Drew Lehe, Eli Bingham
arXiv:2605.12718v2 Announce Type: replace
Abstract: Multi-agent debate has emerged as a promising approach for improving LLM reasoning on ground-truth tasks, yet current methodologies face certain st...
By Tommaso Giovannelli, Griffin D. Kent
arXiv:2607. 15281v1 Announce Type: new Abstract: Causal and intervention-based question answering is fundamental to advancing large language models (LLMs) toward reasoning beyond surface-level correlations and understanding underlying causal mechanisms.
By Su Lan, Xuefei Yin, Yanming Zhu, Alan Wee-Chung Liew
arXiv:2608. 07214v1 Announce Type: cross Abstract: Modern AI is no longer a single model but an ecosystem: classical ML predictors, deep and multimodal models, large language models, and agents, each trained and tuned over different data sources and each producing outputs at scale that become inputs to the others.
By Dazhuo Qiu, Yingli Zhou, Amedeo Pachera, Angela Bonifati, Andrea Mauri
The paper investigates why causal Bayesian networks are labeled as causal, noting that interventions are defined within the model while predictions refer to real-world actions. It introduces a formal framework that separates the causal model from the data-generating process and formalizes how actions map to interventions. The authors prove an impossibility result showing that no non‑circular interpretation can satisfy all desirable properties, and they explore the consequences for falsification, causal discovery, representation learning, and abstraction.
By Frederik Hytting J{\o}rgensen, Luigi Gresele, Sebastian Weichwald
arXiv:2606. 11195v1 Announce Type: cross Abstract: Large language models (LLMs) have transformed how humans access information, but not how we reason with it.
By Rikard Rosenbacke, Carl Rosenbacke, Victor Rosenbacke, Martin McKee
arXiv:2510. 27544v3 Announce Type: replace Abstract: Current training paradigms, optimized for long-horizon reasoning trace execution, have made Large Language Models (LLMs) excel at pattern matching and forward simulation of reasoning, but underperform at counterfactual causal understanding and reasoning.
By Nikolaus Holzer, William Fishell, Baishakhi Ray, Mark Santolucito
arXiv:2605. 24528v2 Announce Type: replace Abstract: Real world decision-making requires constructing mental models under uncertainty over evidence, over the underlying causal rules, and over the state of the world itself.
By Jeffrey Qin, Wasu Top Piriyakulkij, Zhuangfei Gao, Mia Radovanovic, Jessica Sommerville, Kevin Ellis, Marta Kryven
arXiv:2607. 01251v1 Announce Type: cross Abstract: Debate, where AI agents argue opposing positions, has emerged as a key approach to scalable oversight.
By Yuyang Jiang, Chacha Chen, Teng Wu, Liwen Sun, Han Liu, Shi Feng, Chenhao Tan
The paper argues that as Large Language Models transition from chatbots to agentic systems, the current post-hoc interpretability paradigm is insufficient for safe deployment because it cannot audit or intervene before an output is produced. It proposes a shift to generative interpretability, where a model’s inference process inherently exposes semantically meaningful checkpoints that are human-understandable and can be causally intervened upon. The authors illustrate the advantages of this approach and introduce Neuro‑Symbolic Models as a concrete implementation.
By Xiaocong Yang
arXiv:2508. 21010v3 Announce Type: replace-cross Abstract: Existing Causal-Why Video Question Answering (VideoQA) models often struggle with higher-order reasoning, relying on opaque, monolithic pipelines that entangle video understanding, causal inference, and answer generation.
By Paritosh Parmar, Eric Peh, Basura Fernando
arXiv:2610.34094v1 Announce Type: new
Abstract: Large language models (LLMs) generally lack deep causal reasoning capabilities in temporal reasoning. Inspired by document layout reordering, this pape...
By Zhifeng Yang, Wei Luo, Yongjian Chen, Shiqi Ou