aiqbench analysis
This paper introduces PLOT, a gradient-free method that uses optimal transport to efficiently align high-level causal models with low-level neural computations. By progressively localizing intervention sites, the approach significantly reduces computational costs while maintaining accuracy in mechanistic interpretability tasks like binary addition and MCQA. The technique effectively separates correspondence search from subspace learning, offering a scalable framework for constructing executable intervention handles without requiring rotation learning.
PLOT: Progressive Localization via Optimal Transport in Neural Causal Abstraction
arXiv:2605.06979v2 Announce Type: replace-cross Abstract: Causal abstraction offers a principled framework for mechanistic interpretability, aligning a high-level causal model with low-level neural computation through interchange intervention analysis. Finding such an alignment, however, often requires fitting and evaluating separate learned mappings across many candidate neural locations. We introduce PLOT, a gradient-free approach for joint correspondence discovery via a global matching of intervention effects. PLOT represents abstract and neural interventions by geometric signatures of their effects on the shared task output and fits an optimal transport coupling between the two collections. The coupling can be calibrated directly into an executable intervention handle or used progressively to select coarse parent sites for finer localization, reducing the cost of searching broad collections of neural sites. Experiments on hierarchical equality, binary addition, and multiple-choice question answering (MCQA) demonstrate fast and accurate direct handles without learning intervention rotations and show that progressive localization can improve accuracy under intervention-size constraints while reducing computational cost. PLOT can also guide gradient-based subspace learning, with PLOT-guided distributed alignment search (DAS) attaining accuracy comparable to full DAS at approximately $3.6\times$ and $20\times$ lower serial runtime on binary addition and MCQA, respectively. PLOT thus separates the search for candidate correspondences from optional subspace learning and provides a common framework for constructing and testing neural intervention handles for a given causal model.
Abstract:Causal abstraction offers a principled framework for mechanistic interpretability, aligning a high-level causal model with low-level neural computation through interchange intervention analysis. Finding such an alignment, however, often requires fitting and evaluating separate learned mappings across many candidate neural locations. We introduce PLOT, a gradient-free approach for joint correspondence discovery via a global matching of intervention effects. PLOT represents abstract and neural interventions by geometric signatures of their effects on the shared task output and fits an optimal transport coupling between the two collections. The coupling can be calibrated directly into an executable intervention handle or used progressively to select coarse parent sites for finer localization, reducing the cost of searching broad collections of neural sites. Experiments on hierarchical equality, binary addition, and multiple-choice question answering (MCQA) demonstrate fast and accurate direct handles without learning intervention rotations and show that progressive localization can improve accuracy under intervention-size constraints while reducing computational cost. PLOT can also guide gradient-based subspace learning, with PLOT-guided distributed alignment search (DAS) attaining accuracy comparable to full DAS at approximately $3.6\times$ and $20\times$ lower serial runtime on binary addition and MCQA, respectively. PLOT thus separates the search for candidate correspondences from optional subspace learning and provides a common framework for constructing and testing neural intervention handles for a given causal model.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
arXiv:2605.06979 [cs.LG]
(or arXiv:2605.06979v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2605.06979 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 312 words
Related coverage
[AINews] Reality Checks on AI News (Yegge shuts down Gas Town, Databricks’ +60% Astra cost)
Latent Space · Sep 17, 2026
Dharma vs Adharma - The Perplexity of Being Alive
hillpost.in · Aug 25, 2026
Humanity’s Last Invention — Richard Socher of Recursive
Latent Space · Sep 14, 2026
Latent reasoning architectures would likely undermine CoT, our strongest oversight tool
AI Alignment Forum · Sep 23, 2026