The Causal Round Trip: Generating Authentic Counterfactuals by Eliminating Information Loss
Rui Wu [email protected], Lizheng Wang [email protected], Yongjun Li [email protected] · 2025
Read The Causal Round Trip: Generating Authentic Counterfactuals by Eliminating Information Loss by Rui Wu [email protected], Lizheng Wang [email protected], Yongjun Li [email protected] (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-05236.txt.
Introduction
Rui Wu and co-authors tackle a core flaw in using diffusion models for counterfactual work in causal models. Pearl's frame needs faithful abduction, the exact recovery of hidden noise, but standard diffusion designs lose information by design, since they are tuned for looks, not logic.
Their causal round trip aims to erase that loss, so a fact can be lifted to its noise, changed, and brought back as a true counterfactual. The work names why pretty image tools mislead in causal use. It then sets a path to counterfactuals that keep faith with the source case.
Catalog record
| CREATOR | Rui Wu [email protected], Lizheng Wang [email protected], Yongjun Li [email protected] |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 129 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-05236