Semantic Manipulation Localization
Zhenshan Tan, Chenhan Lu, Yuxiang Huang, Ziwen He, Xiang Zhang, Yuzhe Sha, Xianyi Chen, Tianrun Chen, Zhangjie Fu · 2026
Read Semantic Manipulation Localization by Zhenshan Tan, Chenhan Lu, Yuxiang Huang, Ziwen He, Xiang Zhang, Yuzhe Sha, Xianyi Chen, Tianrun Chen, Zhangjie Fu (2026) 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-2604-10132.txt.
Description
Tan and eight colleagues define Semantic Manipulation Localization, a task that finds subtle edits changing what an image means. Modern editing leaves few low-level artifacts, tweaking an object's traits, state, or links while matching surroundings. Standard tools miss these edits, since they lean on artifacts, not meaning. The team builds a benchmark with pixel annotations from a meaning-driven editing pipeline. Their TRACE framework reasons in three stages: anchoring, perturbation sensing, and meaning-constrained reasoning. TRACE beats existing methods on the benchmark with more complete and coherent results.
Catalog record
| CREATOR | Zhenshan Tan, Chenhan Lu, Yuxiang Huang, Ziwen He, Xiang Zhang, Yuzhe Sha, Xianyi Chen, Tianrun Chen, Zhangjie Fu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 61352 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.10132