---
format: "aidr-story-markdown/v1"
id: "76f3c65e366c58cb541c7c7211dd17fdef7ca2ae0a72e5e63e8ca7eb51296b23"
canonical_url: "https://aidr.today/76f3c65e?lang=en"
title: "LensErase: A Foundation Model for Erase Anything"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-10-15T00:00:00.000Z"
category: "Models"
topics: ["inference"]
source_urls: ["https://huggingface.co/papers/2610.18072"]
summary: "Researchers introduced LensErase, a mask-based foundation model for erasing objects, text, and watermarks from images, trained on a 101.07 million example dataset and refined with supervised fine-tuning and reinforcement learning."
---

# LensErase: A Foundation Model for Erase Anything

> [Open the canonical story](<https://aidr.today/76f3c65e?lang=en>)

**Published:** 2026-10-15T00:00:00.000Z
**Category:** Models
**Topics:** inference

## Summary

Researchers introduced LensErase, a mask\-based foundation model for erasing objects, text, and watermarks from images, trained on a 101\.07 million example dataset and refined with supervised fine\-tuning and reinforcement learning\.

## Sources

- [Story source](<https://huggingface.co/papers/2610.18072>)