Task description 📝

MiMIC is focused on multi-modal machine-generated content detection. This first edition focuses specifically on (text, image) pairs, as they are the most prevalent in the web. It consists of two subtasks focused on the same detection task, both for English and Spanish:

Subtask 1: Multimodal Machine Generated Content Detection in English 🇬🇧

Participants must detect whether a pair of (image, English text) has been fully or partially created by a human or generative models.

Subtask 2: Multimodal Machine Generated Content Detection in Spanish 🇪🇸

Participants must detect whether a pair of (image, Spanish text) has been fully or partially created by a human or generative models.

Both subtasks involve multiclass classification of (text, image) pairs into four categories: fully-generated, fully-human, image-generated, and text-generated.

With these two subtasks, our goal is to explore (i) the feasibility of detecting multimodal machine-generated content, (ii) the unique challenges posed by each modality in identifying generated content, and (iii) if text and image signals complement each other to help detectors.

Participants are not limited to multimodal approaches, i.e., uni-modal approaches are also permitted, as understanding how multimodal signals work compared to uni-modal signals is essential.

IMPORTANT: there are restrictions to participate in the task. Please, read the Task constraints section.

Examples 📌

Subtask 1: Multimodal Machine Generated Content Detection in English 🇬🇧

The Cray X-MP was a supercomputer designed, built, and produced by Cray Research. It was the company's first vector processor, shared-memory, and parallel processing computer. It was the 1982 successor to the 1976 Cray-1 and the world's fastest computer between 1983 and 1985. The lead designer was Steve Chen. The CERN CRAY-XMP48 is displayed at EPFL in Switzerland.

Label: image-generated

The Monastery of Santa María de Toloño is an ancient monastery located in the Sierra de Toloño, in La Rioja Alavesa. Founded in the 12th century, it was an important religious and cultural center in the region. Although it is now in ruins, its history and location in a stunning natural setting make it a site of historical and tourist interest. Panoramic view of the Sierra de Toloño and the surrounding vineyards in La Rioja Alavesa.

Label: text-generated

San Jorge (in Basque, Sanduzelai) is a neighborhood in Pamplona, in the Chartered Community of Navarra, Spain, located in the northwest of the city. Evidence of settlement in a nearby area dates back to the late Bronze Age, specifically in the Soltxate area or Santa Lucía Hill. Later, during the Middle Ages, there was a modest population in what is now known as San Jorge, organized into three small settlements: one near the current Cuatro Vientos Bridge, another close to today's health center, and a third next to the Miluze Bridge. For a long time, these settlements were considered part of another Pamplona neighborhood, Rochapea. However, it was from the 1960s onward that San Jorge truly emerged as a distinct neighborhood. From the construction of the first large apartment blocks in the 1960s to the present, the neighborhood has undergone significant revitalization.

Label: fully-human

Cissopis leverianus, known as the conoto, is a bird from the Laurasic family, which inhabits tropical and subtropical regions of America. It is characterized by its white plumage and distinctive vocalizations. The conoto is a bird notable for its bright white plumage.

Label: fully-generated

Subtask 2: Multimodal Machine Generated Content Detection in Spanish 🇪🇸

Phragmipedium caudatum, una orquídea terrestre originaria de América del Sur, se caracteriza por sus grandes flores con pétalos excepcionalmente largos. Es una especie de gran valor ornamental, apreciada por coleccionistas y aficionados. Sus flores varían en color, con tonos verdes, blancos y marrones, mostrando un labelo distintivo. Su cultivo requiere condiciones específicas de humedad y temperatura, siendo una planta epífita o terrestre. Flores de Phragmipedium caudatum

Label: text-generated

El Infrared Background Signature Survey es un estudio que busca caracterizar el fondo infrarrojo del espacio, proporcionando información esencial sobre la radiación infrarroja ambiental. Este proyecto tiene como objetivo permitir una mejor comprensión de fuentes de energía térmica en el cosmos, mejorando así la detección de objetos celestes. Los datos recogidos son importantes para múltiples aplicaciones, incluyendo la astrofísica y la exploración espacial. Un satélite en órbita observando la Tierra desde el espacio en infrarrojo.

Label: fully-generated

Santa Fe 443, también conocido como Torre Santa Fe Pads o Torre Impulso, es un rascacielos ubicado en Avenida Santa Fe nº443, Cruz Manca, en la delegación Cuajimalpa, en la colonia Cruz Manca de Ciudad de México. Cuando finalizó su construcción era el edificio más alto de Santa Fe superando al edificio Torre Panorama Santa Fe. En la Ciudad de México se convirtió en el octavo más alto. Es considerado uno de los edificios más modernos del Distrito Federal. Fue superado en marzo de 2008 por la Torre Paragon Santa Fe. Torre Santa Fe Pads en 2011.

Label: fully-human

El Villar es un pueblo desaparecido, y no solo despoblado, actualmente integrado en el municipio de Robledillo de la Jara, en la Comunidad de Madrid. Se halla en la margen derecha del arroyo también llamado del Villar, afluente del río Lozoya por su izquierda, y junto al cruce de los antiguos caminos de Berzosa del Lozoya a Cervera de Buitrago, de Cervera de Buitrago a Manjirón y de Robledillo al Molino del Villar. Se cree probable que el caserío procediera de un antiguo asentamiento romano, ya que de procedencia romana es el topónimo Villar, precisamente uno de los pocos que se conocen por la zona norte de Madrid con etimología anterior a la Reconquista. Paraje de Las Cerquillas del Villar

Label: image-generated

Page updated

Google Sites

Report abuse