Germán Ortiz

🇲🇽 PhD student in #NLProc at @INSAITinstitute & @UKPLab | Prev. @mbzuai, @UNAM_MX

[Germán Ortiz @jgermanmx]

🤖 Can malicious prompting push LLMs to generate convincing but misleading charts?

🔍 Yes, and we introduce ChartAttack, a framework to evaluate this risk in MLLM-generated charts.

Thanks @gistdotscience for the summary: https://t.co/cSphPLEVJp

[Germán Ortiz @jgermanmx]

🫠🫠🫠

[Xin Eric Wang @xwang_lk]

ICLR ＝ I Can Locate Reviewers
ACL ＝ Authors Can be Located

From Rednote. 😂

[ELLIS Unit Lisbon @Lisbon_ELLIS]

Last week, we proudly hosted the 15th edition of the Lisbon Machine Learning School where we welcomed more than 140 participants for an intensive and inspiring program!

🔗 To read more about LxMLS 2025: https://t.co/lU09dgqppI

🔗 To revisit the talks: https://t.co/SNyicvArOQ

[Jasper Dekoninck @j_dekoninck]

Thrilled to share a major step forward for AI for mathematical proof generation!

We are releasing the Open Proof Corpus: the largest ever public collection of human-annotated LLM-generated math proofs, and a large-scale study over this dataset!

[Germán Ortiz @jgermanmx]

🎯In summary, combining parameter-efficient methods and hypernetwork-based weight generation is a solid alternative to solving Sequence Labelling tasks in a multi-task setting in high—and low-resource scenarios (7/7)

📑 Our paper is available on arxiv: https://t.co/PCVixmMuNr

[Germán Ortiz @jgermanmx]

🚨Happy to introduce our latest work: HyperLoader, a simple approach that combines different parameter-efficient fine-tuning methods in a multitask setting using hypernetworks for Sequence Labelling (1/🧵)

Collab with @helen_py and @thamar_solorio

📑 https://t.co/PCVixmMuNr

[Germán Ortiz @jgermanmx]

📉Our approach ranks third in the percentage of trainable parameters, behind Hyperdecoder and UniPELT with BERT, the latter using a single-task approach. This suggests that HyperLoader's improvement is not solely due to adding more trainable parameters. (6/🧵)

[Barbara Plank @barbara_plank]

Now on stage Mirella Lapata giving the keynote today at #EACL20204

[Germán Ortiz @jgermanmx]

I am thrilled to announce the Mexican Society of Artificial Intelligence ( @smia_mx ) granted me the José Negrete Award for the best master’s thesis in an AI-related field at #MICAI2023 for my work on paraphrase detection and the proposal of the Sentence-CROBI neural network!

[Gpo Ing. Lingüística @GIL_UNAM]

Hoy en #JuevesDeDifusión te hablaremos de "Detección automática de paráfrasis mediante el modelo Sentence-CROBI, una arquitectura de red neuronal simple basada en codificadores cruzados y bi-codificadores", una tesis escrita por Jesús Germán Ortiz Barajas +

[SomosNLP @SomosNLP_]

🔥 #HACKTOBERFEST: Corpus en español (y lenguas cercanas) 🔥

Vamos a recopilar corpus de diferentes países, registros y dominios para impulsar la creación de modelos que realmente hablen como los 600 millones de personas hispanohablantes.

➡️ ¡Únete! https://t.co/y7iqRONwZ4

[SEPLN @sepln]

📢 Comenzamos el overview de “HOMO-MEX at Iberlef 2023: Hate speech detection in Online Messages directed Towards the MEXican Spanish speaking LGBTQ+ population” con Helena Gómez y Sergio Ojeda

🔗Accede al enlace:
https://t.co/6iMIcF6jm8

#sepln2023 #iberlef2023 @iberlef

[Frank Dellaert @fdellaert]

A nice blog post covering ideas and techniques above and beyond “vanilla” transformers: The Transformer Family Version 2.0 | Lil'Log https://t.co/Sg4Ahc3mbT

[Info Quórum @quorum_mx]

🧑‍🔬🗣️👂 Científicos mexicanos se dieron cita en en #Cholula, para participar en el #Mexlef2022, donde presentaron los los más recientes avances en investigación en el tema del tratamiento automático del lenguaje. @inaoe_mx #INAOE
https://t.co/6ZIXNkIAVK

[Juan Luis García Men @juanluis170290]

2da Jornada de #MexLEF2022 Overview of PAR-MEX.
Muchas gracias Jesús Germán Ortiz-Barajas!
@LabTL_INAOE

[SEPLN @sepln]

Jesús Germán Ortiz Barajas (@UNAM_MX) explica "Overview of PAR-MEX at Iberlef 2022: Paraphrase Detection in Spanish Shared Task"

#IberLEF2022 #SEPLN2022

https://t.co/1kCJTZc3HB