Day 3 of @fetc presented Great Scott an AI tool showcase. I met Dan and Frank just two hrs before at their poster session and they came to hype up my #FETC2026 session about #EdTechai. A huge thanks to @Kindermitch@johnwick for my closing line. @LemanskyRachel@gret
[#EdTechAI] El tema de investigación que estoy coordinando, en su edición en Frontiers in Education (@FrontiersIn), finalizará pronto. Algunos datos:
➤ Frontiers in Education: SJR Q2.
➤ Fecha límite: 02/02/2026.
➤ Más de 31.000 vistas y descargas.
➤ Más de 3.400 descargas.
Brief introduction to LLM backdoors Attribution
LLMs are vulnerable to backdoor attacks through data poisoning, yet the internal mechanisms governing these attacks remain a black box.
Previous research on interpretability for LLM safety tends to focus on alignment, jailbreak, and hallucination, but overlooks backdoor mechanisms, making it difficult to understand and fully eliminate the backdoor threat.
In this paper, aiming to bridge this gap, we explore the interpretable mechanisms of LLM backdoors through Backdoor Attribution (BkdAttr), a tripartite causal analysis framework.
• Interpretability Lens - We propose the BkdAttr interpretability framework, which is effective for different LLM architectures and backdoors. We make pioneering efforts to prove and analyze the existence and properties of backdoor components, filling the methodological and theoretical gaps.
• Progressive Techniques - We begin with the Backdoor Probe to detect backdoor features within representations and then propose BAHA to identify the backdoor attention heads for extracting these features, culminating in the Backdoor Vector as a potent backdoor activation controller.
• Instructive Insights - Our research elucidates the underlying mechanism of LLM backdoors: sparse backdoor attention heads transform the trigger presence into backdoor features, which can modulate backdoor activation via simple arithmetic addition or subtraction on LLM representations.
Source: arxiv.org/pdf/2509.21761
Miao Yu, Zhenhong Zhou, @MoayadAloqaily, Kun Wang, Biwei Huang, Stephen Wang, @JinYueming, @qingsongedu - @SquirrelAi_Edu, @USTCGlobal, @UAE_University, @UAEU_NEWS, @NUSingapore, @NTUsg#AIsecurity#LLMSecurity#AgenticAI#PromptInjection#ToolingSecurity#ModelSafety#AITrust#ResponsibleAI#MLOps#DataPrivacy#AcademicResearch#EdTechAI
EdTech founder buried in grading? 😩
We built an AI workflow to automate grading and feedback, freeing up hours every week.
Our client now focuses on teaching, not paperwork.
Ready to reclaim your time? DM "EDTECHAI"
Hello #foxfrens. It's poll time! 🗳️
What would you like to build with Foxsy AI tools?
A) AI teacher assistant
B) Homeschool robot
C) Study companion bot
D) Something we haven’t even imagined yet? 💡✨
@PulsarMvX send 500 FOXSY to 100 reactions
#AIEducation#EdTechAI
Data isn't just fuel — it's DNA 🧬 and Foxsy has 20 years of it 📊
Carefully labeled. Ethically sourced. Ready for you to build smarter AI tutors and even robot teachers.
Foxsy = data you can trust 🔒
@PulsarMvX send 500 FOXSY to 100 reactions
#EdTechAI#TrustedData
Imagine a world where your tutor knows your learning style better than any teacher ever could.
With Foxsy, that reality is near. Learning that adapts to you - not the other way around.
@PulsarMvX send 500 FOXSY to 100 reactions
#AIEducation#EdTech#EdTechAI
[#EdTechAI] El ABP-IA no es una tendencia, es un cambio de paradigma en el aula.
➤ Proyectos guiados por IA desde la idea hasta la evaluación.
➤ Alumnado más implicado y motivado.
➤ Docentes formados para usar tecnología con sentido pedagógico.
🔗 revistas.um.es/reifop/articl…
[#EdTechAI] Integrar inteligencia artificial (#IA) en el aprendizaje basado en proyectos (#ABP) tiene un gran impacto:
➤ Proporciona retroalimentación adaptativa.
➤ Aumenta la motivación estudiantil.
➤ Personaliza el aprendizaje del alumnado.
🔗 Aquí: mdpi.com/2227-7102/15/2/150
[#EdTechAI] ¡Muchas gracias! Poniendo el foco de la investigación en mejorar la experiencia de aprendizaje de todos y de todas a través de la Inteligencia Artificial (#IA) 🚀