Javad Taghia's picture

Building on HF

6 7 40

Javad Taghia PRO

telcom

·

https://deegitals.com

AI & ML interests

text-to-image, image-to-image Unlearning, training models, model evaluation and safety/alignment benchmarking PhD - UNSW Sydney ( alumni )

Recent Activity

reacted to AdinaY's post with 👍 1 day ago

After a VLM, StepFun dropped a new audio model: Step-Audio-R1.1, enabling thinking while speaking 🔥 https://huggingface.co/stepfun-ai/Step-Audio-R1.1 ✨ Apache 2.0 ✨ Combines dual-brain architecture and acoustic-grounded reasoning to enable real-time dialogue with SOTA-level reasoning

posted an update 1 day ago

MAD-GRPO: https://huggingface.co/blog/telcom/mad-grpo In R1-Zero-Like Training *, Dr.GRPO treats GRPO’s by dropping std, but that often comes with a hidden side effect: length-weighted updates that can nudge model toward verbosity. MAD-GRPO provides robust scale (MAD + epsilon) per-token normalization stability without verbosity bias. *https://huggingface.co/papers/2503.20783

commented on a paper 1 day ago

Understanding R1-Zero-Like Training: A Critical Perspective

View all activity

Organizations

commented a paper 1 day ago

Understanding R1-Zero-Like Training: A Critical Perspective

Paper • 2503.20783 • Published Mar 26, 2025 • 59 •

New activity in AIML-TUDA/unsafe-vs-safe-stable-diffusion 17 days ago

Demo - Build error

#1 opened about 2 years ago by

Update app.py

#2 opened 17 days ago by

New activity in Tongyi-MAI/Z-Image-Turbo 25 days ago

Z-model dissect and compare to SDXL models

#113 opened 25 days ago by

commented a paper 26 days ago

Machine Unlearning for Image-to-Image Generative Models

Paper • 2402.00351 • Published Feb 1, 2024 • 15 •

New activity in h2oai/h2ogpt-4096-llama2-13b-chat over 2 years ago

How to use this model in Local H2O installation

#2 opened over 2 years ago by