Tal Daniel

Postdoctoral Researcher · ML & Robotics

Postdoctoral Fellow at the Carnegie Mellon University Robotics Institute, working with Prof. Deepak Pathak and Prof. David Held. Research spanning unsupervised & self-supervised representation learning, generative modeling, reinforcement learning, and robotics.

Ph.D. from the Technion ECE Department, advised by Prof. Aviv Tamar. B.Sc. and M.Sc. both in ECE from the Technion (Cum Laude / Summa Cum Laude).

Fellowships

2024 Irwin & Joan Jacobs Ph.D. Excellence Fellowship

2023 Miriam & Aaron Gutwirth Memorial Ph.D. Excellence Fellowship

✉ tdaniel at andrew dot cmu dot edu

GitHub

Scholar

Publications & Pre-Prints

3D-DLP: Self-Supervised 3D Object-Centric Scene Representation Learning

Ellina Zhang, Madhavan Iyengar, Amir Zadeh, Chuan Li, David Held, Deepak Pathak and Tal Daniel

ICML 2026

TL;DR — A self-supervised 3D object-centric representation that provides keypoints, bounding-boxes and segmentations directly in 3D, useful for multi-object manipulation

Paper Website Code

Latent Particle World Models: Self-supervised Object-centric Stochastic Dynamics Modeling

Tal Daniel, Carl Qi, Dan Haramati, Amir Zadeh, Chuan Li, Aviv Tamar, Deepak Pathak and David Held

ICLR 2026 Oral

TL;DR — A self-supervised object-centric world model that learns keypoints, and masks directly from videos, supports multi-modal conditioning, scaled to real-world multi-object datasets

Paper Website Code Video

Hierarchical Entity-centric Reinforcement Learning with Factored Subgoal Diffusion

Dan Haramati, Carl Qi, Tal Daniel, Amy Zhang, Aviv Tamar and George Konidaris

ICLR 2026

TL;DR — An entity-centric hierarchical RL framework using diffusion subgoals for improved long-horizon tasks.

Paper Website Code Video

EC-Diffuser: Multi-Object Manipulation via Entity-Centric Behavior Generation

Carl Qi, Dan Haramati, Tal Daniel, Aviv Tamar and Amy Zhang

ICLR 2025

TL;DR — Multi-object control policies for robotics from pixels using deep latent particles (DLP) and diffusion models.

Paper Website Code

DDLP: Unsupervised Object-centric Video Prediction with Deep Dynamic Latent Particles

Tal Daniel and Aviv Tamar

TMLR 2024

TL;DR — Object-centric video prediction, generation, and modification via the deep latent particle (DLP) representation.

Paper Website Code Video

Entity-Centric Reinforcement Learning for Object Manipulation from Pixels

Dan Haramati, Tal Daniel and Aviv Tamar

ICLR 2024 Spotlight

TL;DR — Using DLP instead of raw pixels unlocks compositional capabilities in reinforcement learning.

Paper Website Code Video

Unsupervised Image Representation Learning with Deep Latent Particles

Tal Daniel and Aviv Tamar

ICML 2022 Spotlight

TL;DR — Represent images as latent particles for unsupervised object detection, segmentation, and manipulation.

Paper Website Code Video

Soft-IntroVAE: Analyzing and Improving the Introspective Variational Autoencoder

Tal Daniel and Aviv Tamar

CVPR 2021 Oral Best Paper — MLIS-TCE 2022

TL;DR — Stable adversarial VAE training without a discriminator; applicable to density estimation, image generation, OOD detection, and more.

Paper Website Code Video

Deep Variational Semi-Supervised Novelty Detection

Tal Daniel, Thanard Kurutach and Aviv Tamar

NeurIPS 2021 Workshop — Deep Generative Models

TL;DR — Principled incorporation of negative samples in the VAE framework for meaningful representations.

Paper

Beyond Credential Stuffing: Password Similarity Models Using Neural Networks

Bijeeta Pal, Tal Daniel, Rahul Chatterjee and Thomas Ristenpart

IEEE S&P 2019

TL;DR — Cracking passwords with neural networks — and building defenses against such attacks.

Paper

Teaching

Course materials from my time at the Technion — all freely available on GitHub. Tutorials are in Jupyter Notebook format with theory, Python & PyTorch code, and PDF versions.