Onur Celik
@onclk_
Phd student in machine learning and robotics @ Autonomous Learning Robots KIT
I couldn’t make it to #ICML2025, but our work on Diffusion-Based Maximum Entropy RL is there! We introduce DiME, a new approach that swaps the standard actor in MaxEnt RL with a conditional diffusion model. This bypasses the need for tricky entropy approximations and lets our…
Can we teach dexterous robot hands manipulation without human demos or hand-crafted rewards? Our key insight: Use Vision-Language Models (VLMs) to scaffold coarse motion plans, then train an RL agent to execute them with 3D keypoints as the interface. 1/7
Curious how recent #BayesianDeepLearning methods perform on data beyond the training distribution? At #NeurIPS2023, there's large-scale survey on the #WILDS dataset, exploring fine-tuning, transformers, and deep ensembles in real-world scenarios. 📃➡️ bit.ly/3PXR25r
I am happy to share our recent work, “On Uncertainty in Deep State Space Models for Model-Based Reinforcement Learning”, ( openreview.net/pdf?id=UQXdQyo… ) published at TMLR. With @geri_neumann Let me summarize our key contributions in a short thread:
Our work ‚Specializing Versatile Skill Libraries using Local Mixture of Experts‘ is accepted at #CoRL2021 🙂. openreview.net/forum?id=j3Rgu… Thanks to Prof. Neumann(@alr_kit), Zhou, Ge Li and Philipp Becker (@philippb06).
Our work 'Action Conditional Recurrent Kalman Network' is accepted at #CoRL2020. Excited to share more details at the virtual conf. :) Thanks to Prof Gerhard Neuman(@alr_kit) and our collaborators @philippb06, @dtrbchlr, @MarcHanheide, @LCAS_UoL. arxiv.org/abs/2010.10201