[CVPR2023 Highlight] Consistent-Teacher: Towards Reducing Inconsistent Pseudo-targets in Semi-supervised Object Detection

Simple-to-use-and-extend implementation of the DeepMimic Approach using the MuJoCo Physics Engine and Stable Baselines 3, mainly for locomotion tasks.

Official Code for SGRv2 and SGR.

The code corresponding to the paper "Improving Sample Efficiency of Deep Reinforcement Learning for Bipedal Walking".

Code for the "Sample-efficient Integration of New Modalities into Large Language Models" paper

World-model-based molecular optimization with counterfactual planning and 2,500× oracle savings.

Markov Decision Process DQN with Noisy Networks for Exploration (ICLR 2018) - 21.1% performance improvement over ε-greedy.

Evidence-checked survey and reproducible data atlas of sample, robot-time, wall-clock, and lifecycle cost in real-world robot RL.

Model-based vs model-free RL in a gridworld: does planning beat trial and error and how little world-model is actually enough? A clean, seed-averaged reproduction + small empirical study.

Sample-efficiency study for Diffusion Policy on PushT, trained on AMD ROCm. Preliminary: regularization improves 100-demo success 8→19%, with no gain at 200 demos.

Can DQN learn more from each experience and reduce the need for total samples