TD-MPC2
PulseAugur coverage of TD-MPC2 — every cluster mentioning TD-MPC2 across labs, papers, and developer communities, ranked by signal.
3 day(s) with sentiment data
-
New TrojanWorld framework backdoors reinforcement learning agents via imagination steering
Researchers have developed a new framework called TrojanWorld designed to backdoor world-model agents used in reinforcement learning. This framework exploits the predictive core of these agents by steering their interna…
-
New research highlights 'intervention gap' in AI world models
A new research paper titled "The Intervention Gap in Latent World Models" explores a critical property of learned world models: planning-time intervention fidelity. This property measures whether a model's internal tran…
-
New AI method boosts success rate for autonomous endovascular navigation
Researchers have developed a new method called Progressive Experience Fusion (PEF) to train controllers for autonomous endovascular navigation. This technique aims to improve the success rate of delivering mechanical th…
-
Multi-horizon consistency impacts video prediction geometry
Researchers have investigated the impact of multi-horizon latent consistency, a training parameter in video prediction and world models, on transition geometry. Their study, using Moving-MNIST as a primary test case, fo…
-
New research explores reinforcement learning advancements across multiple domains · 10 sources tracked
Multiple research papers published on arXiv explore advancements in reinforcement learning (RL) and its applications. One study focuses on improving the interpretability of RL policies through decision-tree pruning, dem…
-
New diagnostic tool improves world model evaluation for reinforcement learning
Researchers have introduced a new diagnostic tool called operator-on-F to better evaluate world models used in model-based reinforcement learning. This method complements existing value-equivalence checks by focusing on…
-
New latent space planning framework boosts agricultural robot navigation
Researchers have developed LeCropFollow, a novel visual navigation framework for agricultural robots operating in unstructured crop fields. This system utilizes a learned latent representation, integrating a self-superv…
-
FlowMPC framework enhances imitation learning with world models
Researchers have developed FlowMPC, a new framework that enhances the performance of Flow Matching (FM) policies in multimodal action spaces. By integrating a learned world model with an imitation-learned FM policy, Flo…
-
Equivariant World Models Offer Certified Predictability Horizon
A new research paper introduces a method for certifying the predictability horizon of equivariant world models. The approach provides a computable certificate that guarantees error bounds over time, stratified by the mo…
-
ELVIS: Ensemble-Calibrated Latent Imagination for Long-Horizon Visual MPC
Researchers have developed ELVIS, a novel approach to long-horizon visual planning in reinforcement learning that uses a Gaussian-mixture model predictive controller to maintain multiple hypotheses over extended rollout…