The paper introduces Trust Guided Decision Transformer (TGDT), a method that mitigates performance degradation in Decision Transformers during long rollouts by monitoring the model’s next‑state prediction error. TGDT evaluates multiple recent context suffixes, filters out those whose prediction error exceeds a calibrated threshold, and then selects the highest‑value action from the remaining trusted suffixes using a frozen critic. Experiments on D4RL navigation and locomotion tasks demonstrate that TGDT reduces persistent high‑error runs and improves returns compared to vanilla Decision Transformer and other context‑control baselines.
arXiv:2609.23775v1 Announce Type: new Abstract: Obtaining the optimal action-value function in Markov decision processes is computationally intensive in large state--action spaces. In this study, we...
The paper introduces an adaptive rollout truncation method for offline world model training that uses epistemic uncertainty to decide when to stop autoregressive rollouts. By calibrating a threshold during a warm‑up phase, the approach replaces fixed‑horizon rollouts with uncertainty‑driven truncation, evaluated with ensemble and Monte Carlo dropout estimators. Experiments on ANYmal‑D and ANT demonstrate that this strategy matches or surpasses fixed‑horizon training while reducing cumulative rollout steps by about 72%.
arXiv:2608. 02958v1 Announce Type: cross Abstract: Vision-Language-Action (VLA) policies trained by behavior cloning fail silently: from the action stream alone, a collapsing rollout looks much like one making clean progress, because imitation supplies no notion of progress.
arXiv:2605.22164v2 Announce Type: replace Abstract: Latent world models can learn representations that contain information needed for control, while the downstream controller may still rank candidate...
arXiv:2607. 27914v1 Announce Type: new Abstract: Multi-zone variable-air-volume control must balance thermal comfort, indoor air quality, and electricity use across several continuous actuators.
arXiv:2503. 03660v4 Announce Type: replace Abstract: We introduce a sequence-conditioned critic for Soft Actor-Critic (SAC) that models trajectory context with a lightweight Transformer and trains on aggregated $N$-step targets.
arXiv:2602. 11619v2 Announce Type: replace Abstract: Running the same LLM agent on identical inputs yields 2.
arXiv:2608. 07809v1 Announce Type: new Abstract: A world model is only useful for physical AI if it changes what the agent does, and only safe if it declines to do so when it is wrong.
arXiv:2609.24840v1 Announce Type: cross Abstract: Diffusion models offer flexible motion generation, but translating this flexibility into feedback-responsive humanoid control remains challenging. Hi...
The paper proposes a method for selectively querying language‑model advice in reinforcement learning by predicting the value of potential responses and only querying when the expected benefit outweighs the cost. It introduces a certified, response‑contingent metareasoning framework that guarantees near‑optimal advice usage under certain assumptions, and demonstrates that a calibrated controller with Qwen2.5 advisors can improve task performance while drastically reducing the number of advice calls on the BabyAI benchmark.
arXiv:2607. 01218v1 Announce Type: cross Abstract: Transformers use the same forward computation stream to both predict the next token and store useful state for future token predictions.
The paper investigates how small action errors evolve when using action chunking in behavioural cloning. By injecting errors at each state and observing their growth under open‑loop (no replanning) and closed‑loop (replanning) regimes, the authors classify states as contracting, expanding, or unresolved. Across twelve manipulation tasks, they find that stable states are rare, error amplification is common, and that short‑horizon fitting can overestimate long‑horizon propagation. Predictors trained on camera and proprioceptive data can recover open‑loop stability but only partially capture closed‑loop dynamics, indicating that standard imitation learning does not reliably produce policies that contract errors when perturbed.