MiraBench: Evaluating Action-Conditioned Reliability in Robotic World Models

Action-conditioned world models are increasingly used as scalable simulators for robot learning, yet current evaluations provide limited evidence that their predictions are reliable under the actions they condition on. Existing benchmarks largely emphasize visual fidelity, leaving unclear whether predicted futures are physically plausible, faithful to commanded actions, and calibrated to failure when actions should not succeed. We introduce \textsc{MiraBench}, a hierarchical benchmark that defines \emph{action-conditioned reliability} as a core evaluation target for robotic world models. MiraBench decomposes this target into three progressively demanding levels: \emph{Physics Adherence}, which evaluates reference-free physical consistency; \emph{Action-Following Fidelity}, which measures whether predictions respect task-relevant action inputs; and \emph{Optimism Bias Detection}, which probes the tendency to predict successful outcomes under failure-inducing actions. To support this evaluation, we curate a human-annotated corpus with over 16,000 judgments across tasks, failure categories, and leading world models. We evaluate 12 representative model configurations spanning vector-conditioned robotic world models, text-conditioned generative world models, open-weight systems, closed-source systems, and multiple model scales. Across this broad model landscape, MiraBench reveals three central findings: visual fidelity is a poor proxy for action fidelity; increasing model scale does not reliably improve action following; and optimism bias is pervasive across current systems. By shifting evaluation from appearance to action-conditioned reliability, MiraBench provides a diagnostic foundation for assessing and improving robotic world models as faithful simulators.

Tournament-GRPO: Group-Wise Tournament Rewards for Reinforcement Learning in Open-Ended Long-Form Generation

Reinforcement learning in open-ended long-form generation is challenging because reliable reference answers and automatic metrics are often unavailable. Existing rubric-based methods typically rely on pointwise LLM-as-a-judge scoring, but absolute scores are difficult to calibrate across complex responses, may provide weak discrimination among same-query rollouts, and can become saturated during optimization. We propose Tournament-GRPO, a group-wise reward framework that converts rubric-guided LLM judgments into relative rewards through repeated multi-round tournaments among same-query rollouts. Tournament-GRPO compares candidates within groups, accumulates tournament outcomes, and normalizes them into group-wise rewards for GRPO training. Experiments on Deep Research Bench show that Tournament-GRPO consistently outperforms existing reward-design baselines, achieving a 4.52-point overall-score improvement over the strongest baseline. Further analyses show that tournament rewards provide a favorable effectiveness--efficiency trade-off and that tournament design affects training dynamics. These results suggest that rubric-guided tournament comparison provides an effective reward signal for reinforcement learning in open-ended long-form generation.

Joint DOA and Non-circular Phase Estimation of Non-circular Signals for Antenna Arrays: Block Sparse Bayesian Learning Method

This letter proposes a block sparse Bayesian learning (BSBL) algorithm of non-circular (NC) signals for direction-of-arrival (DOA) estimation, which is suitable for arbitrary unknown NC phases. The block sparse NC signal representation model is constructed through a permutation strategy, capturing the available intra-block structure information to enhance recovery performance. After that, we create the sparse probability model and derive the cost function under BSBL framework. Finally, the fast marginal likelihood maximum (FMLM) algorithm is introduced, enabling the rapid implementation of signal recovery by the addition and removal of basis functions. Simulation results demonstrate the effectiveness and the superior performance of our proposed method.

Wavelet-Enhanced Desnowing: A Novel Single Image Restoration Approach for Traffic Surveillance under Adverse Weather Conditions

Image restoration under adverse weather conditions refers to the process of removing degradation caused by weather particles while improving visual quality. Most existing deweathering methods rely on increasing the network scale and data volume to achieve better performance which requires more expensive computing power. Also, many methods lack generalization for specific applications. In the traffic surveillance screener, the main challenges are snow removal and veil effect elimination. In this paper, we propose a wavelet-enhanced snow removal method that use a Dual-Tree Complex Wavelet Transform feature enhancement module and a dynamic convolution acceleration module to address snow degradation in surveillance images. We also use a residual learning restoration module to remove veil effects caused by rain, snow, and fog. The proposed architecture extracts and analyzes information from snow-covered regions, significantly improving snow removal performance. And the residual learning restoration module removes veiling effects in images, enhancing clarity and detail. Experiments show that it performs better than some popular desnowing methods. Our approach also demonstrates effectiveness and accuracy when applied to real traffic surveillance images.