Multimodal representation learning combines complementary signals to capture shared semantic information. For dynamic scenes, these representations need to remain consistent across views while preserving distinctions between temporal states. In visual speech representation learning, view variation changes the visibility and reliability of articulation cues, making it difficult to learn such representations from synchronized audio and video. To address this challenge, we propose MuViRL, a self-supervised framework for multimodal view-invariant visual speech representation learning via calibrated view aggregation. The framework combines synchronized 3D Gaussian view synthesis with multimodal self-supervision to learn consistent representations from real and synthesized data. A multimodal decision model learns relative view utility from teacher prediction losses, and its calibrated scores guide the aggregation of complementary information. Aggregation-to-Single Distillation transfers the aggregated predictions to a single branch. Real-synthetic alignment matches synchronized windows across views while preserving their temporal relations. Experiments on real and synthesized multiview data demonstrate that MuViRL improves multimodal representation consistency and visual speech recognition performance across views.
Purpose: To develop a phase-aware self-supervised framework for accelerated high-resolution multi-echo gradient-echo (mGRE) MRI that preserves anatomical detail and complex echo information for quantitative mapping. Methods: We developed multi-stage phase-aware self-supervised reconstruction (MSPR-GRE), integrating progressive k-space expansion with phase estimation and refinement. Sequential unrolled networks reconstruct complex multi-echo images and update echo-dependent phase priors. A relaxation-informed temporal subspace is used for acquisitions with at least four echoes, while shorter echo trains are reconstructed in the image domain. Training uses held-out acquired k-space samples without fully sampled targets. Evaluation included 9.4 T ex vivo mouse-brain data 30um eight-echo and 18um dual-echo) and retrospectively undersampled 7 T AHEAD human-brain data, with supervised ablations. Results:MSPR-GRE reduced artifacts in regions of rapid phase variation and preserved fine mouse-brain anatomy. Across five AHEAD test datasets at R = 4, 8, and 12, it generally improved magnitude-image, T2* map, and QSM metrics, particularly at higher acceleration. At R = 12, fourth-echo magnitude NRMSE was 12.09$\pm$0.60% versus 21.55$\pm$1.12% for SSDU. Ablations indicated benefits of phase refinement, while supervised experiments supported multi-stage phase-aware reconstruction. Conclusion: MSPR-GRE improves accelerated mGRE reconstruction and quantitative-map agreement without fully sampled training targets. Further validation of generalizability and phase-estimation accuracy is needed.
Reinforcement learning with verifiable rewards (RLVR) has delivered impressive gains in mathematical and multimodal reasoning and has become a standard post-training paradigm for contemporary language and vision-language models. However, the RLVR recipe introduces a significant risk of capability regression, in which models forget foundational skills after prolonged training without employing regularization strategies. We empirically confirm this concern, observing that open-source reasoning models suffer performance degradation on core capabilities such as perception and faithfulness. While imposing regularization terms like KL divergence can help prevent deviation from the base model, these terms are computed on the current task and therefore do not guarantee preservation of broader knowledge. Meanwhile, commonly used experience replay across heterogeneous domains makes it nontrivial to decide how much training emphasis each objective should receive. To address this, we propose RECAP-a replay strategy with dynamic objective reweighting for general knowledge preservation. Our reweighting mechanism adapts online using short-horizon signals of convergence and instability, shifting the post-training focus away from saturated objectives and toward underperforming or volatile ones. Our method is end-to-end and readily applicable to existing RLVR pipelines without training additional models or heavy tuning. Extensive experiments on benchmarks using Qwen2.5-VL-3B and Qwen2.5-VL-7B demonstrate the effectiveness of our method, which not only preserves general capabilities but also improves reasoning by enabling more flexible trade-offs among in-task rewards.