AMUSE is a new optimizer integrating Muon orthogonalization with Schedule-Free averaging via adaptive interpolation for schedule-free anytime training that improves Pareto frontiers on vision and LLM tasks.
hub
Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
21 Pith papers cite this work. Polarity classification is still indexing.
21
Pith papers citing it
hub tools
citation-role summary
background 3
method 1
citation-polarity summary
representative citing papers
Thermo-VL augments a frozen Molmo-7B VLM with a trainable thermal encoder and prompt-conditioned dual-attention fusion to improve cross-spectrum visual reasoning.
Self-pretraining improves Transformer sequence classification by enabling learning of proximity-biased attention from positional encodings that label supervision alone cannot easily acquire from random starts.
Imagining in 360° decouples visual search into a single-step probabilistic semantic layout predictor and an actor, removing the need for multi-turn CoT reasoning and trajectory annotations while improving efficiency in 360° environments.
MUSE decouples reconstruction and semantic learning in visual tokenization via topological orthogonality, yielding SOTA generation quality and improved semantic performance over its teacher model.
Video Prediction Policy conditions robot action learning on future-frame predictions inside fine-tuned video diffusion models, yielding 18.6% relative gains on Calvin ABC-D and 31.6% higher real-world success rates.
Video-LLaVA creates a unified visual representation for images and videos via pre-projection alignment, enabling mutual enhancement from joint training and strong results on image and video benchmarks.
TD-MPC2 scales an implicit world-model RL method to a 317M-parameter agent that masters 80 tasks across four domains with a single hyperparameter configuration.
MVProbe is a multi-perspective probing framework for weight-space learning that combines first-order and Gram-based views and outperforms ProbeX on the Model Jungle benchmark.
TriForces adds a model-agnostic three-stream architecture plus self-supervised objectives to atomistic GNNs, improving transfer performance on MatBench, QM9, and limited-data OMat24 without DFT labels.
TAPE applies temporal-aware token pruning with smoothing, reselection, and timestep scheduling to speed up video diffusion models while preserving visual fidelity and coherence.
Introduces a staged pretrain-to-alignment workflow for geophysical AI that improves relative geologic time estimation across global field surveys despite limited labels and domain gaps.
A self-supervised approach uses consistent spatial relationships of anatomical structures across patients to improve 3D multi-modal medical image representations, yielding modest gains on segmentation and classification tasks.
Cosine similarity poorly predicts performance degradation from layer removal in LLMs, making direct accuracy-drop ablation a more reliable relevance metric.
SSL clustering is derived as KL-divergence optimization where a teacher-distribution constraint normalizes via inverse cluster priors and simplifies to batch centering by Jensen's inequality.
TaTok is a theoretically grounded adaptive tokenization method that uses global tokens and cumulative conditional entropy filtering to reduce redundancy while improving reconstruction quality over fixed-rate patch tokenization.
SplAttN uses Gaussian soft splatting and attention to avoid sparse projection collapse in point cloud completion, achieving SOTA results and demonstrating genuine visual cue reliance on KITTI.
A self-supervised method learns a fixed set of disentangled fingerprint tokens from medical time series by combining reconstruction loss with a total coding rate diversity penalty, framed as a disentangled rate-distortion problem.
mPLUG-Owl3 introduces hyper attention blocks to integrate vision and language for long image-sequence understanding and reports SOTA results on single-image, multi-image, and video benchmarks.