REAM: Merging Improves Pruning of Experts in LLMs
Paper • 2604.04356 • Published • 7
AI & ML interests
At the Frontier of Weight Space Learning
bknyazÂ
submitted a
paper to Daily Papers 6 months ago
bknyazÂ
authored 6
papers 6 months ago
(Almost) Free Modality Stitching of Foundation Models
Paper • 2507.10015 • Published • 2
M^4olGen: Multi-Agent, Multi-Stage Molecular Generation under Precise Multi-Property Constraints
Paper • 2601.10131 • Published • 17
Recursive Autoconvolution for Unsupervised Learning of Convolutional Neural Networks
Paper • 1606.00611 • Published
Celo2: Towards Learned Optimization Free Lunch
Paper • 2602.19142 • Published
Celo: Training Versatile Learned Optimizers on a Compute Diet
Paper • 2501.12670 • Published
REAM: Merging Improves Pruning of Experts in LLMs
Paper • 2604.04356 • Published • 7
bknyazÂ
authored 2
papers over 1 year ago
bknyazÂ
authored 4
papers about 2 years ago
Parameter Prediction for Unseen Deep Architectures
Paper • 2110.13100 • Published
Graph Neural Networks for Learning Equivariant Representations of Neural Networks
Paper • 2403.12143 • Published • 1
Accelerating Training with Neuron Interaction and Nowcasting Networks
Paper • 2409.04434 • Published
Can We Scale Transformers to Predict Parameters of Diverse ImageNet Models?
Paper • 2303.04143 • Published