Learning Multimodal Embeddings with Evidence-Aligned Readout Paper • 2609.33659 • Published Sep 27 • 12
What Really Works for Universal Multimodal Embeddings? A Controlled Study of Contrastive Training Strategies Paper • 2610.15035 • Published 20 days ago • 17
Learning to Read Out: Unembedding Dynamics in Language Model Pretraining Paper • 2610.21025 • Published 16 days ago • 10
Beyond Policy Rollouts: Embedding-Space Directional Exploration with Anchored Noise for LLM RLVR Paper • 2610.16105 • Published 18 days ago • 17
SpecMRL: Redesign Matryoshka Representation Learning within Parametric Spectrum Paper • 2610.26824 • Published 11 days ago • 89
Embeddings Are Overpriced? A Frozen 16 MB Vocabulary for Small and Ternary Language Models Paper • 2610.22428 • Published 15 days ago • 21
Periodic Weak Spots: Phase Sensitivity from Chunked KV-Cache Compression Paper • 2609.36322 • Published Sep 28 • 120
Redesign Mixture-of-Experts Routers with Manifold Power Iteration Paper • 2606.12397 • Published Jun 10 • 91
SPARKLING: Balancing Signal Preservation and Symmetry Breaking for Width-Progressive Learning Paper • 2602.02472 • Published Feb 2 • 48
Coupling Experts and Routers in Mixture-of-Experts via an Auxiliary Loss Paper • 2512.23447 • Published Dec 29, 2025 • 100