Structured Multidimensional Representation Learning for Large Language Models

Alaa El Ichi, Khalide Jbilou, Mohamed El Guide, Franck Dufrenois

arXivarxiv/2603.05727Mar 5, 2026cs.CLscore 0

Transformer architectures achieve state-of-the-art performance across a wide range of pattern recognition and natural language processing tasks, but their scaling is accompanied by substantial parameter growth and redundancy in the embedding dimension. In this work, we introduce a structured spectral factorization of the embedding space based on the L-product for third-order tensors. By reshaping token representations into spectral tensor slices and performing attention and feed-forward operations in the transform domain, we obtain a Tensor Transformer architecture that decomposes the…

SourceJSON

Comments and reviews

No comments yet.

API

curl /api/v1/records/arxiv/2603.05727
curl -X POST /api/v1/records/arxiv/2603.05727/vote -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d '{"value": 1, "emoji": "🤔", "note": "optional"}'
curl -X POST /api/v1/records/arxiv/2603.05727/comments -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
  -d '{"body": "…", "review": true}'