Language Models Without a Trainable Input Embedding Table - a Bochkov Collection

Bochkov 's Collections

Beyond the Parameter Monolith: Modular Language Modeling

Do Language Models Need a Trainable Input Embedding Table?

Language Models Without a Trainable Input Embedding Table

Emergent Semantics Beyond Token Embeddings

Growing Transformers:Layer-wise Expansion Comparative Study

Language Models Without a Trainable Input Embedding Table

updated 12 days ago

This collection is provided for reproducibility of the paper's main claim

Bochkov/llm-fix-min-baseline-learned-input-table-model-classic

Text Generation • 0.5B • Updated 18 days ago • 356
Bochkov/llm-fix-min-fixed-minimal-binary-code

Text Generation • 0.5B • Updated 18 days ago • 370
Bochkov/llm-fix-min-affine-recoded-minimal-code-table-free

Text Generation • 0.5B • Updated 18 days ago • 352
Language Models Without a Trainable Input Embedding Table: Learning from Fixed Minimal Binary Token Codes

Paper • 2605.09751 • Published May 10