File size: 1,245 Bytes
2138d01 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 | cff-version: 1.2.0
message: "If you use TMT in your research, please cite it as below."
type: software
title: "TemporalMesh Transformer (TMT)"
abstract: >
A novel autoregressive language model architecture that simultaneously fuses
dynamic graph topology (Mesh Attention), token-level temporal semantic decay,
and per-token adaptive depth routing into a single unified model.
Achieves ~50% compute reduction and lower perplexity vs. parameter-matched baselines.
authors:
- name: "Vignesh"
alias: "vignesh2027"
repository-code: "https://github.com/vignesh2027/TemporalMesh-Transformer"
url: "https://huggingface.co/vigneshwar234/TemporalMesh-Transformer"
license: MIT
version: "1.0.0"
date-released: "2026-05-19"
keywords:
- transformer
- mesh-attention
- temporal-decay
- adaptive-depth
- graph-neural-network
- efficient-transformer
- language-model
- PyTorch
- NLP
- deep-learning
preferred-citation:
type: generic
title: >
TemporalMesh Transformer: Dynamic Graph Attention with Temporal Decay
and Adaptive Depth Routing
authors:
- name: "Vignesh"
year: 2026
url: "https://huggingface.co/vigneshwar234/TemporalMesh-Transformer"
notes: "Preprint. Available at GitHub and Hugging Face."
|