Pinned Loading

1. dsan6650_final dsan6650_final Public

   Recursive reasoning with tiny transformers (<1M params): TRM + MoE + MCTS for Sudoku solving

   Jupyter Notebook 2

2. 6600_FinalProject 6600_FinalProject Public

   DSAN 6600 Final Project

3. rfgsa rfgsa Public

   Forked from piotrpiekos/MoSA

   AR-supported Mixture-of-Sparse-Attention (MoSA) using the Routing-Free method.

   Python

4. routing-in-sparse-attention routing-in-sparse-attention Public

   Comparison of routing strategies in sparse attention: Dense, MoSA, RFGSA, PRSA on WikiText-103 (10-epoch, equal-budget) with per-method routing visualizations and poster source.

   Python