armenjeddi/LoopFormer-3block-8iterations-FineWeb300K
0.3B • Updated
• 82
LoopFormer
Models trained in the ICLR2026 paper: LoopFormer: Elastic-Depth Looped Transformers for Latent Reasoning via Shortcut Modulation
-
-
armenjeddi/NanoGPT-Base24-FineWeb300K
1B • Updated • 50 -
armenjeddi/LoopFormer-3block-8iterations
0.3B • Updated • 425 -
armenjeddi/LoopFormer-2block-12iterations
0.2B • Updated • 137 -
armenjeddi/LoopFormer-1block-24iterations
0.2B • Updated • 199 -
armenjeddi/NanoGPT-Base24
1B • Updated • 132 -
armenjeddi/NanoGPT-Base12
0.6B • Updated • 100 -
armenjeddi/NanoGPT-Base6
0.3B • Updated • 81 -
armenjeddi/NanoGPT-Base3
0.2B • Updated • 44 -
armenjeddi/Base-Loop-3block-8iterations
0.2B • Updated • 109 -
armenjeddi/Base-Loop-EE-3block-8iterations
0.2B • Updated • 159 -
armenjeddi/Base-Loop-EE-2block-12iterations
0.2B • Updated • 118 -
armenjeddi/Base-Loop-EE-1block-24iterations
0.1B • Updated • 128 -
armenjeddi/TMLT-3block-8iterations
0.3B • Updated • 69 -
armenjeddi/TMLT-EE-3block-8iterations
0.3B • Updated • 101 -
armenjeddi/TMLT-EE-2block-12iterations
0.2B • Updated • 96 -
LoopFormer: Elastic-Depth Looped Transformers for Latent Reasoning via Shortcut Modulation
Paper • 2602.11451 • Published • 15