Our 22 open source Gemstone models for scaling laws range from 50M to 2B parameters, spanning 11 widths from 256 to 3072 and 18 depths from 3 to 80.
Tom Goldstein's Lab at University of Maryland, College Park
university
AI & ML interests
AI security & privacy, algorithmic bias, foundations of ML
Recent Activity
Collections
7
These are checkpoints for recurrent LLMs developed to scale test-time compute by recurring in latent space.
-
tomg-group-umd/huginn-0125
Text Generation • Updated • 2.71k • 255 -
Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach
Paper • 2502.05171 • Published • 131 -
tomg-group-umd/huginn_swa_100_10_avg_0.9_merge
Text Generation • Updated • 25 -
tomg-group-umd/step-00010752-recurrence_full_512_0
Text Generation • Updated • 23
spaces
4
models
87
tomg-group-umd/huginn-0125
Text Generation
•
Updated
•
2.71k
•
255
tomg-group-umd/Gemstone-2560x8_cooldown
Text Generation
•
Updated
•
12
tomg-group-umd/Gemstone-3072x12_lr_ablation
Text Generation
•
Updated
•
13
tomg-group-umd/Gemstone-256x23_lr_ablation
Text Generation
•
Updated
•
20
tomg-group-umd/Gemstone-1024x28_lr_ablation
Text Generation
•
Updated
•
14
tomg-group-umd/Gemstone-384x13_lr_ablation
Text Generation
•
Updated
•
11
tomg-group-umd/Gemstone-2048x27_lr_ablation
Text Generation
•
Updated
•
14
tomg-group-umd/Gemstone-512x16_lr_ablation
Text Generation
•
Updated
•
15
tomg-group-umd/Gemstone-384x36_lr_ablation
Text Generation
•
Updated
•
11
tomg-group-umd/Gemstone-1792x18_lr_ablation
Text Generation
•
Updated
•
13
datasets
23
tomg-group-umd/huginn-dataset
Viewer
•
Updated
•
274M
•
666
•
1
tomg-group-umd/fictional_qa_03-19-25_training_splits
Viewer
•
Updated
•
36.9k
•
45
tomg-group-umd/trivia_qa_03-19-25_training_splits
Viewer
•
Updated
•
16.4k
•
27
tomg-group-umd/fictional_qa_03-19-25_processed_flat
Viewer
•
Updated
•
31.7k
•
37
tomg-group-umd/fictional_qa_03-12-25_processed_flat
Viewer
•
Updated
•
24.2k
•
61
tomg-group-umd/alpaca_cleaned_dataset_short
Viewer
•
Updated
•
32
•
45
•
1
tomg-group-umd/riftV1
Viewer
•
Updated
•
478k
•
4
tomg-group-umd/fictional_qa_11-08-24_refolded_1-10-25_txt
Viewer
•
Updated
•
417k
•
4
tomg-group-umd/fictional_qa_11-08-24_refolded_1-10-25
Viewer
•
Updated
•
444k
•
9
tomg-group-umd/rift
Viewer
•
Updated
•
714k
•
5