Scheduling Recursive Reasoning in Looped Transformers Collection TRM checkpoints trained with progress–stability co-design for TAPS adaptive recurrent inference. Includes Sudoku and Maze models, config • 4 items • Updated 24 days ago
Thinking Outside the Box: Can Language Models Rely on External Guidance Selectively? Paper • 2609.39578 • Published Sep 30 • 20
Thinking Outside the Box: Can Language Models Rely on External Guidance Selectively? Paper • 2609.39578 • Published Sep 30 • 20