5 papers
A two-level metacognitive framework for test-time adaptation that separates task execution from control, with a taxonomy of existing methods and the open challenges ahead.
Implicit Sheaf Operators compute sheaf diffusion as local operator-vector products instead of building the sheaf Laplacian, with up to 4.87× lower latency and 3.92× less peak GPU memory.
We turn the discriminative CLIP model into a generator capable of addressing multiple tasks, such as text-to-image synthesis, style transfer, and image editing, without fine-tuning.
The 2025 Topological Deep Learning Challenge asked for new datasets, benchmark tasks, and data pipelines in TopoBench, and drew 44 qualifying submissions.
Language models take turns writing tasks, answering them, and judging each other's answers. Weighting each judge by its reliability gives rankings that correlate with MMLU-Pro (78%) and GPQA (63%).
