What's the most promising successor to the Transformer?
MAMBA emerges as a leading contender for Transformer successors, boasting linear inference efficiency, yet remains untested on large-scale models.
A Survey of Logical Reasoning Capabilities in Large Language Models: Frameworks, Methods, and Evaluation
This survey reveals that LLMs excel in basic logical tasks but face challenges with complex multi-step reasoning, indicating a need for targeted research in specific reasoning types.
Evaluating Physical Concept Understanding in LLMs Through Abstract Grid-Based Tasks
This study presents a structured assessment framework for evaluating physics understanding in LLMs, utilizing a test suite that spans mechanics, thermodynamics, and electromagnetism, with both quantitative and qualitative questions.
torch.compile using hidet compiler
Hidet serves as an alternative backend to torch inductor for torch.compile, potentially enhancing performance and flexibility in PyTorch applications.