Why can't transformers learn multiplication?(arxiv.org)139 points by PaulHoule 1 year ago | 80 comments
Representing web applications as knowledge graphs(arxiv.org)66 points by PaulHoule 2 years ago | 7 comments
Controllable Fast and Slow Thinking by Learning with Randomized Reasoning Traces(arxiv.org)30 points by hislaziness 2 years ago | 12 comments
ChunkLLM: A Lightweight Pluggable Framework for Accelerating LLMs Inference(arxiv.org)85 points by PaulHoule 1 year ago | 6 comments
Universal optimality of Dijkstra via beyond-worst-case heaps(arxiv.org)182 points by foweltschmerz 2 years ago | 43 comments
Numerical Precision Affects Mathematical Reasoning Capabilities of LLMs(arxiv.org)55 points by belter 2 years ago | 39 comments