Search papers, labs, and topics across Lattice.
University of California
4
0
6
LRMs allocate more resources to problems they get wrong, while humans engage more with problems they expect to solve, revealing a fundamental difference in reasoning strategies.
Dynamic scheduling in SegFold unlocks nearly 2x speedup over traditional SpGEMM accelerators by optimizing data reuse and load balance.
SkillGrad's gradient-descent-inspired approach makes optimizing LLM agent skills as straightforward as training neural networks, achieving a 6.7% improvement over training-based baselines.
LLMs' impressive reasoning may be an illusion: truncating their reasoning steps reveals memorization-based shortcuts that current contamination detection methods miss.
