ArXiv paper demonstrates FurnitureVLA, a vision-language-action model for learning long-horizon bimanual furniture assembly tasks. Advances multi-step robotic reasoning.
ArXiv paper presents AutoMem, a method for automated learning of memory mechanisms as a core cognitive skill in AI systems. Addresses memory optimization in neural architectures.
New ArXiv paper empirically measures the gap between research ideas generated by humans versus LLMs. Provides quantitative analysis of LLM research capability limitations.
ArXiv paper demonstrates that training a single transformer layer can match full-parameter RL training performance. Challenges conventional assumptions about model depth requirements.
arXiv announced it will spin out from Cornell University after 25 years to become an independent nonprofit, with major funding from Simons Foundation and Schmidt Sciences.