Here are the top ML papers of the week (Dec 11 - Dec 17): 1. LLMs for Discoveries in Mathematical Sciences - Uses LLMs to find new solutions in math and computer science, emphasizing the importance of safeguarding against LLM hallucinations. 2. Weak-to-Strong Generalization - Explores whether weak model supervision can unlock the full potential of stronger models, achieving impressive results. 3. Audiobox - A unified model for generating various audio modalities, with innovative prompting techniques and infilling objectives. 4. Mathematical LLMs - A survey on LLM progress in mathematical tasks, covering prompting techniques and math problem-solving. 5. Transparent Open-Source LLMs - Introduces LLM360 to make LLM training transparent and reproducible, releasing 7B parameter LLMs. 6. LLMs in Medicine - A comprehensive survey of LLMs in medicine, analyzing 300+ papers and their applications. 7. Beyond Human Data for LLMs - Proposes self-training with feedback to reduce dependence on human-generated data for LLMs. 8. Gaussian-SLAM - A neural RGBD SLAM method for photorealistic scene reconstruction without sacrificing speed. 9. Pearl - Introduces a production-ready RL agent software package for adaptive AI agents in challenging environments. 10. Quip - Compresses trained model weights into lower precision formats, bridging the gap between quantized LLMs and unquantized models. Explore these groundbreaking papers in machine learning!