Search papers, labs, and topics across Lattice.
4
0
8
Anchoring contrastive learning to valid Markov transition kernels could revolutionize how we model dynamic systems in AI.
Naive RL in recommender systems suffers from biased gradients that favor longer paths, but ProRL fixes this with a novel reward centering and advantage estimation scheme.
A diffusion model can generate high-quality synthetic chromosome images, boosting anomaly detection by nearly 14% F1 score and reducing reliance on scarce real-world abnormal samples.
Stop training LLMs on lucky guesses: this new RL method uses the model's own in-context learning ability to identify and upweight high-quality reasoning traces, leading to better performance.
