The chain structure is also convenient to handle streaming data as we will explain later. We first give a brief introduction to the EP and CEP framework. Step 2. We construct a tilted distribution to combine the true likelihood, Step 3. We project the tilted distribution back to the exponential family, q KL( null p nullq) where q belongs to the exponential family. Step 4. We update the approximation term by's in parallel, and uses damping to avoid divergence. The above computation are very conveniently to implement.

Neural Information Processing Systems http://nips.cc/

Tensor decomposition is an important tool for multiway data analysis.

However, it is often argued that correct predictions in the tail are more "interesting" or "rewarding," but the community has not yet settled on a metric capturing this intuitive concept.

In this work, we introduce ChatQA, a suite of models that outperform GPT -4 on retrieval-augmented generation (RAG) and conversational question answering (QA).

Large language models are increasingly trained in continual or open-ended settings, where the total training horizon is not known in advance. Despite this, most existing pretraining recipes are not anytime: they rely on horizon-dependent learning rate schedules and extensive tuning under a fixed compute budget. In this work, we provide a theoretical analysis demonstrating the existence of anytime learning schedules for overparameterized linear regression, and we highlight the central role of weight averaging - also known as model merging - in achieving the minimax convergence rates of stochastic gradient descent. We show that these anytime schedules polynomially decay with time, with the decay rate determined by the source and capacity conditions of the problem. Empirically, we evaluate 150M and 300M parameter language models trained at 1-32x Chinchilla scale, comparing constant learning rates with weight averaging and $1/\sqrt{t}$ schedules with weight averaging against a well-tuned cosine schedule. Across the full training range, the anytime schedules achieve comparable final loss to cosine decay. Taken together, our results suggest that weight averaging combined with simple, horizon-free step sizes offers a practical and effective anytime alternative to cosine learning rate schedules for large language model pretraining.