The scaling of inference computation has unlocked the potential of long-context large language models (LLMs) across diverse settings. For knowledge-intensive tasks, the increased compute is often allocated to incorporate more external knowledge. However, without effectively utilizing such knowledge, solely expanding context does not always enhance performance. In this work, we investigate inference scaling for retrieval augmented generation (RAG), exploring the combination of multiple strategies beyond simply increasing the quantity of knowledge, including in-context learning and iterative prompting. These strategies provide additional flexibility to scale test-time computation (e.g., by increasing retrieved documents or generation steps), thereby enhancing LLMs’ ability to effectively acquire and utilize contextual information. We address two key questions: (1) How does RAG performance benefit from the scaling of inference computation when optimally configured? (2) Can we predict the optimal test-time compute allocation for a given budget by modeling the relationship between RAG performance and inference parameters? Our observations reveal that increasing inference computation leads to nearly linear gains in RAG performance when optimally allocated, a relationship we describe as the inference scaling laws for RAG. Building on this, we further develop the computation allocation model to estimate RAG performance across different inference configurations. The model predicts optimal inference parameters under various computation constraints, which align closely with the experimental results. By applying these optimal configurations, we demonstrate that scaling inference compute on long-context LLMs achieves up to 58.9% gains on benchmark datasets compared to standard RAG.

As Learning-to-Rank (LTR) approaches primarily seek to improve ranking quality, their output scores are not scale-calibrated by design. This fundamentally limits LTR usage in score-sensitive applications. Though a simple multi-objective approach that combines a regression and a ranking objective can effectively learn scale-calibrated scores, we argue that the two objectives are not necessarily compatible, which makes the trade-off less ideal for either of them. In this paper, we propose a practical regression compatible ranking (RCR) approach that achieves a better trade-off, where the two ranking and regression components are proved to be mutually aligned. Although the same idea applies to ranking with both binary and graded relevance, we mainly focus on binary labels in this paper. We evaluate the proposed approach on several public LTR benchmarks and show that it consistently achieves either best or competitive result in terms of both regression and ranking metrics, and significantly improves the Pareto frontiers in the context of multi-objective optimization. Furthermore, we evaluated the proposed approach on YouTube Search and found that it not only improved the ranking quality of the production pCTR model, but also brought gains to the click prediction accuracy. The proposed approach has been successfully deployed in the YouTube production system.

The distillation of ranking models has become an important topic in both academia and industry. In recent years, several advanced methods have been proposed to tackle this problem, often leveraging ranking information from teacher rankers that is absent in traditional classification settings. To date, there is no well-established consensus on how to evaluate this class of models. Moreover, inconsistent benchmarking on a wide range of tasks and datasets make it difficult to assess or invigorate advances in this field. This paper first examines representative prior arts on ranking distillation, and raises three questions to be answered around methodology and reproducibility. To that end, we propose a systematic and unified benchmark, Ranking Distillation Suite (RD-Suite), which is a suite of tasks with 4 large real-world datasets, encompassing two major modalities (textual and numeric) and two applications (standard distillation and distillation transfer). RD-Suite consists of benchmark results that challenge some of the common wisdom in the field, and the release of datasets with teacher scores and evaluation scripts for future research. RD-Suite paves the way towards better understanding of ranking distillation, facilities more research in this direction, and presents new challenges.

In the context of hierarchical reinforcement learning, the idea of hierarchies of abstract machines (HAMs) is to write a partial policy as a set of hierarchical finite state machines with unspecified choice states, and use reinforcement learning to learn an optimal completion of this partial policy. Given a HAM with potentially deep hierarchical structure, there often exist many internal transitions where a machine calls another machine with the environment state unchanged. In this paper, we propose a new hierarchical reinforcement learning algorithm that discovers such internal transitions automatically, and shortcircuits them recursively in computation of Q values. The resulting HAMQ-INT algorithm outperforms the state of the art significantly on the benchmark Taxi domain and a much more complex RoboCup Keepaway domain.

This paper provides a novel POMDP planning method, called Palm LEAf SEarch (PLEASE), which allows the selection of more than one outcome when their potential impacts are close to the highest one during its forward exploration. Compared with existing trial-based algorithms, PLEASE can save considerable time to propagate the bound improvements of beliefs in deep levels of the search tree to the root belief because of fewer backup operations. Experiments showed that PLEASE scales up SARSOP, one of the fastest algorithms, by orders of magnitude on some POMDP tasks with large observation spaces.

Markov decision processes (MDPs) provide a rich framework for planning under uncertainty. However, exactly solving a large MDP is usually intractable due to the “curse of dimensionality”— the state space grows exponentially with the number of state variables. Online algorithms tackle this problem by avoiding computing a policy for the entire state space. On the other hand, since online algorithm has to find a near-optimal action online in almost real time, the computation time is often very limited. In the context of reinforcement learning, MAXQ is a value function decomposition method that exploits the underlying structure of the original MDP and decomposes it into a combination of smaller subproblems arranged over a task hierarchy. In this article, we present MAXQ-OP—a novel online planning algorithm for large MDPs that utilizes MAXQ hierarchical decomposition in online settings. Compared to traditional online planning algorithms, MAXQ-OP is able to reach much more deeper states in the search tree with relatively less computation time by exploiting MAXQ hierarchical decomposition online. We empirically evaluate our algorithm in the standard Taxi domain—a common benchmark for MDPs—to show the effectiveness of our approach. We have also conducted a long-term case study in a highly complex simulated soccer domain and developed a team named WrightEagle that has won five world champions and five runners-up in the recent 10 years of RoboCup Soccer Simulation 2D annual competitions. The results in the RoboCup domain confirm the scalability of MAXQ-OP to very large domains.

Monte-Carlo tree search is drawing great interest in the domain of planning under uncertainty, particularly when little or no domain knowledge is available. One of the central problems is the trade-off between exploration and exploitation. In this paper we present a novel Bayesian mixture modelling and inference based Thompson sampling approach to addressing this dilemma. The proposed Dirichlet-NormalGamma MCTS (DNG-MCTS) algorithm represents the uncertainty of the accumulated reward for actions in the MCTS search tree as a mixture of Normal distributions and inferences on it in Bayesian settings by choosing conjugate priors in the form of combinations of Dirichlet and NormalGamma distributions. Thompson sampling is used to select the best action at each decision node. Experimental results show that our proposed algorithm has achieved the state-of-the-art comparing with popular UCT algorithm in the context of online planning for general Markov decision processes.