Aviral Kumar

Mitsuhiko Nakamoto, Yuexiang Zhai, Anikait Singh et al. · stanford

A compelling use case of offline reinforcement learning (RL) is to obtain a policy initialization from existing datasets followed by fast online fine-tuning with limited interaction. However, existing offline RL methods tend to behave poorly during fine-tuning. In this paper, we devise an approach for learning an effective initialization from offline data that also enables fast online fine-tuning capabilities. Our approach, calibrated Q-learning (Cal-QL), accomplishes this by learning a conservative value function initialization that underestimates the value of the learned policy from offline data, while also being calibrated, in the sense that the learned Q-values are at a reasonable scale. We refer to this property as calibration, and define it formally as providing a lower bound on the true value function of the learned policy and an upper bound on the value of some other (suboptimal) reference policy, which may simply be the behavior policy. We show that offline RL algorithms that learn such calibrated value functions lead to effective online fine-tuning, enabling us to take the benefits of offline initializations in online fine-tuning. In practice, Cal-QL can be implemented on top of the conservative Q learning (CQL) for offline RL within a one-line code change. Empirically, Cal-QL outperforms state-of-the-art methods on 9/11 fine-tuning benchmark tasks that we study in this paper. Code and video are available at https://nakamotoo.github.io/Cal-QL

Aviral Kumar, Joey Hong, Anikait Singh et al. · stanford

Offline reinforcement learning (RL) algorithms can acquire effective policies by utilizing previously collected experience, without any online interaction. It is widely understood that offline RL is able to extract good policies even from highly suboptimal data, a scenario where imitation learning finds suboptimal solutions that do not improve over the demonstrator that generated the dataset. However, another common use case for practitioners is to learn from data that resembles demonstrations. In this case, one can choose to apply offline RL, but can also use behavioral cloning (BC) algorithms, which mimic a subset of the dataset via supervised learning. Therefore, it seems natural to ask: when can an offline RL method outperform BC with an equal amount of expert data, even when BC is a natural choice? To answer this question, we characterize the properties of environments that allow offline RL methods to perform better than BC methods, even when only provided with expert data. Additionally, we show that policies trained on sufficiently noisy suboptimal data can attain better performance than even BC algorithms with expert data, especially on long-horizon problems. We validate our theoretical results via extensive experiments on both diagnostic and high-dimensional domains including robotic manipulation, maze navigation, and Atari games, with a variety of data distributions. We observe that, under specific but common conditions such as sparse rewards or noisy data sources, modern offline RL methods can significantly outperform BC.

Aviral Kumar, Anikait Singh, Frederik Ebert et al. · stanford

Progress in deep learning highlights the tremendous potential of utilizing diverse robotic datasets for attaining effective generalization and makes it enticing to consider leveraging broad datasets for attaining robust generalization in robotic learning as well. However, in practice, we often want to learn a new skill in a new environment that is unlikely to be contained in the prior data. Therefore we ask: how can we leverage existing diverse offline datasets in combination with small amounts of task-specific data to solve new tasks, while still enjoying the generalization benefits of training on large amounts of data? In this paper, we demonstrate that end-to-end offline RL can be an effective approach for doing this, without the need for any representation learning or vision-based pre-training. We present pre-training for robots (PTR), a framework based on offline RL that attempts to effectively learn new tasks by combining pre-training on existing robotic datasets with rapid fine-tuning on a new task, with as few as 10 demonstrations. PTR utilizes an existing offline RL method, conservative Q-learning (CQL), but extends it to include several crucial design decisions that enable PTR to actually work and outperform a variety of prior methods. To our knowledge, PTR is the first RL method that succeeds at learning new tasks in a new domain on a real WidowX robot with as few as 10 task demonstrations, by effectively leveraging an existing dataset of diverse multi-task robot data collected in a variety of toy kitchens. We also demonstrate that PTR can enable effective autonomous fine-tuning and improvement in a handful of trials, without needing any demonstrations. An accompanying overview video can be found in the supplementary material and at thi URL: https://sites.google.com/view/ptr-final/

Aviral Kumar, Rishabh Agarwal, Xinyang Geng et al. · deepmind

The potential of offline reinforcement learning (RL) is that high-capacity models trained on large, heterogeneous datasets can lead to agents that generalize broadly, analogously to similar advances in vision and NLP. However, recent works argue that offline RL methods encounter unique challenges to scaling up model capacity. Drawing on the learnings from these works, we re-examine previous design choices and find that with appropriate choices: ResNets, cross-entropy based distributional backups, and feature normalization, offline Q-learning algorithms exhibit strong performance that scales with model capacity. Using multi-task Atari as a testbed for scaling and generalization, we train a single policy on 40 games with near-human performance using up-to 80 million parameter networks, finding that model performance scales favorably with capacity. In contrast to prior work, we extrapolate beyond dataset performance even when trained entirely on a large (400M transitions) but highly suboptimal dataset (51% human-level performance). Compared to return-conditioned supervised approaches, offline Q-learning scales similarly with model capacity and has better performance, especially when the dataset is suboptimal. Finally, we show that offline Q-learning with a diverse dataset is sufficient to learn powerful representations that facilitate rapid transfer to novel games and fast online learning on new variations of a training game, improving over existing state-of-the-art representation learning approaches.

Qiyang Li, Aviral Kumar, Ilya Kostrikov et al. · berkeley

Deep reinforcement learning algorithms that learn policies by trial-and-error must learn from limited amounts of data collected by actively interacting with the environment. While many prior works have shown that proper regularization techniques are crucial for enabling data-efficient RL, a general understanding of the bottlenecks in data-efficient RL has remained unclear. Consequently, it has been difficult to devise a universal technique that works well across all domains. In this paper, we attempt to understand the primary bottleneck in sample-efficient deep RL by examining several potential hypotheses such as non-stationarity, excessive action distribution shift, and overfitting. We perform thorough empirical analysis on state-based DeepMind control suite (DMC) tasks in a controlled and systematic way to show that high temporal-difference (TD) error on the validation set of transitions is the main culprit that severely affects the performance of deep RL algorithms, and prior methods that lead to good performance do in fact, control the validation TD error to be low. This observation gives us a robust principle for making deep RL efficient: we can hill-climb on the validation TD error by utilizing any form of regularization techniques from supervised learning. We show that a simple online model selection method that targets the validation TD error is effective across state-based DMC and Gym tasks.

Anikait Singh, Aviral Kumar, Quan Vuong et al. · stanford

Offline reinforcement learning (RL) learns policies entirely from static datasets, thereby avoiding the challenges associated with online data collection. Practical applications of offline RL will inevitably require learning from datasets where the variability of demonstrated behaviors changes non-uniformly across the state space. For example, at a red light, nearly all human drivers behave similarly by stopping, but when merging onto a highway, some drivers merge quickly, efficiently, and safely, while many hesitate or merge dangerously. Both theoretically and empirically, we show that typical offline RL methods, which are based on distribution constraints fail to learn from data with such non-uniform variability, due to the requirement to stay close to the behavior policy to the same extent across the state space. Ideally, the learned policy should be free to choose per state how closely to follow the behavior policy to maximize long-term return, as long as the learned policy stays within the support of the behavior policy. To instantiate this principle, we reweight the data distribution in conservative Q-learning (CQL) to obtain an approximate support constraint formulation. The reweighted distribution is a mixture of the current policy and an additional policy trained to mine poor actions that are likely under the behavior policy. Our method, CQL (ReDS), is simple, theoretically motivated, and improves performance across a wide range of offline RL problems in Atari games, navigation, and pixel-based manipulation.

Rafael Rafailov, Kyle Hatch, Anikait Singh et al. · berkeley, stanford

Offline reinforcement learning algorithms hold the promise of enabling data-driven RL methods that do not require costly or dangerous real-world exploration and benefit from large pre-collected datasets. This in turn can facilitate real-world applications, as well as a more standardized approach to RL research. Furthermore, offline RL methods can provide effective initializations for online finetuning to overcome challenges with exploration. However, evaluating progress on offline RL algorithms requires effective and challenging benchmarks that capture properties of real-world tasks, provide a range of task difficulties, and cover a range of challenges both in terms of the parameters of the domain (e.g., length of the horizon, sparsity of rewards) and the parameters of the data (e.g., narrow demonstration data or broad exploratory data). While considerable progress in offline RL in recent years has been enabled by simpler benchmark tasks, the most widely used datasets are increasingly saturating in performance and may fail to reflect properties of realistic tasks. We propose a new benchmark for offline RL that focuses on realistic simulations of robotic manipulation and locomotion environments, based on models of real-world robotic systems, and comprising a variety of data sources, including scripted data, play-style data collected by human teleoperators, and other data sources. Our proposed benchmark covers state-based and image-based domains, and supports both offline RL and online fine-tuning evaluation, with some of the tasks specifically designed to require both pre-training and fine-tuning. We hope that our proposed benchmark will facilitate further progress on both offline RL and fine-tuning algorithms. Website with code, examples, tasks, and data is available at \url{https://sites.google.com/view/d5rl/}

Aviral Kumar, Vincent Zhuang, Rishabh Agarwal et al. · deepmind

Self-correction is a highly desirable capability of large language models (LLMs), yet it has consistently been found to be largely ineffective in modern LLMs. Current methods for training self-correction typically depend on either multiple models, a more advanced model, or additional forms of supervision. To address these shortcomings, we develop a multi-turn online reinforcement learning (RL) approach, SCoRe, that significantly improves an LLM's self-correction ability using entirely self-generated data. To build SCoRe, we first show that variants of supervised fine-tuning (SFT) on offline model-generated correction traces are often insufficient for instilling self-correction behavior. In particular, we observe that training via SFT falls prey to either a distribution mismatch between mistakes made by the data-collection policy and the model's own responses, or to behavior collapse, where learning implicitly prefers only a certain mode of correction behavior that is often not effective at self-correction on test problems. SCoRe addresses these challenges by training under the model's own distribution of self-generated correction traces and using appropriate regularization to steer the learning process into learning a self-correction behavior that is effective at test time as opposed to fitting high-reward responses for a given prompt. This regularization process includes an initial phase of multi-turn RL on a base model to generate a policy initialization that is less susceptible to collapse, followed by using a reward bonus to amplify self-correction. With Gemini 1.0 Pro and 1.5 Flash models, we find that SCoRe achieves state-of-the-art self-correction performance, improving the base models' self-correction by 15.6% and 9.1% respectively on MATH and HumanEval.

Chethan Bhateja, Derek Guo, Dibya Ghosh et al. · berkeley, stanford

Pre-training on Internet data has proven to be a key ingredient for broad generalization in many modern ML systems. What would it take to enable such capabilities in robotic reinforcement learning (RL)? Offline RL methods, which learn from datasets of robot experience, offer one way to leverage prior data into the robotic learning pipeline. However, these methods have a "type mismatch" with video data (such as Ego4D), the largest prior datasets available for robotics, since video offers observation-only experience without the action or reward annotations needed for RL methods. In this paper, we develop a system for leveraging large-scale human video datasets in robotic offline RL, based entirely on learning value functions via temporal-difference learning. We show that value learning on video datasets learns representations that are more conducive to downstream robotic offline RL than other approaches for learning from video data. Our system, called V-PTR, combines the benefits of pre-training on video data with robotic offline RL approaches that train on diverse robot data, resulting in value functions and policies for manipulation tasks that perform better, act robustly, and generalize broadly. On several manipulation tasks on a real WidowX robot, our framework produces policies that greatly improve over prior methods. Our video and additional details can be found at https://dibyaghosh.com/vptr/

Zhang-Wei Hong, Aviral Kumar, Sathwik Karnik et al.

Offline policy learning is aimed at learning decision-making policies using existing datasets of trajectories without collecting additional data. The primary motivation for using reinforcement learning (RL) instead of supervised learning techniques such as behavior cloning is to find a policy that achieves a higher average return than the trajectories constituting the dataset. However, we empirically find that when a dataset is dominated by suboptimal trajectories, state-of-the-art offline RL algorithms do not substantially improve over the average return of trajectories in the dataset. We argue this is due to an assumption made by current offline RL algorithms of staying close to the trajectories in the dataset. If the dataset primarily consists of sub-optimal trajectories, this assumption forces the policy to mimic the suboptimal actions. We overcome this issue by proposing a sampling strategy that enables the policy to only be constrained to ``good data" rather than all actions in the dataset (i.e., uniform sampling). We present a realization of the sampling strategy and an algorithm that can be used as a plug-and-play module in standard offline RL algorithms. Our evaluation demonstrates significant performance gains in 72 imbalanced datasets, D4RL dataset, and across three different offline RL algorithms. Code is available at https://github.com/Improbable-AI/dw-offline-rl.

Yevgen Chebotar, Quan Vuong, Alex Irpan et al.

In this work, we present a scalable reinforcement learning method for training multi-task policies from large offline datasets that can leverage both human demonstrations and autonomously collected data. Our method uses a Transformer to provide a scalable representation for Q-functions trained via offline temporal difference backups. We therefore refer to the method as Q-Transformer. By discretizing each action dimension and representing the Q-value of each action dimension as separate tokens, we can apply effective high-capacity sequence modeling techniques for Q-learning. We present several design decisions that enable good performance with offline RL training, and show that Q-Transformer outperforms prior offline RL algorithms and imitation learning techniques on a large diverse real-world robotic manipulation task suite. The project's website and videos can be found at https://qtransformer.github.io

Homer Walke, Jonathan Yang, Albert Yu et al.

Reinforcement learning (RL) algorithms hold the promise of enabling autonomous skill acquisition for robotic systems. However, in practice, real-world robotic RL typically requires time consuming data collection and frequent human intervention to reset the environment. Moreover, robotic policies learned with RL often fail when deployed beyond the carefully controlled setting in which they were learned. In this work, we study how these challenges can all be tackled by effective utilization of diverse offline datasets collected from previously seen tasks. When faced with a new task, our system adapts previously learned skills to quickly learn to both perform the new task and return the environment to an initial state, effectively performing its own environment reset. Our empirical results demonstrate that incorporating prior data into robotic reinforcement learning enables autonomous learning, substantially improves sample-efficiency of learning, and enables better generalization. Project website: https://sites.google.com/view/ariel-berkeley/

Han Qi, Yi Su, Aviral Kumar et al.

The goal in offline data-driven decision-making is synthesize decisions that optimize a black-box utility function, using a previously-collected static dataset, with no active interaction. These problems appear in many forms: offline reinforcement learning (RL), where we must produce actions that optimize the long-term reward, bandits from logged data, where the goal is to determine the correct arm, and offline model-based optimization (MBO) problems, where we must find the optimal design provided access to only a static dataset. A key challenge in all these settings is distributional shift: when we optimize with respect to the input into a model trained from offline data, it is easy to produce an out-of-distribution (OOD) input that appears erroneously good. In contrast to prior approaches that utilize pessimism or conservatism to tackle this problem, in this paper, we formulate offline data-driven decision-making as domain adaptation, where the goal is to make accurate predictions for the value of optimized decisions ("target domain"), when training only on the dataset ("source domain"). This perspective leads to invariant objective models (IOM), our approach for addressing distributional shift by enforcing invariance between the learned representations of the training dataset and optimized decisions. In IOM, if the optimized decisions are too different from the training dataset, the representation will be forced to lose much of the information that distinguishes good designs from bad ones, making all choices seem mediocre. Critically, when the optimizer is aware of this representational tradeoff, it should choose not to stray too far from the training distribution, leading to a natural trade-off between distributional shift and learning performance.

Joey Hong, Aviral Kumar, Sergey Levine

Offline reinforcement learning (RL) promises the ability to learn effective policies solely using existing, static datasets, without any costly online interaction. To do so, offline RL methods must handle distributional shift between the dataset and the learned policy. The most common approach is to learn conservative, or lower-bound, value functions, which underestimate the return of out-of-distribution (OOD) actions. However, such methods exhibit one notable drawback: policies optimized on such value functions can only behave according to a fixed, possibly suboptimal, degree of conservatism. However, this can be alleviated if we instead are able to learn policies for varying degrees of conservatism at training time and devise a method to dynamically choose one of them during evaluation. To do so, in this work, we propose learning value functions that additionally condition on the degree of conservatism, which we dub confidence-conditioned value functions. We derive a new form of a Bellman backup that simultaneously learns Q-values for any degree of confidence with high probability. By conditioning on confidence, our value functions enable adaptive strategies during online evaluation by controlling for confidence level using the history of observations thus far. This approach can be implemented in practice by conditioning the Q-function from existing conservative algorithms on the confidence.We theoretically show that our learned value functions produce conservative estimates of the true value at any desired confidence. Finally, we empirically show that our algorithm outperforms existing conservative offline RL algorithms on multiple discrete control domains.

LM-Provers, Yuxiao Qu, Amrith Setlur et al.

Proprietary AI systems have recently demonstrated impressive capabilities on complex proof-based problems, with gold-level performance reported at the 2025 International Mathematical Olympiad (IMO). However, the training pipelines behind these systems remain largely undisclosed, and their reliance on large "internal" models and scaffolds makes them expensive to run, difficult to reproduce, and hard to study or improve upon. This raises a central question: can small, open models also be trained to achieve competitive reasoning performance on difficult Olympiad-level math? In this paper, we answer this question by building QED-Nano, a 4B model post-trained for Olympiad-level proofs. Our training recipe has three stages: (1) supervised fine-tuning to imbue good proof-writing styles by distilling from DeepSeek-Math-V2, (2) reinforcement learning (RL) with rubric-based rewards, and (3) expanding RL with a reasoning cache, which decomposes long proofs into iterative summarize-and-refine cycles and enables stronger test-time reasoning. QED-Nano surpasses the proof-generation performance of much larger open models, including Nomos-1 and GPT-OSS-120B, and approaches the performance of proprietary models like Gemini 3 Pro, at a fraction of the inference cost. To support further research on open mathematical reasoning, we release the full QED-Nano pipeline, including the QED-Nano and QED-Nano-SFT models, the FineProofs-SFT and FineProofs-RL datasets, and the training and evaluation code.

Jianlan Luo, Perry Dong, Jeffrey Wu et al.

The offline reinforcement learning (RL) paradigm provides a general recipe to convert static behavior datasets into policies that can perform better than the policy that collected the data. While policy constraints, conservatism, and other methods for mitigating distributional shifts have made offline reinforcement learning more effective, the continuous action setting often necessitates various approximations for applying these techniques. Many of these challenges are greatly alleviated in discrete action settings, where offline RL constraints and regularizers can often be computed more precisely or even exactly. In this paper, we propose an adaptive scheme for action quantization. We use a VQ-VAE to learn state-conditioned action quantization, avoiding the exponential blowup that comes with naïve discretization of the action space. We show that several state-of-the-art offline RL methods such as IQL, CQL, and BRAC improve in performance on benchmarks when combined with our proposed discretization scheme. We further validate our approach on a set of challenging long-horizon complex robotic manipulation tasks in the Robomimic environment, where our discretized offline RL algorithms are able to improve upon their continuous counterparts by 2-3x. Our project page is at https://saqrl.github.io/

Quan Vuong, Aviral Kumar, Sergey Levine et al.

In offline RL, constraining the learned policy to remain close to the data is essential to prevent the policy from outputting out-of-distribution (OOD) actions with erroneously overestimated values. In principle, generative adversarial networks (GAN) can provide an elegant solution to do so, with the discriminator directly providing a probability that quantifies distributional shift. However, in practice, GAN-based offline RL methods have not performed as well as alternative approaches, perhaps because the generator is trained to both fool the discriminator and maximize return -- two objectives that can be at odds with each other. In this paper, we show that the issue of conflicting objectives can be resolved by training two generators: one that maximizes return, with the other capturing the ``remainder'' of the data distribution in the offline dataset, such that the mixture of the two is close to the behavior policy. We show that not only does having two generators enable an effective GAN-based offline RL method, but also approximates a support constraint, where the policy does not need to match the entire data distribution, but only the slice of the data that leads to high long term performance. We name our method DASCO, for Dual-Generator Adversarial Support Constrained Offline RL. On benchmark tasks that require learning from sub-optimal data, DASCO significantly outperforms prior methods that enforce distribution constraint.

Lunjun Zhang, Arian Hosseini, Hritik Bansal et al.

Verifiers or reward models are often used to enhance the reasoning performance of large language models (LLMs). A common approach is the Best-of-N method, where N candidate solutions generated by the LLM are ranked by a verifier, and the best one is selected. While LLM-based verifiers are typically trained as discriminative classifiers to score solutions, they do not utilize the text generation capabilities of pretrained LLMs. To overcome this limitation, we instead propose training verifiers using the ubiquitous next-token prediction objective, jointly on verification and solution generation. Compared to standard verifiers, such generative verifiers (GenRM) can benefit from several advantages of LLMs: they integrate seamlessly with instruction tuning, enable chain-of-thought reasoning, and can utilize additional test-time compute via majority voting for better verification. We demonstrate that GenRM outperforms discriminative, DPO verifiers, and LLM-as-a-Judge, resulting in large performance gains with Best-of-N, namely 5% $\rightarrow$ 45.3% on algorithmic tasks and 73% $\rightarrow$ 93.4% on GSM8K. In easy-to-hard generalization settings, we observe improvements of 28% $\rightarrow$ 44.6% on MATH, and 37.9% $\rightarrow$ 53.5% on MMLU abstract algebra. Furthermore, we find that training GenRM with synthetic verification rationales is sufficient to pick out subtle errors on math problems. Finally, we demonstrate that GenRM scales favorably with model size and test-time compute.

Charlie Snell, Jaehoon Lee, Kelvin Xu et al.

Enabling LLMs to improve their outputs by using more test-time computation is a critical step towards building generally self-improving agents that can operate on open-ended natural language. In this paper, we study the scaling of inference-time computation in LLMs, with a focus on answering the question: if an LLM is allowed to use a fixed but non-trivial amount of inference-time compute, how much can it improve its performance on a challenging prompt? Answering this question has implications not only on the achievable performance of LLMs, but also on the future of LLM pretraining and how one should tradeoff inference-time and pre-training compute. Despite its importance, little research attempted to understand the scaling behaviors of various test-time inference methods. Moreover, current work largely provides negative results for a number of these strategies. In this work, we analyze two primary mechanisms to scale test-time computation: (1) searching against dense, process-based verifier reward models; and (2) updating the model's distribution over a response adaptively, given the prompt at test time. We find that in both cases, the effectiveness of different approaches to scaling test-time compute critically varies depending on the difficulty of the prompt. This observation motivates applying a "compute-optimal" scaling strategy, which acts to most effectively allocate test-time compute adaptively per prompt. Using this compute-optimal strategy, we can improve the efficiency of test-time compute scaling by more than 4x compared to a best-of-N baseline. Additionally, in a FLOPs-matched evaluation, we find that on problems where a smaller base model attains somewhat non-trivial success rates, test-time compute can be used to outperform a 14x larger model.

Han Qi, Xinyang Geng, Stefano Rando et al.

In computational chemistry, crystal structure prediction (CSP) is an optimization problem that involves discovering the lowest energy stable crystal structure for a given chemical formula. This problem is challenging as it requires discovering globally optimal designs with the lowest energies on complex manifolds. One approach to tackle this problem involves building simulators based on density functional theory (DFT) followed by running search in simulation, but these simulators are painfully slow. In this paper, we study present and study an alternate, data-driven approach to crystal structure prediction: instead of directly searching for the most stable structures in simulation, we train a surrogate model of the crystal formation energy from a database of existing crystal structures, and then optimize this model with respect to the parameters of the crystal structure. This surrogate model is trained to be conservative so as to prevent exploitation of its errors by the optimizer. To handle optimization in the non-Euclidean space of crystal structures, we first utilize a state-of-the-art graph diffusion auto-encoder (CD-VAE) to convert a crystal structure into a vector-based search space and then optimize a conservative surrogate model of the crystal energy, trained on top of this vector representation. We show that our approach, dubbed LCOMs (latent conservative objective models), performs comparably to the best current approaches in terms of success rate of structure prediction, while also drastically reducing computational cost.

Yuxiao Qu, Tianjun Zhang, Naman Garg et al.

A central piece in enabling intelligent agentic behavior in foundation models is to make them capable of introspecting upon their behavior, reasoning, and correcting their mistakes as more computation or interaction is available. Even the strongest proprietary large language models (LLMs) do not quite exhibit the ability of continually improving their responses sequentially, even in scenarios where they are explicitly told that they are making a mistake. In this paper, we develop RISE: Recursive IntroSpEction, an approach for fine-tuning LLMs to introduce this capability, despite prior work hypothesizing that this capability may not be possible to attain. Our approach prescribes an iterative fine-tuning procedure, which attempts to teach the model how to alter its response after having executed previously unsuccessful attempts to solve a hard test-time problem, with optionally additional environment feedback. RISE poses fine-tuning for a single-turn prompt as solving a multi-turn Markov decision process (MDP), where the initial state is the prompt. Inspired by principles in online imitation learning and reinforcement learning, we propose strategies for multi-turn data collection and training so as to imbue an LLM with the capability to recursively detect and correct its previous mistakes in subsequent iterations. Our experiments show that RISE enables Llama2, Llama3, and Mistral models to improve themselves with more turns on math reasoning tasks, outperforming several single-turn strategies given an equal amount of inference-time computation. We also find that RISE scales well, often attaining larger benefits with more capable models. Our analysis shows that RISE makes meaningful improvements to responses to arrive at the correct solution for challenging prompts, without disrupting one-turn abilities as a result of expressing more complex distributions.

Mian Wu, Gavin Zhang, Sewon Min et al.

Open-ended generation tasks require outputs to satisfy diverse and often implicit task-specific evaluation rubrics. The sheer number of relevant rubrics leads to prohibitively high verification costs and incomplete assessments of a response, making reinforcement learning (RL) post-training with rubric-based rewards difficult to scale. This problem is exacerbated by the fact that often the best way to combine these rubrics into one single reward is also highly prompt-specific. We propose Reinforcement Learning with Adversarial Critic (RLAC), a post-training approach that addresses these challenges via dynamic rubric verification. Our approach employs a large language model (LLM) as a critic that dynamically identifies only the most likely failure modes (e.g., a factual error or unhandled edge case), which are then verified by an external validator to optimize both generator and critic jointly. By training both the generator and the critic, this game enhances the critic's error detection and the generator's output quality while reducing required verifications. Our experiments demonstrate that RLAC improves factual accuracy in text generation and correctness in code generation, while also outperforming exhaustive verification and reward model methods. We show that dynamic critics are more effective than fixed critics, showcasing the potential of RLAC for scaling RL post-training to free-form generation tasks.

Hao Bai, Alexey Taymanov, Tong Zhang et al.

We present WebGym, the largest-to-date open-source environment for training realistic visual web agents. Real websites are non-stationary and diverse, making artificial or small-scale task sets insufficient for robust policy learning. WebGym contains nearly 300,000 tasks with rubric-based evaluations across diverse, real-world websites and difficulty levels. We train agents with a simple reinforcement learning (RL) recipe, which trains on the agent's own interaction traces (rollouts), using task rewards as feedback to guide learning. To enable scaling RL, we speed up sampling of trajectories in WebGym by developing a high-throughput asynchronous rollout system, designed specifically for web agents. Our system achieves a 4-5x rollout speedup compared to naive implementations. Second, we scale the task set breadth, depth, and size, which results in continued performance improvement. Fine-tuning a strong base vision-language model, Qwen-3-VL-8B-Instruct, on WebGym results in an improvement in success rate on an out-of-distribution test set from 26.2% to 42.9%, significantly outperforming agents based on proprietary models such as GPT-4o and GPT-5-Thinking that achieve 27.1% and 29.8%, respectively. This improvement is substantial because our test set consists only of tasks on websites never seen during training, unlike many other prior works on training visual web agents.

Junhong Shen, Hao Bai, Lunjun Zhang et al. · cmu

The current paradigm of test-time scaling relies on generating long reasoning traces ("thinking" more) before producing a response. In agent problems that require interaction, this can be done by generating thinking traces before acting in the world. However, this process does not allow agents to acquire new information from the environment or adapt their behavior over time. In this work, we propose to scale test-time interaction, an untapped dimension of test-time scaling that increases the agent's interaction horizon to enable running rich behaviors such as exploration, backtracking, and dynamic re-planning within a single rollout. To demonstrate the promise of this scaling dimension, we study the domain of web agents. We first show that even prompting-based interaction scaling without any training can improve task success on web benchmarks non-trivially. Building on this, we introduce TTI (Test-Time Interaction), a curriculum-based online reinforcement learning (RL) approach that trains agents by adaptively adjusting their rollout lengths. Using a Gemma 3 12B model, TTI produces state-of-the-art open-source, open-data web agents on WebVoyager and WebArena benchmarks. We further show that TTI enables agents to balance exploration and exploitation adaptively. Our results establish interaction scaling as a powerful, complementary axis to scaling per-step compute, offering new avenues for training adaptive agents.

Seohong Park, Kevin Frans, Deepinder Mann et al.

In this work, we study the scalability of offline reinforcement learning (RL) algorithms. In principle, a truly scalable offline RL algorithm should be able to solve any given problem, regardless of its complexity, given sufficient data, compute, and model capacity. We investigate if and how current offline RL algorithms match up to this promise on diverse, challenging, previously unsolved tasks, using datasets up to 1000x larger than typical offline RL datasets. We observe that despite scaling up data, many existing offline RL algorithms exhibit poor scaling behavior, saturating well below the maximum performance. We hypothesize that the horizon is the main cause behind the poor scaling of offline RL. We empirically verify this hypothesis through several analysis experiments, showing that long horizons indeed present a fundamental barrier to scaling up offline RL. We then show that various horizon reduction techniques substantially enhance scalability on challenging tasks. Based on our insights, we also introduce a minimal yet scalable method named SHARSA that effectively reduces the horizon. SHARSA achieves the best asymptotic performance and scaling behavior among our evaluation methods, showing that explicitly reducing the horizon unlocks the scalability of offline RL. Code: https://github.com/seohongpark/horizon-reduction

Hao Bai, Yifei Zhou, Li Erran Li et al.

While a number of existing approaches for building foundation model agents rely on prompting or fine-tuning with human demonstrations, it is not sufficient in dynamic environments (e.g., mobile device control). On-policy reinforcement learning (RL) should address these limitations, but collecting actual rollouts in an environment is often undesirable in truly open-ended agentic problems such as mobile device control or interacting with humans, where each unit of interaction is associated with a cost. In such scenarios, a method for policy learning that can utilize off-policy experience by learning a trained action-value function is much more effective. In this paper, we develop an approach, called Digi-Q, to train VLM-based action-value Q-functions which are then used to extract the agent policy. We study our approach in the mobile device control setting. Digi-Q trains the Q-function using offline temporal-difference (TD) learning, on top of frozen, intermediate-layer features of a VLM. Compared to fine-tuning the whole VLM, this approach saves us compute and enhances scalability. To make the VLM features amenable for representing the Q-function, we need to employ an initial phase of fine-tuning to amplify coverage over actionable information needed for value function. Once trained, we use this Q-function via a Best-of-N policy extraction operator that imitates the best action out of multiple candidate actions from the current policy as ranked by the value function, enabling policy improvement without environment interaction. Digi-Q outperforms several prior methods on user-scale device control tasks in Android-in-the-Wild, attaining 21.2% improvement over prior best-performing method. In some cases, our Digi-Q approach already matches state-of-the-art RL methods that require interaction. The project is open-sourced at https://github.com/DigiRL-agent/digiq

Vansh Kapoor, Aman Gupta, Hao Chen et al.

Multi-step reasoning tasks like mathematical problem solving are vulnerable to cascading failures, where a single incorrect step leads to complete solution breakdown. Current LLM routing methods assign entire queries to one model, treating all reasoning steps as equal. We propose TRIM (Targeted routing in multi-step reasoning tasks), which routes only critical steps$\unicode{x2013}$those likely to derail the solution$\unicode{x2013}$to larger models while letting smaller models handle routine continuations. Our key insight is that targeted step-level interventions can fundamentally transform inference efficiency by confining expensive calls to precisely those steps where stronger models prevent cascading errors. TRIM operates at the step-level: it uses process reward models to identify erroneous steps and makes routing decisions based on step-level uncertainty and budget constraints. We develop several routing strategies within TRIM, ranging from a simple threshold-based policy to more expressive policies that reason about long-horizon accuracy-cost trade-offs and uncertainty in step-level correctness estimates. On MATH-500, even the simplest thresholding strategy surpasses prior routing methods with 5x higher cost efficiency, while more advanced policies match the strong, expensive model's performance using 80% fewer expensive model tokens. On harder benchmarks such as AIME, TRIM achieves up to 6x higher cost efficiency. All methods generalize effectively across math reasoning tasks, demonstrating that step-level difficulty represents fundamental characteristics of reasoning.

Fahim Tajwar, Anikait Singh, Archit Sharma et al. · stanford

Learning from preference labels plays a crucial role in fine-tuning large language models. There are several distinct approaches for preference fine-tuning, including supervised learning, on-policy reinforcement learning (RL), and contrastive learning. Different methods come with different implementation tradeoffs and performance differences, and existing empirical findings present different conclusions, for instance, some results show that online RL is quite important to attain good fine-tuning results, while others find (offline) contrastive or even purely supervised methods sufficient. This raises a natural question: what kind of approaches are important for fine-tuning with preference data and why? In this paper, we answer this question by performing a rigorous analysis of a number of fine-tuning techniques on didactic and full-scale LLM problems. Our main finding is that, in general, approaches that use on-policy sampling or attempt to push down the likelihood on certain responses (i.e., employ a "negative gradient") outperform offline and maximum likelihood objectives. We conceptualize our insights and unify methods that use on-policy sampling or negative gradient under a notion of mode-seeking objectives for categorical distributions. Mode-seeking objectives are able to alter probability mass on specific bins of a categorical distribution at a fast rate compared to maximum likelihood, allowing them to relocate masses across bins more effectively. Our analysis prescribes actionable insights for preference fine-tuning of LLMs and informs how data should be collected for maximal improvement.

Yifei Zhou, Andrea Zanette, Jiayi Pan et al. · berkeley

A broad use case of large language models (LLMs) is in goal-directed decision-making tasks (or "agent" tasks), where an LLM needs to not just generate completions for a given prompt, but rather make intelligent decisions over a multi-turn interaction to accomplish a task (e.g., when interacting with the web, using tools, or providing customer support). Reinforcement learning (RL) provides a general paradigm to address such agent tasks, but current RL methods for LLMs largely focus on optimizing single-turn rewards. By construction, most single-turn RL methods cannot endow LLMs with the ability to intelligently seek information over multiple turns, perform credit assignment, or reason about their past actions -- all of which are critical in agent tasks. This raises the question: how can we design effective and efficient multi-turn RL algorithms for LLMs? In this paper, we develop a framework for building multi-turn RL algorithms for fine-tuning LLMs, that preserves the flexibility of existing single-turn RL methods for LLMs (e.g., proximal policy optimization), while accommodating multiple turns, long horizons, and delayed rewards effectively. To do this, our framework adopts a hierarchical RL approach and runs two RL algorithms in parallel: a high-level off-policy value-based RL algorithm to aggregate reward over utterances, and a low-level RL algorithm that utilizes this high-level value function to train a token policy within each utterance or turn. Our hierarchical framework, Actor-Critic Framework with a Hierarchical Structure (ArCHer), can also give rise to other RL methods. Empirically, we find that ArCHer significantly improves efficiency and performance on agent tasks, attaining a sample efficiency of about 100x over existing methods, while also improving with larger model capacity (upto the 7 billion scale that we tested on).

Brandon Trabucco, Xinyang Geng, Aviral Kumar et al.

Black-box model-based optimization (MBO) problems, where the goal is to find a design input that maximizes an unknown objective function, are ubiquitous in a wide range of domains, such as the design of proteins, DNA sequences, aircraft, and robots. Solving model-based optimization problems typically requires actively querying the unknown objective function on design proposals, which means physically building the candidate molecule, aircraft, or robot, testing it, and storing the result. This process can be expensive and time consuming, and one might instead prefer to optimize for the best design using only the data one already has. This setting -- called offline MBO -- poses substantial and different algorithmic challenges than more commonly studied online techniques. A number of recent works have demonstrated success with offline MBO for high-dimensional optimization problems using high-capacity deep neural networks. However, the lack of standardized benchmarks in this emerging field is making progress difficult to track. To address this, we present Design-Bench, a benchmark for offline MBO with a unified evaluation protocol and reference implementations of recent methods. Our benchmark includes a suite of diverse and realistic tasks derived from real-world optimization problems in biology, materials science, and robotics that present distinct challenges for offline MBO. Our benchmark and reference implementations are released at github.com/rail-berkeley/design-bench and github.com/rail-berkeley/design-baselines.

Justin Fu, Mohammad Norouzi, Ofir Nachum et al.

Off-policy evaluation (OPE) holds the promise of being able to leverage large, offline datasets for both evaluating and selecting complex policies for decision making. The ability to learn offline is particularly important in many real-world domains, such as in healthcare, recommender systems, or robotics, where online data collection is an expensive and potentially dangerous process. Being able to accurately evaluate and select high-performing policies without requiring online interaction could yield significant benefits in safety, time, and cost for these applications. While many OPE methods have been proposed in recent years, comparing results between papers is difficult because currently there is a lack of a comprehensive and unified benchmark, and measuring algorithmic progress has been challenging due to the lack of difficult evaluation tasks. In order to address this gap, we present a collection of policies that in conjunction with existing offline datasets can be used for benchmarking off-policy evaluation. Our tasks include a range of challenging high-dimensional continuous control problems, with wide selections of datasets and policies for performing policy selection. The goal of our benchmark is to provide a standardized measure of progress that is motivated from a set of principles designed to challenge and test the limits of existing OPE methods. We perform an evaluation of state-of-the-art algorithms and provide open-source access to our data and code to foster future research in this area.

Justin Fu, Aviral Kumar, Ofir Nachum et al.

The offline reinforcement learning (RL) setting (also known as full batch RL), where a policy is learned from a static dataset, is compelling as progress enables RL methods to take advantage of large, previously-collected datasets, much like how the rise of large datasets has fueled results in supervised learning. However, existing online RL benchmarks are not tailored towards the offline setting and existing offline RL benchmarks are restricted to data generated by partially-trained agents, making progress in offline RL difficult to measure. In this work, we introduce benchmarks specifically designed for the offline setting, guided by key properties of datasets relevant to real-world applications of offline RL. With a focus on dataset collection, examples of such properties include: datasets generated via hand-designed controllers and human demonstrators, multitask datasets where an agent performs different tasks in the same environment, and datasets collected with mixtures of policies. By moving beyond simple benchmark tasks and data collected by partially-trained RL agents, we reveal important and unappreciated deficiencies of existing algorithms. To facilitate research, we have released our benchmark tasks and datasets with a comprehensive evaluation of existing algorithms, an evaluation protocol, and open-source examples. This serves as a common starting point for the community to identify shortcomings in existing offline RL methods and a collaborative route for progress in this emerging area.

Jesse Farebrother, Jordi Orbay, Quan Vuong et al. · deepmind

Value functions are a central component of deep reinforcement learning (RL). These functions, parameterized by neural networks, are trained using a mean squared error regression objective to match bootstrapped target values. However, scaling value-based RL methods that use regression to large networks, such as high-capacity Transformers, has proven challenging. This difficulty is in stark contrast to supervised learning: by leveraging a cross-entropy classification loss, supervised methods have scaled reliably to massive networks. Observing this discrepancy, in this paper, we investigate whether the scalability of deep RL can also be improved simply by using classification in place of regression for training value functions. We demonstrate that value functions trained with categorical cross-entropy significantly improves performance and scalability in a variety of domains. These include: single-task RL on Atari 2600 games with SoftMoEs, multi-task RL on Atari with large-scale ResNets, robotic manipulation with Q-transformers, playing Chess without search, and a language-agent Wordle task with high-capacity Transformers, achieving state-of-the-art results on these domains. Through careful analysis, we show that the benefits of categorical cross-entropy primarily stem from its ability to mitigate issues inherent to value-based RL, such as noisy targets and non-stationarity. Overall, we argue that a simple shift to training value functions with categorical cross-entropy can yield substantial improvements in the scalability of deep RL at little-to-no cost.

Katie Kang, Eric Wallace, Claire Tomlin et al.

Large language models are known to hallucinate when faced with unfamiliar queries, but the underlying mechanism that govern how models hallucinate are not yet fully understood. In this work, we find that unfamiliar examples in the models' finetuning data -- those that introduce concepts beyond the base model's scope of knowledge -- are crucial in shaping these errors. In particular, we find that an LLM's hallucinated predictions tend to mirror the responses associated with its unfamiliar finetuning examples. This suggests that by modifying how unfamiliar finetuning examples are supervised, we can influence a model's responses to unfamiliar queries (e.g., say ``I don't know''). We empirically validate this observation in a series of controlled experiments involving SFT, RL, and reward model finetuning on TriviaQA and MMLU. Our work further investigates RL finetuning strategies for improving the factuality of long-form model generations. We find that, while hallucinations from the reward model can significantly undermine the effectiveness of RL factuality finetuning, strategically controlling how reward models hallucinate can minimize these negative effects. Leveraging our previous observations on controlling hallucinations, we propose an approach for learning more reliable reward models, and show that they improve the efficacy of RL factuality finetuning in long-form biography and book/movie plot generation tasks.

Amrith Setlur, Nived Rajaraman, Sergey Levine et al. · cmu

Despite substantial advances in scaling test-time compute, an ongoing debate in the community is how it should be scaled up to enable continued and efficient improvements with scaling. There are largely two approaches: first, distilling successful search or thinking traces; and second, using verification (e.g., 0/1 outcome rewards, reward models, or verifiers) to guide reinforcement learning (RL) and search algorithms. In this paper, we prove that finetuning LLMs with verifier-based (VB) methods based on RL or search is far superior to verifier-free (VF) approaches based on distilling or cloning search traces, given a fixed amount of compute/data budget. Further, we show that as we scale test-time compute (measured as the output token length) and training data, suboptimality of VF methods scales poorly compared to VB when the base pre-trained LLM presents a heterogeneous distribution over correct solution traces (e.g., different lengths, styles, etc.) and admits a non-sharp distribution over rewards on traces sampled from it. We formalize this condition using anti-concentration [Erdős, 1945]. This implies a stronger result that VB methods scale better asymptotically, with the performance gap between VB and VF methods widening as test-time budget grows. We corroborate our theory empirically on both didactic and math reasoning problems with 3/8/32B-sized pre-trained LLMs, where we find verification is crucial for scaling test-time compute.

Mitsuhiko Nakamoto, Oier Mees, Aviral Kumar et al.

Large, general-purpose robotic policies trained on diverse demonstration datasets have been shown to be remarkably effective both for controlling a variety of robots in a range of different scenes, and for acquiring broad repertoires of manipulation skills. However, the data that such policies are trained on is generally of mixed quality -- not only are human-collected demonstrations unlikely to perform the task perfectly, but the larger the dataset is, the harder it is to curate only the highest quality examples. It also remains unclear how optimal data from one embodiment is for training on another embodiment. In this paper, we present a general and broadly applicable approach that enhances the performance of such generalist robot policies at deployment time by re-ranking their actions according to a value function learned via offline RL. This approach, which we call Value-Guided Policy Steering (V-GPS), is compatible with a wide range of different generalist policies, without needing to fine-tune or even access the weights of the policy. We show that the same value function can improve the performance of five different state-of-the-art policies with different architectures, even though they were trained on distinct datasets, attaining consistent performance improvement on multiple robotic platforms across a total of 12 tasks. Code and videos can be found at: https://nakamotoo.github.io/V-GPS

Yinlam Chow, Guy Tennenholtz, Izzeddin Gur et al.

Recent studies have indicated that effectively utilizing inference-time compute is crucial for attaining better performance from large language models (LLMs). In this work, we propose a novel inference-aware fine-tuning paradigm, in which the model is fine-tuned in a manner that directly optimizes the performance of the inference-time strategy. We study this paradigm using the simple yet effective Best-of-N (BoN) inference strategy, in which a verifier selects the best out of a set of LLM-generated responses. We devise the first imitation learning and reinforcement learning~(RL) methods for BoN-aware fine-tuning, overcoming the challenging, non-differentiable argmax operator within BoN. We empirically demonstrate that our BoN-aware models implicitly learn a meta-strategy that interleaves best responses with more diverse responses that might be better suited to a test-time input -- a process reminiscent of the exploration-exploitation trade-off in RL. Our experiments demonstrate the effectiveness of BoN-aware fine-tuning in terms of improved performance and inference-time compute. In particular, we show that our methods improve the Bo32 performance of Gemma 2B on Hendrycks MATH from 26.8% to 30.8%, and pass@32 from 60.0% to 67.0%, as well as the pass@16 on HumanEval from 61.6% to 67.1%.

William Chen, Oier Mees, Aviral Kumar et al.

Humans can quickly learn new behaviors by leveraging background world knowledge. In contrast, agents trained with reinforcement learning (RL) typically learn behaviors from scratch. We thus propose a novel approach that uses the vast amounts of general and indexable world knowledge encoded in vision-language models (VLMs) pre-trained on Internet-scale data for embodied RL. We initialize policies with VLMs by using them as promptable representations: embeddings that encode semantic features of visual observations based on the VLM's internal knowledge and reasoning capabilities, as elicited through prompts that provide task context and auxiliary information. We evaluate our approach on visually-complex, long horizon RL tasks in Minecraft and robot navigation in Habitat. We find that our policies trained on embeddings from off-the-shelf, general-purpose VLMs outperform equivalent policies trained on generic, non-promptable image embeddings. We also find our approach outperforms instruction-following methods and performs comparably to domain-specific embeddings. Finally, we show that our approach can use chain-of-thought prompting to produce representations of common-sense semantic reasoning, improving policy performance in novel scenes by 1.5 times.

Max Sobol Mark, Tian Gao, Georgia Gabriela Sampaio et al. · stanford

Recent advances in learning decision-making policies can largely be attributed to training expressive policy models, largely via imitation learning. While imitation learning discards non-expert data, reinforcement learning (RL) can still learn from suboptimal data. However, instantiating RL training of a new policy class often presents a different challenge: most deep RL machinery is co-developed with assumptions on the policy class and backbone, resulting in poor performance when the policy class changes. For instance, SAC utilizes a low-variance reparameterization policy gradient for Gaussian policies, but this is unstable for diffusion policies and intractable for autoregressive categorical policies. To address this issue, we develop an offline RL and online fine-tuning approach called policy-agnostic RL (PA-RL) that can effectively train multiple policy classes, with varying architectures and sizes. We build off the basic idea that a universal supervised learning loss can replace the policy improvement step in RL, as long as it is applied on "optimized" actions. To obtain these optimized actions, we first sample multiple actions from a base policy, and run global optimization (i.e., re-ranking multiple action samples using the Q-function) and local optimization (i.e., running gradient steps on an action sample) to maximize the critic on these candidates. PA-RL enables fine-tuning diffusion and transformer policies with either autoregressive tokens or continuous action outputs, at different sizes, entirely via actor-critic RL. Moreover, PA-RL improves the performance and sample-efficiency by up to 2 times compared to existing offline RL and online fine-tuning methods. We show the first result that successfully fine-tunes OpenVLA, a 7B generalist robot policy, autonomously with Cal-QL, an online RL fine-tuning algorithm, improving from 40% to 70% in the real world in 40 minutes.

Zhiyuan Zhou, Andy Peng, Qiyang Li et al.

The modern paradigm in machine learning involves pre-training on diverse data, followed by task-specific fine-tuning. In reinforcement learning (RL), this translates to learning via offline RL on a diverse historical dataset, followed by rapid online RL fine-tuning using interaction data. Most RL fine-tuning methods require continued training on offline data for stability and performance. However, this is undesirable because training on diverse offline data is slow and expensive for large datasets, and in principle, also limit the performance improvement possible because of constraints or pessimism on offline data. In this paper, we show that retaining offline data is unnecessary as long as we use a properly-designed online RL approach for fine-tuning offline RL initializations. To build this approach, we start by analyzing the role of retaining offline data in online fine-tuning. We find that continued training on offline data is mostly useful for preventing a sudden divergence in the value function at the onset of fine-tuning, caused by a distribution mismatch between the offline data and online rollouts. This divergence typically results in unlearning and forgetting the benefits of offline pre-training. Our approach, Warm-start RL (WSRL), mitigates the catastrophic forgetting of pre-trained initializations using a very simple idea. WSRL employs a warmup phase that seeds the online RL run with a very small number of rollouts from the pre-trained policy to do fast online RL. The data collected during warmup helps ``recalibrate'' the offline Q-function to the online distribution, allowing us to completely discard offline data without destabilizing the online RL fine-tuning. We show that WSRL is able to fine-tune without retaining any offline data, and is able to learn faster and attains higher performance than existing algorithms irrespective of whether they retain offline data or not.

Amrith Setlur, Matthew Y. R. Yang, Charlie Snell et al. · cmu

Test-time scaling offers a promising path to improve LLM reasoning by utilizing more compute at inference time; however, the true promise of this paradigm lies in extrapolation (i.e., improvement in performance on hard problems as LLMs keep "thinking" for longer, beyond the maximum token budget they were trained on). Surprisingly, we find that most existing reasoning models do not extrapolate well. We show that one way to enable extrapolation is by training the LLM to perform in-context exploration: training the LLM to effectively spend its test time budget by chaining operations (such as generation, verification, refinement, etc.), or testing multiple hypotheses before it commits to an answer. To enable in-context exploration, we identify three key ingredients as part of our recipe e3: (1) chaining skills that the base LLM has asymmetric competence in, e.g., chaining verification (easy) with generation (hard), as a way to implement in-context search; (2) leveraging "negative" gradients from incorrect traces to amplify exploration during RL, resulting in longer search traces that chains additional asymmetries; and (3) coupling task difficulty with training token budget during training via a specifically-designed curriculum to structure in-context exploration. Our recipe e3 produces the best known 1.7B model according to AIME'25 and HMMT'25 scores, and extrapolates to 2x the training token budget. Our e3-1.7B model not only attains high pass@1 scores, but also improves pass@k over the base model.

Katie Kang, Amrith Setlur, Dibya Ghosh et al. · berkeley, cmu

Despite the remarkable capabilities of modern large language models (LLMs), the mechanisms behind their problem-solving abilities remain elusive. In this work, we aim to better understand how the learning dynamics of LLM finetuning shapes downstream generalization. Our analysis focuses on reasoning tasks, whose problem structure allows us to distinguish between memorization (the exact replication of reasoning steps from the training data) and performance (the correctness of the final solution). We find that a model's generalization behavior can be effectively characterized by a training metric we call pre-memorization train accuracy: the accuracy of model samples on training queries before they begin to copy the exact reasoning steps from the training set. On the dataset level, this metric is able to reliably predict test accuracy, achieving $R^2$ of around or exceeding 0.9 across various models (Llama3 8, Gemma2 9B), datasets (GSM8k, MATH), and training configurations. On a per-example level, this metric is also indicative of whether individual model predictions are robust to perturbations in the training query. By connecting a model's learning behavior to its generalization, pre-memorization train accuracy can guide targeted improvements to training strategies. We focus on data curation as an example, and show that prioritizing examples with low pre-memorization accuracy leads to 1.5-2x improvements in data efficiency compared to i.i.d. data scaling, and outperforms other standard data curation techniques.

Oleh Rybkin, Michal Nauman, Preston Fu et al.

Scaling data and compute is critical to the success of modern ML. However, scaling demands predictability: we want methods to not only perform well with more compute or data, but also have their performance be predictable from small-scale runs, without running the large-scale experiment. In this paper, we show that value-based off-policy RL methods are predictable despite community lore regarding their pathological behavior. First, we show that data and compute requirements to attain a given performance level lie on a Pareto frontier, controlled by the updates-to-data (UTD) ratio. By estimating this frontier, we can predict this data requirement when given more compute, and this compute requirement when given more data. Second, we determine the optimal allocation of a total resource budget across data and compute for a given performance and use it to determine hyperparameters that maximize performance for a given budget. Third, this scaling is enabled by first estimating predictable relationships between hyperparameters, which is used to manage effects of overfitting and plasticity loss unique to RL. We validate our approach using three algorithms: SAC, BRO, and PQL on DeepMind Control, OpenAI gym, and IsaacGym, when extrapolating to higher levels of data, compute, budget, or performance.

Taeyoun Kim, Fahim Tajwar, Aditi Raghunathan et al.

Reasoning methods that adaptively allocate test-time compute have advanced LLM performance on easy to verify domains such as math and code. In this work, we study how to utilize this approach to train models that exhibit a degree of robustness to safety vulnerabilities, and show that doing so can provide benefits. We build a recipe called $\textit{TARS}$ (Training Adaptive Reasoners for Safety), a reinforcement learning (RL) approach that trains models to reason about safety using chain-of-thought traces and a reward signal that balances safety with task completion. To build TARS, we identify three critical design choices: (1) a ``lightweight'' warmstart SFT stage, (2) a mix of harmful, harmless, and ambiguous prompts to prevent shortcut behaviors such as too many refusals, and (3) a reward function to prevent degeneration of reasoning capabilities during training. Models trained with TARS exhibit adaptive behaviors by spending more compute on ambiguous queries, leading to better safety-refusal trade-offs. They also internally learn to better distinguish between safe and unsafe prompts and attain greater robustness to both white-box (e.g., GCG) and black-box attacks (e.g., PAIR). Overall, our work provides an effective, open recipe for training LLMs against jailbreaks and harmful requests by reasoning per prompt.

Michal Nauman, Marek Cygan, Carmelo Sferrazza et al.

Recent advances in language modeling and vision stem from training large models on diverse, multi-task data. This paradigm has had limited impact in value-based reinforcement learning (RL), where improvements are often driven by small models trained in a single-task context. This is because in multi-task RL sparse rewards and gradient conflicts make optimization of temporal difference brittle. Practical workflows for generalist policies therefore avoid online training, instead cloning expert trajectories or distilling collections of single-task policies into one agent. In this work, we show that the use of high-capacity value models trained via cross-entropy and conditioned on learnable task embeddings addresses the problem of task interference in online RL, allowing for robust and scalable multi-task training. We test our approach on 7 multi-task benchmarks with over 280 unique tasks, spanning high degree-of-freedom humanoid control and discrete vision-based RL. We find that, despite its simplicity, the proposed approach leads to state-of-the-art single and multi-task performance, as well as sample-efficient transfer to new tasks.

Zheyuan Hu, Robyn Wu, Naveen Enock et al.

Modern paradigms for robot imitation train expressive policy architectures on large amounts of human demonstration data. Yet performance on contact-rich, deformable-object, and long-horizon tasks plateau far below perfect execution, even with thousands of expert demonstrations. This is due to the inefficiency of existing ``expert'' data collection procedures based on human teleoperation. To address this issue, we introduce RaC, a new phase of training on human-in-the-loop rollouts after imitation learning pre-training. In RaC, we fine-tune a robotic policy on human intervention trajectories that illustrate recovery and correction behaviors. Specifically, during a policy rollout, human operators intervene when failure appears imminent, first rewinding the robot back to a familiar, in-distribution state and then providing a corrective segment that completes the current sub-task. Training on this data composition expands the robotic skill repertoire to include retry and adaptation behaviors, which we show are crucial for boosting both efficiency and robustness on long-horizon tasks. Across three real-world bimanual control tasks: shirt hanging, airtight container lid sealing, takeout box packing, and a simulated assembly task, RaC outperforms the prior state-of-the-art using 10$\times$ less data collection time and samples. We also show that RaC enables test-time scaling: the performance of the trained RaC policy scales linearly in the number of recovery maneuvers it exhibits. Videos of the learned policy are available at https://rac-scaling-robot.github.io/.

Preston Fu, Oleh Rybkin, Zhiyuan Zhou et al.

As models grow larger and training them becomes expensive, it becomes increasingly important to scale training recipes not just to larger models and more data, but to do so in a compute-optimal manner that extracts maximal performance per unit of compute. While such scaling has been well studied for language modeling, reinforcement learning (RL) has received less attention in this regard. In this paper, we investigate compute scaling for online, value-based deep RL. These methods present two primary axes for compute allocation: model capacity and the update-to-data (UTD) ratio. Given a fixed compute budget, we ask: how should resources be partitioned across these axes to maximize sample efficiency? Our analysis reveals a nuanced interplay between model size, batch size, and UTD. In particular, we identify a phenomenon we call TD-overfitting: increasing the batch quickly harms Q-function accuracy for small models, but this effect is absent in large models, enabling effective use of large batch size at scale. We provide a mental model for understanding this phenomenon and build guidelines for choosing batch size and UTD to optimize compute usage. Our findings provide a grounded starting point for compute-optimal scaling in deep RL, mirroring studies in supervised learning but adapted to TD learning.

Yuxiao Qu, Anikait Singh, Yoonho Lee et al. · cmu, stanford

Reasoning requires going beyond pattern matching or memorization of solutions to identify and implement "algorithmic procedures" that can be used to deduce answers to hard problems. Doing so requires realizing the most relevant primitives, intermediate results, or shared procedures, and building upon them. While RL post-training on long chains of thought ultimately aims to uncover this kind of algorithmic behavior, most reasoning traces learned by large models fail to consistently capture or reuse procedures, instead drifting into verbose and degenerate exploration. To address more effective reasoning, we introduce reasoning abstractions: concise natural language descriptions of procedural and factual knowledge that guide the model toward learning successful reasoning. We train models to be capable of proposing multiple abstractions given a problem, followed by RL that incentivizes building a solution while using the information provided by these abstractions. This results in a two-player RL training paradigm, abbreviated as RLAD, that jointly trains an abstraction generator and a solution generator. This setup effectively enables structured exploration, decouples learning signals of abstraction proposal and solution generation, and improves generalization to harder problems. We also show that allocating more test-time compute to generating abstractions is more beneficial for performance than generating more solutions at large test budgets, illustrating the role of abstractions in guiding meaningful exploration.

Grace Liu, Yuxiao Qu, Jeff Schneider et al.

Many tasks require learned models to strategically gather relevant information over multiple rounds of interaction before actually acting on a task. Strategic information gathering requires models to know not only how to effectively acquire information, but also when to stop gathering information and make a decision, in order to avoid overthinking or getting derailed when acting. In this paper, we formalize this problem and introduce Counterfactuals and Reasoning for Termination (CaRT), an approach for teaching LLMs when to stop seeking information. To appropriately learn when to terminate, CaRT fine-tunes LLMs using counterfactual pairs of trajectories, one where termination is appropriate and a minimally modified version of the same trajectory where it is not. It trains the LLM to explain the rationale for the termination decision in either case via verbal reasoning, and imbues this capability into the base LLM via fine-tuning. We instantiate CaRT in two domains: interactive medical diagnosis and math problem solving. In both domains, we find that CaRT improves the efficiency of information gathering and task success rate compared to other fine-tuning methods.

Bhavya Agrawalla, Michal Nauman, Khush Agrawal et al.

A hallmark of modern large-scale machine learning techniques is the use of training objectives that provide dense supervision to intermediate computations, such as teacher forcing the next token in language models or denoising step-by-step in diffusion models. This enables models to learn complex functions in a generalizable manner. Motivated by this observation, we investigate the benefits of iterative computation for temporal difference (TD) methods in reinforcement learning (RL). Typically they represent value functions in a monolithic fashion, without iterative compute. We introduce floq (flow-matching Q-functions), an approach that parameterizes the Q-function using a velocity field and trains it using techniques from flow-matching, typically used in generative modeling. This velocity field underneath the flow is trained using a TD-learning objective, which bootstraps from values produced by a target velocity field, computed by running multiple steps of numerical integration. Crucially, floq allows for more fine-grained control and scaling of the Q-function capacity than monolithic architectures, by appropriately setting the number of integration steps. Across a suite of challenging offline RL benchmarks and online fine-tuning tasks, floq improves performance by nearly 1.8x. floq scales capacity far better than standard TD-learning architectures, highlighting the potential of iterative computation for value learning.