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% → 45.3% on algorithmic tasks, 73% → 93.4% on GSM8K, and 28% →44.6% on easy-to-hard generalization on MATH. Furthermore, we find that training GenRM with synthetic verification rationales is sufficient to pick out subtle errors on math problems. Finally, we demonstrate that generative verifiers scale favorably with model size and inference-time compute.

Test-time scaling in agentic tasks often relies on generating long reasoning traces ("think" more) before acting, but this does not allow agents to acquire new information from the environment or adapt behavior over time. In this work, we propose scaling test-time interaction, an untapped dimension for test-time scaling that increases the agent's interaction horizon to enable rich behaviors such as exploration, backtracking, and dynamic re-planning within a single rollout. To demonstrate the promise of this scaling dimension, we situate our study in the domain of web agents. We first show that even prompting-based interaction scaling can improve task success on web benchmarks non-trivially. Building on this, we introduce TTI, a curriculum-based online reinforcement learning (RL) approach that trains agents by adaptively adjusting their interaction lengths during rollout. Using a Gemma 3 12B model, TTI sets a new state-of-the-art among open-source agents trained on public data on WebVoyager and WebArena. Case studies further reveal that TTI enables agents to balance exploration and exploitation adaptively. Our results establish interaction scaling as a powerful, complementary axis to scaling per-action compute, offering new avenues for training robust and adaptive agents.

Learning world models can teach an agent how the world works in an unsupervised manner. Even though it can be viewed as a special case of sequence modeling, progress for scaling world models on robotic applications such as autonomous driving has been somewhat less rapid than scaling language models with Generative Pre-trained Transformers (GPT). We identify two reasons as major bottlenecks: dealing with complex and unstructured observation space, and having a scalable generative model. Consequently, we propose Copilot4D, a novel world modeling approach that first tokenizes sensor observations with VQVAE, then predicts the future via discrete diffusion. To efficiently decode and denoise tokens in parallel, we recast Masked Generative Image Transformer as discrete diffusion and enhance it with a few simple changes, resulting in notable improvement. When applied to learning world models on point cloud observations, Copilot4D reduces prior SOTA Chamfer distance by more than 65% for 1s prediction, and more than 50% for 3s prediction, across NuScenes, KITTI Odometry, and Argoverse2 datasets. Our results demonstrate that discrete diffusion on tokenized agent experience can unlock the power of GPT-like unsupervised learning for robotics.

In this paper, we study the problem of unsupervised object detection from 3D point clouds in self-driving scenes. We present a simple yet effective method that exploits (i) point clustering in near-range areas where the point clouds are dense, (ii) temporal consistency to filter out noisy unsupervised detections, (iii) translation equivariance of CNNs to extend the auto-labels to long range, and (iv) self-supervision for improving on its own. Our approach, OYSTER (Object Discovery via Spatio-Temporal Refinement), does not impose constraints on data collection (such as repeated traversals of the same location), is able to detect objects in a zero-shot manner without supervised fine-tuning (even in sparse, distant regions), and continues to self-improve given more rounds of iterative self-training. To better measure model performance in self-driving scenarios, we propose a new planning-centric perception metric based on distance-to-collision. We demonstrate that our unsupervised object detector significantly outperforms unsupervised baselines on PandaSet and Argoverse 2 Sensor dataset, showing promise that self-supervision combined with object priors can enable object discovery in the wild. For more information, visit the project website: https://waabi.ai/research/oyster.

Planning, the ability to analyze the structure of a problem in the large and decompose it into interrelated subproblems, is a hallmark of human intelligence. While deep reinforcement learning (RL) has shown great promise for solving relatively straightforward control tasks, it remains an open problem how to best incorporate planning into existing deep RL paradigms to handle increasingly complex environments. One prominent framework, Model-Based RL, learns a world model and plans using step-by-step virtual rollouts. This type of world model quickly diverges from reality when the planning horizon increases, thus struggling at long-horizon planning. How can we learn world models that endow agents with the ability to do temporally extended reasoning? In this work, we propose to learn graph-structured world models composed of sparse, multi-step transitions. We devise a novel algorithm to learn latent landmarks that are scattered (in terms of reachability) across the goal space as the nodes on the graph. In this same graph, the edges are the reachability estimates distilled from Q-functions. On a variety of high-dimensional continuous control tasks ranging from robotic manipulation to navigation, we demonstrate that our method, named L3P, significantly outperforms prior work, and is oftentimes the only method capable of leveraging both the robustness of model-free RL and generalization of graph-search algorithms. We believe our work is an important step towards scalable planning in reinforcement learning.

We reinterpret the problem of finding intrinsic rewards in reinforcement learning (RL) as a bilevel optimization problem. Using this interpretation, we can make use of recent advancements in the hyperparameter optimization literature, mainly from Self-Tuning Networks (STN), to learn intrinsic rewards. To facilitate our methods, we introduces a new general conditioning layer: Conditional Layer Normalization (CLN). We evaluate our method on several continuous control benchmarks in the Mujoco physics simulator. On all of these benchmarks, the intrinsic rewards learned on the fly lead to higher final rewards.