Abstract
We introduce PPOPT - Proximal Policy Optimization using Pretraining, a novel, model-free deep-reinforcement-learning algorithm that leverages pretraining to achieve high training efficiency and stability on very small training samples in physics-based environments. Reinforcement learning agents typically rely on large samples of environment interactions to learn a policy. However, frequent interactions with a (computer-simulated) environment may incur high computational costs, especially when the environment is complex. Our main innovation is a new policy neural network architecture that consists of a pretrained neural network middle section sandwiched between two fully-connected networks. Pretraining part of the network on a different environment with similar physics will help the agent learn the target environment with high efficiency because it will leverage a general understanding of the transferrable physics characteristics from the pretraining environment. We demonstrate that PPOPT outperforms baseline classic PPO on small training samples both in terms of rewards gained and general training stability. While PPOPT underperforms against classic model-based methods such as DYNA DDPG, the model-free nature of PPOPT allows it to train in significantly less time than its model-based counterparts. Finally, we present our implementation of PPOPT as open-source software, available at github.com/Davidrxyang/PPOPT.

Abstract
This paper explores the application of deep Q-learning to hedging at-the-money options on the S\&P~500 index. We develop an agent based on the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm, trained to simulate hedging decisions without making explicit model assumptions on price dynamics. The agent was trained on historical intraday prices of S\&P~500 call options across years 2004--2024, using a single time series of six predictor variables: option price, underlying asset price, moneyness, time to maturity, realized volatility, and current hedge position. A walk-forward procedure was applied for training, which led to nearly 17~years of out-of-sample evaluation. The performance of the deep reinforcement learning (DRL) agent is benchmarked against the Black--Scholes delta-hedging strategy over the same period. We assess both approaches using metrics such as annualized return, volatility, information ratio, and Sharpe ratio. To test the models' adaptability, we performed simulations across varying market conditions and added constraints such as transaction costs and risk-awareness penalties. Our results show that the DRL agent can outperform traditional hedging methods, particularly in volatile or high-cost environments, highlighting its robustness and flexibility in practical trading contexts. While the agent consistently outperforms delta-hedging, its performance deteriorates when the risk-awareness parameter is higher. We also observed that the longer the time interval used for volatility estimation, the more stable the results.

Abstract
Data lakehouses run sensitive workloads, where AI-driven automation raises concerns about trust, correctness, and governance. We argue that API-first, programmable lakehouses provide the right abstractions for safe-by-design, agentic workflows. Using Bauplan as a case study, we show how data branching and declarative environments extend naturally to agents, enabling reproducibility and observability while reducing the attack surface. We present a proof-of-concept in which agents repair data pipelines using correctness checks inspired by proof-carrying code. Our prototype demonstrates that untrusted AI agents can operate safely on production data and outlines a path toward a fully agentic lakehouse.

Abstract
A long-term goal of language agents is to learn and improve through their own experience, ultimately outperforming humans in complex, real-world tasks. However, training agents from experience data with reinforcement learning remains difficult in many environments, which either lack verifiable rewards (e.g., websites) or require inefficient long-horizon rollouts (e.g., multi-turn tool use). As a result, most current agents rely on supervised fine-tuning on expert data, which is challenging to scale and generalizes poorly. This limitation stems from the nature of expert demonstrations: they capture only a narrow range of scenarios and expose the agent to limited environment diversity. We address this limitation with a middle-ground paradigm we call early experience: interaction data generated by the agent's own actions, where the resulting future states serve as supervision without reward signals. Within this paradigm we study two strategies of using such data: (1) Implicit world modeling, which uses collected states to ground the policy in environment dynamics; and (2) Self-reflection, where the agent learns from its suboptimal actions to improve reasoning and decision-making. We evaluate across eight diverse environments and multiple model families. Our approaches consistently improve effectiveness and out-of-domain generalization, highlighting the value of early experience. Moreover, in environments with verifiable rewards, our results provide promising signals that early experience offers a strong foundation for subsequent reinforcement learning, positioning it as a practical bridge between imitation learning and fully experience-driven agents.

AI Insights

• WebArena and its lightweight subset WebArena‑Lite benchmark realistic web‑interaction tasks.
• Llama‑3.2‑3B‑Instruct, Qwen‑2.5‑7B‑Instruct, and GPT‑4‑Turbo all benefit from early‑experience methods.
• Implicit world‑modeling trains a latent dynamics predictor from self‑generated states, enabling reward‑free anticipation.
• Self‑reflection replays suboptimal actions, letting the agent learn corrective policies from its own traces.
• Experiment code is hosted at https://github.com/your-repo-name/webarena-benchmark for quick replication.
• Key papers include “WebArena: A Benchmark for Evaluating Web Interaction Capabilities” and “Implicit World Modeling and Self‑Reflection for Improving AI Performance on Web Tasks.”
• The study shows early experience bridges imitation learning and full‑experience RL, boosting out‑of‑domain generalization.

Abstract
Flow Matching (FM) has shown remarkable ability in modeling complex distributions and achieves strong performance in offline imitation learning for cloning expert behaviors. However, despite its behavioral cloning expressiveness, FM-based policies are inherently limited by their lack of environmental interaction and exploration. This leads to poor generalization in unseen scenarios beyond the expert demonstrations, underscoring the necessity of online interaction with environment. Unfortunately, optimizing FM policies via online interaction is challenging and inefficient due to instability in gradient computation and high inference costs. To address these issues, we propose to let a student policy with simple MLP structure explore the environment and be online updated via RL algorithm with a reward model. This reward model is associated with a teacher FM model, containing rich information of expert data distribution. Furthermore, the same teacher FM model is utilized to regularize the student policy's behavior to stabilize policy learning. Due to the student's simple architecture, we avoid the gradient instability of FM policies and enable efficient online exploration, while still leveraging the expressiveness of the teacher FM model. Extensive experiments show that our approach significantly enhances learning efficiency, generalization, and robustness, especially when learning from suboptimal expert data.

Abstract
Reinforcement learning (RL) is a powerful paradigm for learning to make sequences of decisions. However, RL has yet to be fully leveraged in robotics, principally due to its lack of scalability. Offline RL offers a promising avenue by training agents on large, diverse datasets, avoiding the costly real-world interactions of online RL. Scaling offline RL to increasingly complex datasets requires expressive generative models such as diffusion and flow matching. However, existing methods typically depend on either backpropagation through time (BPTT), which is computationally prohibitive, or policy distillation, which introduces compounding errors and limits scalability to larger base policies. In this paper, we consider the question of how to develop a scalable offline RL approach without relying on distillation or backpropagation through time. We introduce Expressive Value Learning for Offline Reinforcement Learning (EVOR): a scalable offline RL approach that integrates both expressive policies and expressive value functions. EVOR learns an optimal, regularized Q-function via flow matching during training. At inference-time, EVOR performs inference-time policy extraction via rejection sampling against the expressive value function, enabling efficient optimization, regularization, and compute-scalable search without retraining. Empirically, we show that EVOR outperforms baselines on a diverse set of offline RL tasks, demonstrating the benefit of integrating expressive value learning into offline RL.

AI Insights

• EVOR’s policy and value networks are 4‑layer MLPs with 512 units, using layer‑norm on the value side.
• Actions are sampled via a forward‑Euler integrator before a weighted softmax from the Q approximation.
• Rejection sampling at inference replaces back‑propagation, enabling fast policy extraction.
• Hyperparameters β and β are pivotal; careful tuning is required for top OGBench success rates.
• EVOR beats QC‑5 and QC‑1 on diverse offline RL tasks, proving flow‑matching value learning’s edge.
• The code reuses Li et al. (2025)’s QC implementation, itself adapted from Park et al. (2024a), showing a growing RL backbone ecosystem.
• Flow matching may not suit all environments, and hyperparameter sensitivity remains a practical hurdle.