Pierre-Luc Bacon

Gauthier Gidel

A major bottleneck in scientific discovery involves narrowing a large combinatorial set of objects, such as proteins or molecules, to a smal… (voir plus)l set of promising candidates. While this process largely relies on expert knowledge, recent methods leverage reinforcement learning (RL) to enhance this filtering. They achieve this by estimating proxy reward functions from available datasets and using regularization to generate more diverse candidates. These reward functions are inherently uncertain, raising a particularly salient challenge for scientific discovery. In this work, we show that existing methods, often framed as sampling proportional to a reward function, are inadequate and yield suboptimal candidates, especially in large search spaces. To remedy this issue, we take a robust RL approach and introduce a unified operator that seeks robustness to the uncertainty of the proxy reward function. This general operator targets peakier sampling distributions while encompassing known soft RL operators. It also leads us to a novel algorithm that identifies higher-quality, diverse candidates in both synthetic and real-world tasks. Ultimately, our work offers a new, flexible perspective on discrete compositional generation tasks. Code: https://github.com/marcojira/tgm.

2025-06-20

ArXiv (prépublication)

Stable Gradients for Stable Learning at Scale in Deep Reinforcement Learning

Roger Creus Castanyer

Johan Samir Obando Ceron

Lu Li

Scaling deep reinforcement learning networks is challenging and often results in degraded performance, yet the root causes of this failure m… (voir plus)ode remain poorly understood. Several recent works have proposed mechanisms to address this, but they are often complex and fail to highlight the causes underlying this difficulty. In this work, we conduct a series of empirical analyses which suggest that the combination of non-stationarity with gradient pathologies, due to suboptimal architectural choices, underlie the challenges of scale. We propose a series of direct interventions that stabilize gradient flow, enabling robust performance across a range of network depths and widths. Our interventions are simple to implement and compatible with well-established algorithms, and result in an effective mechanism that enables strong performance even at large scales. We validate our findings on a variety of agents and suites of environments.

2025-06-18

ArXiv (prépublication)

State Entropy Regularization for Robust Reinforcement Learning

Uri Koren

Yonatan Ashlag

Mirco Mutti

Esther Derman

Shie Mannor

State entropy regularization has empirically shown better exploration and sample complexity in reinforcement learning (RL). However, its the… (voir plus)oretical guarantees have not been studied. In this paper, we show that state entropy regularization improves robustness to structured and spatially correlated perturbations. These types of variation are common in transfer learning but often overlooked by standard robust RL methods, which typically focus on small, uncorrelated changes. We provide a comprehensive characterization of these robustness properties, including formal guarantees under reward and transition uncertainty, as well as settings where the method performs poorly. Much of our analysis contrasts state entropy with the widely used policy entropy regularization, highlighting their different benefits. Finally, from a practical standpoint, we illustrate that compared with policy entropy, the robustness advantages of state entropy are more sensitive to the number of rollouts used for policy evaluation.

2025-06-08

ArXiv (prépublication)

State Entropy Regularization for Robust Reinforcement Learning

Uri Koren

Yonatan Ashlag

Mirco Mutti

Esther Derman

Shie Mannor

State entropy regularization has empirically shown better exploration and sample complexity in reinforcement learning (RL). However, its the… (voir plus)oretical guarantees have not been studied. In this paper, we show that state entropy regularization improves robustness to structured and spatially correlated perturbations. These types of variation are common in transfer learning but often overlooked by standard robust RL methods, which typically focus on small, uncorrelated changes. We provide a comprehensive characterization of these robustness properties, including formal guarantees under reward and transition uncertainty, as well as settings where the method performs poorly. Much of our analysis contrasts state entropy with the widely used policy entropy regularization, highlighting their different benefits. Finally, from a practical standpoint, we illustrate that compared with policy entropy, the robustness advantages of state entropy are more sensitive to the number of rollouts used for policy evaluation.

2025-06-08

ArXiv (prépublication)

Understanding Behavioral Metric Learning: A Large-Scale Study on Distracting Reinforcement Learning Environments

Ziyan Luo

Tianwei Ni

Doina Precup

Xujie Si

2025-05-31

ArXiv (prépublication)

Understanding the Effectiveness of Learning Behavioral Metrics in Deep Reinforcement Learning

Ziyan Luo

Tianwei Ni

Doina Precup

Xujie Si

A key approach to state abstraction is approximating behavioral metrics (notably, bisimulation metrics) in the observation space, and embed … (voir plus)these learned distances in the representation space. While promising for robustness to task-irrelevant noise shown in prior work, accurately estimating these metrics remains challenging, requiring various design choices that create gaps between theory and practice. Prior evaluations focus mainly on final returns, leaving the quality of learned metrics and the source of performance gains unclear. To systematically assess how metric learning works in deep RL, we evaluate five recent approaches. We unify them under isometric embedding, identify key design choices, and benchmark them with baselines across 20 state-based and 14 pixel-based tasks, spanning 250+ configurations with diverse noise settings. Beyond final returns, we introduce the denoising factor to quantify the encoder’s ability to filter distractions. To further isolate the effect of metric learning, we propose an isolated metric estimation setting, where the encoder is influenced solely by the metric loss. Our results show that metric learning improves return and denoising only marginally, as its benefits fade when key design choices, such as layer normalization and self-prediction loss, are incorporated into the baseline. We also find that commonly used benchmarks (e.g., grayscale videos, varying state-based Gaussian noise dimensions) add little difficulty, while Gaussian noise with random projection and pixel-based Gaussian noise remain challenging even for the best methods. Finally, we release an open-source, modular codebase to improve reproducibility and support future research on metric learning in deep RL.

2025-05-09

rl-conference.cc/RLC/2025/Conference (accepté)

Understanding the Effectiveness of Learning Behavioral Metrics in Deep Reinforcement Learning

Ziyan Luo

Tianwei Ni

Doina Precup

Xujie Si

A key approach to state abstraction is approximating behavioral metrics (notably, bisimulation metrics) in the observation space, and embed … (voir plus)these learned distances in the representation space. While promising for robustness to task-irrelevant noise shown in prior work, accurately estimating these metrics remains challenging, requiring various design choices that create gaps between theory and practice. Prior evaluations focus mainly on final returns, leaving the quality of learned metrics and the source of performance gains unclear. To systematically assess how metric learning works in deep RL, we evaluate five recent approaches. We unify them under isometric embedding, identify key design choices, and benchmark them with baselines across 20 state-based and 14 pixel-based tasks, spanning 250+ configurations with diverse noise settings. Beyond final returns, we introduce the denoising factor to quantify the encoder’s ability to filter distractions. To further isolate the effect of metric learning, we propose an isolated metric estimation setting, where the encoder is influenced solely by the metric loss. Our results show that metric learning improves return and denoising only marginally, as its benefits fade when key design choices, such as layer normalization and self-prediction loss, are incorporated into the baseline. We also find that commonly used benchmarks (e.g., grayscale videos, varying state-based Gaussian noise dimensions) add little difficulty, while Gaussian noise with random projection and pixel-based Gaussian noise remain challenging even for the best methods. Finally, we release an open-source, modular codebase to improve reproducibility and support future research on metric learning in deep RL.

2025-05-09

rl-conference.cc/RLC/2025/Conference (publié)

Network Sparsity Unlocks the Scaling Potential of Deep Reinforcement Learning

Guozheng Ma

Lu Li

Zilin Wang

Li Shen

Dacheng Tao

Effectively scaling up deep reinforcement learning models has proven notoriously difficult due to network pathologies during training, moti… (voir plus)vating various targeted interventions such as periodic reset and architectural advances such as layer normalization. Instead of pursuing more complex modifications, we show that introducing static network sparsity alone can unlock further scaling potential beyond their dense counterparts with state-of-the-art architectures. This is achieved through simple one-shot random pruning, where a predetermined percentage of network weights are randomly removed once before training. Our analysis reveals that, in contrast to naively scaling up dense DRL networks, such sparse networks achieve both higher parameter efficiency for network expressivity and stronger resistance to optimization challenges like plasticity loss and gradient interference. We further extend our evaluation to visual and streaming RL scenarios, demonstrating the consistent benefits of network sparsity.

2025-05-01

ICML.cc/2025/Conference (présentation orale)

Scaling Trends in Language Model Robustness

Nikolaus H. R. Howe

Ian R. McKenzie

Oskar John Hollinsworth

Michał Zając

Tom Tseng

Aaron David Tucker

Adam Gleave

2025-05-01

ICML.cc/2025/Conference (poster)

Mol-MoE: Training Preference-Guided Routers for Molecule Generation

Diego Calanzone

Pierluca D'Oro

Recent advances in language models have enabled framing molecule generation as sequence modeling. However, existing approaches often rely on… (voir plus) single-objective reinforcement learning, limiting their applicability to real-world drug design, where multiple competing properties must be optimized. Traditional multi-objective reinforcement learning (MORL) methods require costly retraining for each new objective combination, making rapid exploration of trade-offs impractical. To overcome these limitations, we introduce Mol-MoE, a mixture-of-experts (MoE) architecture that enables efficient test-time steering of molecule generation without retraining. Central to our approach is a preference-based router training objective that incentivizes the router to combine experts in a way that aligns with user-specified trade-offs. This provides improved flexibility in exploring the chemical property space at test time, facilitating rapid trade-off exploration. Benchmarking against state-of-the-art methods, we show that Mol-MoE achieves superior sample quality and steerability.

2025-03-05

ICLR.cc/2025/Workshop/GEM (publié)

doi.org

Maxwell's Demon at Work: Efficient Pruning by Leveraging Saturation of Neurons

Simon Dufort-Labbé

Pierluca D'Oro

Evgenii Nikishin

Razvan Pascanu

Irina Rish

Aristide Baratin

2025-02-13

TMLR (accepté)

doi.org

Mol-MoE: Training Preference-Guided Routers for Molecule Generation

Diego Calanzone

Pierluca D'Oro

Recent advances in language models have enabled framing molecule generation as sequence modeling. However, existing approaches often rely on… (voir plus) single-objective reinforcement learning, limiting their applicability to real-world drug design, where multiple competing properties must be optimized. Traditional multi-objective reinforcement learning (MORL) methods require costly retraining for each new objective combination, making rapid exploration of trade-offs impractical. To overcome these limitations, we introduce Mol-MoE, a mixture-of-experts (MoE) architecture that enables efficient test-time steering of molecule generation without retraining. Central to our approach is a preference-based router training objective that incentivizes the router to combine experts in a way that aligns with user-specified trade-offs. This provides improved flexibility in exploring the chemical property space at test time, facilitating rapid trade-off exploration. Benchmarking against state-of-the-art methods, we show that Mol-MoE achieves superior sample quality and steerability.

2025-02-08

ArXiv (prépublication)