Pierre-Luc Bacon

Adam Gleave

Increasing model size has unlocked a dazzling array of capabilities in language models. At the same time, even frontier models remain vulner… (voir plus)able to jailbreaks and prompt injections, despite concerted efforts to make them robust. As both attackers and defenders gain access to more compute, and as models become larger, what will be the effect on robustness? We argue that to answer this question requires a scaling lens, which we adopt in an extensive study of language model robustness across several classification tasks, model families, and adversarial attacks. We find that in the absence of explicit safety training, larger models are not consistently more robust; however, scale improves sample efficiency in adversarial training, though it worsens compute efficiency. Further, we find that increasing attack compute smoothly improves attack success rate against both undefended and adversarially trained models. Finally, after exploring robustness transfer across attacks and threat models, we combine attack and defense scaling rates to study the offense-defense balance. We find that while attack scaling outpaces adversarial training across all models studied, larger adversarially trained models might give defense the advantage in the long run. These results underscore the utility of the scaling lens, and provide a paradigm for evaluating future attacks and defenses on frontier models. Code for this project is available at https://github.com/AlignmentResearch/scaling-llm-robustness-paper.

2025-10-06

Proceedings of the 42nd International Conference on Machine Learning (publié)

proceedings.mlr.press

The Three Regimes of Offline-to-Online Reinforcement Learning

Li Li

Tianwei Ni

Offline-to-online reinforcement learning (RL) has emerged as a practical paradigm that leverages offline datasets for pretraining and online… (voir plus) interactions for fine-tuning. However, its empirical behavior is highly inconsistent: design choices of online-fine tuning that work well in one setting can fail completely in another. We propose a stability--plasticity principle that can explain this inconsistency: we should preserve the knowledge of pretrained policy or offline dataset during online fine-tuning, whichever is better, while maintaining sufficient plasticity. This perspective identifies three regimes of online fine-tuning, each requiring distinct stability properties. We validate this framework through a large-scale empirical study, finding that the results strongly align with its predictions in 45 of 63 cases. This work provides a principled framework for guiding design choices in offline-to-online RL based on the relative performance of the offline dataset and the pretrained policy.

2025-10-01

ArXiv (prépublication)

The Three Regimes of Offline-to-Online Reinforcement Learning

Li Li

Tianwei Ni

Offline-to-online reinforcement learning (RL) has emerged as a practical paradigm that leverages offline datasets for pretraining and online… (voir plus) interactions for fine-tuning. However, its empirical behavior is highly inconsistent: design choices of online-fine tuning that work well in one setting can fail completely in another. We propose a stability--plasticity principle that can explain this inconsistency: we should preserve the knowledge of pretrained policy or offline dataset during online fine-tuning, whichever is better, while maintaining sufficient plasticity. This perspective identifies three regimes of online fine-tuning, each requiring distinct stability properties. We validate this framework through a large-scale empirical study, finding that the results strongly align with its predictions in 45 of 63 cases. This work provides a principled framework for guiding design choices in offline-to-online RL based on the relative performance of the offline dataset and the pretrained policy.

2025-10-01

ArXiv (prépublication)

The Three Regimes of Offline-to-Online Reinforcement Learning

Li Li

Tianwei Ni

Offline-to-online reinforcement learning (RL) has emerged as a practical paradigm that leverages offline datasets for pretraining and online… (voir plus) interactions for fine-tuning. However, its empirical behavior is highly inconsistent: design choices of online-fine tuning that work well in one setting can fail completely in another. We propose a stability--plasticity principle that can explain this inconsistency: we should preserve the knowledge of pretrained policy or offline dataset during online fine-tuning, whichever is better, while maintaining sufficient plasticity. This perspective identifies three regimes of online fine-tuning, each requiring distinct stability properties. We validate this framework through a large-scale empirical study, finding that the results strongly align with its predictions in 45 of 63 cases. This work provides a principled framework for guiding design choices in offline-to-online RL based on the relative performance of the offline dataset and the pretrained policy.

2025-10-01

ArXiv (prépublication)

Planning with Unified Multimodal Models

Zhilong Zhang

Yang Yu

With the powerful reasoning capabilities of large language models (LLMs) and vision-language models (VLMs), many recent works have explored … (voir plus)using them for decision-making. However, most of these approaches rely solely on language-based reasoning, which limits their ability to reason and make informed decisions. Recently, a promising new direction has emerged with unified multimodal models (UMMs), which support both multimodal inputs and outputs. We believe such models have greater potential for decision-making by enabling reasoning through generated visual content. To this end, we propose Uni-Plan, a planning framework built on UMMs. Within this framework, a single model simultaneously serves as the policy, dynamics model, and value function. In addition, to avoid hallucinations in dynamics predictions, we present a novel approach self-discriminated filtering, where the generative model serves as a self-discriminator to filter out invalid dynamics predictions. Experiments on long-horizon planning tasks show that Uni-Plan substantially improves success rates compared to VLM-based methods, while also showing strong data scalability, requiring no expert demonstrations and achieving better performance under the same training-data size. This work lays a foundation for future research in reasoning and decision-making with UMMs.

2025-09-27

ArXiv (prépublication)

doi.org

Planning with Unified Multimodal Models

Zhilong Zhang

Yang Yu

With the powerful reasoning capabilities of large language models (LLMs) and vision-language models (VLMs), many recent works have explored … (voir plus)using them for decision-making. However, most of these approaches rely solely on language-based reasoning, which limits their ability to reason and make informed decisions. Recently, a promising new direction has emerged with unified multimodal models (UMMs), which support both multimodal inputs and outputs. We believe such models have greater potential for decision-making by enabling reasoning through generated visual content. To this end, we propose Uni-Plan, a planning framework built on UMMs. Within this framework, a single model simultaneously serves as the policy, dynamics model, and value function. In addition, to avoid hallucinations in dynamics predictions, we present a novel approach self-discriminated filtering, where the generative model serves as a self-discriminator to filter out invalid dynamics predictions. Experiments on long-horizon planning tasks show that Uni-Plan substantially improves success rates compared to VLM-based methods, while also showing strong data scalability, requiring no expert demonstrations and achieving better performance under the same training-data size. This work lays a foundation for future research in reasoning and decision-making with UMMs.

2025-09-27

ArXiv (prépublication)

Discovery of Sustainable Refrigerants through Physics-Informed RL Fine-Tuning of Sequence Models

Most refrigerants currently used in air-conditioning systems, such as hydrofluorocarbons, are potent greenhouse gases and are being phased d… (voir plus)own. Large-scale molecular screening has been applied to the search for alternatives, but in practice only about 300 refrigerants are known, and only a few additional candidates have been suggested without experimental validation. This scarcity of reliable data limits the effectiveness of purely data-driven methods. We present Refgen, a generative pipeline that integrates machine learning with physics-grounded inductive biases. Alongside fine-tuning for valid molecular generation, Refgen incorporates predictive models for critical properties, equations of state, thermochemical polynomials, and full vapor compression cycle simulations. These models enable reinforcement learning fine-tuning under thermodynamic constraints, enforcing consistency and guiding discovery toward molecules that balance efficiency, safety, and environmental impact. By embedding physics into the learning process, Refgen leverages scarce data effectively and enables de novo refrigerant discovery beyond the known set of compounds.

2025-09-23

ArXiv (prépublication)

Discovery of Sustainable Refrigerants through Physics-Informed RL Fine-Tuning of Sequence Models

Most refrigerants currently used in air-conditioning systems, such as hydrofluorocarbons, are potent greenhouse gases and are being phased d… (voir plus)own. Large-scale molecular screening has been applied to the search for alternatives, but in practice only about 300 refrigerants are known, and only a few additional candidates have been suggested without experimental validation. This scarcity of reliable data limits the effectiveness of purely data-driven methods. We present Refgen, a generative pipeline that integrates machine learning with physics-grounded inductive biases. Alongside fine-tuning for valid molecular generation, Refgen incorporates predictive models for critical properties, equations of state, thermochemical polynomials, and full vapor compression cycle simulations. These models enable reinforcement learning fine-tuning under thermodynamic constraints, enforcing consistency and guiding discovery toward molecules that balance efficiency, safety, and environmental impact. By embedding physics into the learning process, Refgen leverages scarce data effectively and enables de novo refrigerant discovery beyond the known set of compounds.

2025-09-23

ArXiv (prépublication)

Network Sparsity Unlocks the Scaling Potential of Deep Reinforcement Learning

Guozheng Ma

Li Li

Zilin Wang

Li Shen

Dacheng Tao

Effectively scaling up deep reinforcement learning models has proven notoriously difficult due to network pathologies during training, motiv… (voir plus)ating 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.

2025-06-21

ArXiv (prépublication)

doi.org

Network Sparsity Unlocks the Scaling Potential of Deep Reinforcement Learning

Guozheng Ma

Li Li

Zilin Wang

Li Shen

Dacheng Tao

Effectively scaling up deep reinforcement learning models has proven notoriously difficult due to network pathologies during training, motiv… (voir plus)ating 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.

2025-06-21

ArXiv (prépublication)

doi.org

Discrete Compositional Generation via General Soft Operators and Robust Reinforcement Learning

Bilun Sun

A major bottleneck in scientific discovery consists of narrowing an exponentially large set of objects, such as proteins or molecules, to a … (voir plus)small set of promising candidates with desirable properties. While this process can rely on expert knowledge, recent methods leverage reinforcement learning (RL) guided by a proxy reward function to enable this filtering. By employing various forms of entropy regularization, these methods aim to learn samplers that generate diverse candidates that are highly rated by the proxy function. In this work, we make two main contributions. First, we show that these methods are liable to generate overly diverse, suboptimal candidates in large search spaces. To address this issue, we introduce a novel unified operator that combines several regularized RL operators into a general framework that better targets peakier sampling distributions. Secondly, we offer a novel, robust RL perspective of this filtering process. The regularization can be interpreted as robustness to a compositional form of uncertainty in the proxy function (i.e., the true evaluation of a candidate differs from the proxy's evaluation). Our analysis leads us to a novel, easy-to-use algorithm we name trajectory general mellowmax (TGM): we show it identifies higher quality, diverse candidates than baselines in both synthetic and real-world tasks. Code: https://github.com/marcojira/tgm.

2025-06-20

ArXiv (prépublication)