Scaling up self-supervised learning has driven breakthroughs in language and vision, yet comparable progress has remained elusive in reinforcement learning (RL). In this paper, we study building blocks for self-supervised RL that unlock substantial improvements in scalability, with network depth serving as a critical factor. Whereas most RL papers in recent years have relied on shallow architectures (around 2 - 5 layers), we demonstrate that increasing the depth up to 1024 layers can significantly boost performance. Our experiments are conducted in an unsupervised goal-conditioned setting, where no demonstrations or rewards are provided, so an agent must explore (from scratch) and learn how to maximize the likelihood of reaching commanded goals. Evaluated on simulated locomotion and manipulation tasks, our approach increases performance by $2\times$ - $50\times$. Increasing the model depth not only increases success rates but also qualitatively changes the behaviors learned.

Document-level relation extraction (DocRE) is the process of identifying and extracting relations between entities that span multiple sentences within a document. Due to its realistic settings, DocRE has garnered increasing research attention in recent years. Previous research has mostly focused on developing sophisticated encoding models to better capture the intricate patterns between entity pairs. While these advancements are undoubtedly crucial, an even more foundational challenge lies in the data itself. The complexity inherent in DocRE makes the labeling process prone to errors, compounded by the extreme sparsity of positive relation samples, which is driven by both the limited availability of positive instances and the broad diversity of positive relation types. These factors can lead to biased optimization processes, further complicating the task of accurate relation extraction. Recognizing these challenges, we have developed a robust framework called \textit{\textbf{COMM}} to better solve DocRE. \textit{\textbf{COMM}} operates by initially employing an instance-aware reasoning method to dynamically capture pertinent information of entity pairs within the document and extract relational features. Following this, \textit{\textbf{COMM}} takes into account the distribution of relations and the difficulty of samples to dynamically adjust the margins between prediction logits and the decision threshold, a process we call Concentrated Margin Maximization. In this way, \textit{\textbf{COMM}} not only enhances the extraction of relevant relational features but also boosts DocRE performance by addressing the specific challenges posed by the data. Extensive experiments and analysis demonstrate the versatility and effectiveness of \textit{\textbf{COMM}}, especially its robustness when trained on low-quality data (achieves \textgreater 10\% performance gains).

Semi-supervised learning is of great significance in medical image segmentation by exploiting unlabeled data. Among its strategies, the co-training framework is prominent. However, previous co-training studies predominantly concentrate on network initialization variances and pseudo-label generation, while overlooking the equilibrium between information interchange and model diversity preservation. In this paper, we propose the Masked Image Consistency and Discrepancy Learning (MICD) framework with three key modules. The Masked Cross Pseudo Consistency (MCPC) module enriches context perception and small sample learning via pseudo-labeling across masked-input branches. The Cross Feature Consistency (CFC) module fortifies information exchange and model robustness by ensuring decoder feature consistency. The Cross Model Discrepancy (CMD) module utilizes EMA teacher networks to oversee outputs and preserve branch diversity. Together, these modules address existing limitations by focusing on fine-grained local information and maintaining diversity in a heterogeneous framework. Experiments on two public medical image datasets, AMOS and Synapse, demonstrate that our approach outperforms state-of-the-art methods.

Compression is at the heart of intelligence. A theoretically optimal way to compress any sequence of data is to find the shortest program that outputs that sequence and then halts. However, such 'Kolmogorov compression' is uncomputable, and code generating LLMs struggle to approximate this theoretical ideal, as it requires reasoning, planning and search capabilities beyond those of current models. In this work, we introduce the KoLMogorov-Test (KT), a compression-as-intelligence test for code generating LLMs. In KT a model is presented with a sequence of data at inference time, and asked to generate the shortest program that produces the sequence. We identify several benefits of KT for both evaluation and training: an essentially infinite number of problem instances of varying difficulty is readily available, strong baselines already exist, the evaluation metric (compression) cannot be gamed, and pretraining data contamination is highly unlikely. To evaluate current models, we use audio, text, and DNA data, as well as sequences produced by random synthetic programs. Current flagship models perform poorly - both GPT4-o and Llama-3.1-405B struggle on our natural and synthetic sequences. On our synthetic distribution, we are able to train code generation models with lower compression rates than previous approaches. Moreover, we show that gains on synthetic data generalize poorly to real data, suggesting that new innovations are necessary for additional gains on KT.

Effective long-term memory management is crucial for language models handling extended contexts. We introduce a novel framework that dynamically ranks memory entries based on relevance. Unlike previous works, our model introduces a novel relevance scoring and a pointwise re-ranking model for key-value embeddings, inspired by learning-to-rank techniques in information retrieval. Enhanced Ranked Memory Augmented Retrieval ERMAR achieves state-of-the-art results on standard benchmarks.

Many-shot in-context learning has recently shown promise as an alternative to finetuning, with the major advantage that the same model can be served for multiple tasks. However, this shifts the computational burden from training-time to inference-time, making deployment of many-shot ICL challenging to justify in-practice. This cost is further increased if a custom demonstration set is retrieved for each inference example. We present Dynamic Block-Sparse Attention, a training-free framework for retrieval-based many-shot in-context learning. By combining carefully designed block-sparse attention and retrieval of cached groups of demonstrations, we achieve comparable per-example latency to finetuning while maintaining on average >95% of the best method's accuracy across strong ICL and finetuning baselines. We hope that this will further enable the deployment of many-shot ICL at scale.

The adversarial Bandit with Knapsack problem is a multi-armed bandits problem with budget constraints and adversarial rewards and costs. In each round, a learner selects an action to take and observes the reward and cost of the selected action. The goal is to maximize the sum of rewards while satisfying the budget constraint. The classical benchmark to compare against is the best fixed distribution over actions that satisfies the budget constraint in expectation. Unlike its stochastic counterpart, where rewards and costs are drawn from some fixed distribution (Badanidiyuru et al., 2018), the adversarial BwK problem does not admit a no-regret algorithm for every problem instance due to the "spend-or-save" dilemma (Immorlica et al., 2022). A key problem left open by existing works is whether there exists a weaker but still meaningful benchmark to compare against such that no-regret learning is still possible. In this work, we present a new benchmark to compare against, motivated both by real-world applications such as autobidding and by its underlying mathematical structure. The benchmark is based on the Earth Mover's Distance (EMD), and we show that sublinear regret is attainable against any strategy whose spending pattern is within EMD $o(T^2)$ of any sub-pacing spending pattern. As a special case, we obtain results against the "pacing over windows" benchmark, where we partition time into disjoint windows of size $w$ and allow the benchmark strategies to choose a different distribution over actions for each window while satisfying a pacing budget constraint. Against this benchmark, our algorithm obtains a regret bound of $\tilde{O}(T/\sqrt{w}+\sqrt{wT})$. We also show a matching lower bound, proving the optimality of our algorithm in this important special case. In addition, we provide further evidence of the necessity of the EMD condition for obtaining a sublinear regret.

Distributed denial-of-service (DDoS) attacks remain a critical threat to Internet services, causing costly disruptions. While machine learning (ML) has shown promise in DDoS detection, current solutions struggle with multi-domain environments where attacks must be detected across heterogeneous networks and organizational boundaries. This limitation severely impacts the practical deployment of ML-based defenses in real-world settings. This paper introduces Anomaly-Flow, a novel framework that addresses this critical gap by combining Federated Learning (FL) with Generative Adversarial Networks (GANs) for privacy-preserving, multi-domain DDoS detection. Our proposal enables collaborative learning across diverse network domains while preserving data privacy through synthetic flow generation. Through extensive evaluation across three distinct network datasets, Anomaly-Flow achieves an average F1-score of $0.747$, outperforming baseline models. Importantly, our framework enables organizations to share attack detection capabilities without exposing sensitive network data, making it particularly valuable for critical infrastructure and privacy-sensitive sectors. Beyond immediate technical contributions, this work provides insights into the challenges and opportunities in multi-domain DDoS detection, establishing a foundation for future research in collaborative network defense systems. Our findings have important implications for academic research and industry practitioners working to deploy practical ML-based security solutions.

Social dilemmas pose a significant challenge in the field of multi-agent reinforcement learning (MARL). Melting Pot is an extensive framework designed to evaluate social dilemma environments, providing an evaluation protocol that measures generalization to new social partners across various test scenarios. However, running reinforcement learning algorithms in the official Melting Pot environments demands substantial computational resources. In this paper, we introduce SocialJax, a suite of sequential social dilemma environments implemented in JAX. JAX is a high-performance numerical computing library for Python that enables significant improvements in the operational efficiency of SocialJax on GPUs and TPUs. Our experiments demonstrate that the training pipeline of SocialJax achieves a 50\texttimes{} speedup in real-time performance compared to Melting Pot's RLlib baselines. Additionally, we validate the effectiveness of baseline algorithms within the SocialJax environments. Finally, we use Schelling diagrams to verify the social dilemma properties of these environments, ensuring they accurately capture the dynamics of social dilemmas.

Graph Neural Networks have significantly advanced research in recommender systems over the past few years. These methods typically capture global interests using aggregated past interactions and rely on static embeddings of users and items over extended periods of time. While effective in some domains, these methods fall short in many real-world scenarios, especially in finance, where user interests and item popularity evolve rapidly over time. To address these challenges, we introduce a novel extension to Light Graph Convolutional Network (LightGCN) designed to learn temporal node embeddings that capture dynamic interests. Our approach employs causal convolution to maintain a forward-looking model architecture. By preserving the chronological order of user-item interactions and introducing a dynamic update mechanism for embeddings through a sliding window, the proposed model generates well-timed and contextually relevant recommendations. Extensive experiments on a real-world dataset from BNP Paribas demonstrate that our approach significantly enhances the performance of LightGCN while maintaining the simplicity and efficiency of its architecture. Our findings provide new insights into designing graph-based recommender systems in time-sensitive applications, particularly for financial product recommendations.