LLMs have made impressive progress, but their growing capabilities also expose them to highly flexible jailbreaking attacks designed to bypass safety alignment. While many existing defenses focus on known types of attacks, it is more critical to prepare LLMs for unseen attacks that may arise during deployment. To address this, we propose a lifelong safety alignment framework that enables LLMs to continuously adapt to new and evolving jailbreaking strategies. Our framework introduces a competitive setup between two components: a Meta-Attacker, trained to actively discover novel jailbreaking strategies, and a Defender, trained to resist them. To effectively warm up the Meta-Attacker, we first leverage the GPT-4o API to extract key insights from a large collection of jailbreak-related research papers. Through iterative training, the first iteration Meta-Attacker achieves a 73% attack success rate (ASR) on RR [80] and a 57% transfer ASR on LAT [53] using only single-turn attacks. Meanwhile, the Defender progressively improves its robustness and ultimately reduces the Meta-Attacker's success rate to just 7%, enabling safer and more reliable deployment of LLMs in open-ended environments.

Training agents to operate under strict constraints during deployment, such as limited resource budgets or stringent safety requirements, presents significant challenges, especially when these constraints render the task complex. In this work, we propose a curriculum learning strategy that gradually tightens constraints during training, enabling the agent to incrementally master the deployment requirements. Inspired by self-paced learning techniques in unconstrained reinforcement learning (RL), our approach facilitates a smoother transition to challenging environments by initially training on simplified versions of the constraints and progressively introducing the full deployment conditions. We provide a theoretical analysis using an RL agent in a binary-tree Markov Decision Process (MDP) to demonstrate that our curriculum strategy can accelerate training relative to a baseline approach that imposes the trajectory constraints from the outset.

Large-scale machine learning models deliver strong performance across a wide range of tasks but come with significant computational and resource constraints. To mitigate these challenges, local smaller models are often deployed alongside larger models, relying on routing and deferral mechanisms to offload complex tasks.

We introduce ReplaceMe, a generalized training-free depth pruning method that effectively replaces transformer blocks with a linear operation, while maintaining high performance for low compression ratios. In contrast to conventional pruning approaches that require additional training or fine-tuning, our approach requires only a small calibration dataset that is used to estimate a linear transformation, which approximates the pruned blocks. The estimated linear mapping can be seamlessly merged with the remaining transformer blocks, eliminating the need for any additional network parameters. Our experiments show that ReplaceMe consistently outperforms other training-free approaches and remains highly competitive with state-of-the-art pruning methods that involve extensive retraining/fine-tuning and architectural modifications. Applied to several large language models (LLMs), ReplaceMe achieves up to 25% pruning while retaining approximately 90% of the original model's performance on open benchmarks--without any training or healing steps, resulting in minimal computational overhead.

Large Language Models (LLMs) exhibit impressive performance across diverse domains but often suffer from overconfidence, limiting their reliability in critical applications. We propose SteerConf, a novel framework that systematically steers LLMs' confidence scores to improve their calibration and reliability. SteerConf introduces three key components: (1) a steering prompt strategy that guides LLMs to produce confidence scores in specified directions (e.g., conservative or optimistic) by leveraging prompts with varying steering levels; (2) a steered confidence consistency measure that quantifies alignment across multiple steered confidences to enhance calibration; and (3) a steered confidence calibration method that aggregates confidence scores using consistency measures and applies linear quantization for answer selection. SteerConf operates without additional training or fine-tuning, making it broadly applicable to existing LLMs. Experiments on seven benchmarks spanning professional knowledge, common sense, ethics, and reasoning tasks, using advanced LLM models (GPT-3.5, LLaMA 3, GPT-4), demonstrate that SteerConf significantly outperforms existing methods, often by a significant margin. Our findings highlight the potential of steering the confidence of LLMs to enhance their reliability for safer deployment in real-world applications.

The quadratic complexity of the attention mechanism in Transformer models has motivated the development of alternative architectures with sub-quadratic scaling, such as state-space models. Among these, Mamba has emerged as a leading architecture, achieving state-of-the-art results across a range of language modeling tasks. However, Mambas performance significantly deteriorates when applied to contexts longer than those seen during pre-training, revealing a sharp sensitivity to context length extension. Through detailed analysis, we attribute this limitation to the out-of-distribution behavior of its state-space dynamics, particularly within the parameterization of the state transition matrix A. Unlike recent works which attribute this sensitivity to the vanished accumulation of discretization time steps, exp( PN t=1 t), we establish a connection between state convergence behavior as the input length approaches infinity and the spectrum of the transition matrix A, offering a well-founded explanation of its role in length extension. Next, to overcome this challenge, we propose an approach that applies spectrum scaling to pre-trained Mamba models to enable robust long-context generalization by selectively modulating the spectrum of Amatrices in each layer. We show that this can significantly improve performance in settings where simply modulating t fails, validating our insights and providing avenues for better length generalization of state-space models with structured transition matrices.

When Did White-Collar Work Start to Look So Bleak? In the nineteen-eighties, an office job promised security and fulfillment. For graduates starting careers today, the prospect is often tinged with dread. The workplace's sense of control can prove illusory--as it did in the era of yuppie-wrought corporate consolidation, and as it does now for graduates entering an economy destabilized by new uncertainties. This spring, across the nation's auditoriums and quadrangles, members of the class of 2026 took their seats to receive remarks from distinguished guests. The graduation speech is a thankless form: generalized, impersonal exhortation/congratulation is almost guaranteed to be forgettable, if all goes well. But this year, on at least a few American campuses, all did not go well. At the University of Arizona, Eric Schmidt, the former C.E.O. of Google, told the crowd that artificial intelligence "will touch every profession, every classroom, every hospital, every laboratory, every person, and every relationship you have," a sweeping promise that landed like a threat.

We propose Fast Long-context Adaptive Memory for Event (FLAME), a novel scalable architecture that combines neuro-inspired feature extraction with robust structured sequence modeling to efficiently process asynchronous and sparse event camera data. As a departure from conventional input encoding methods, FLAME presents Event Attention Layer, a novel feature extractor that leverages neuromorphic dynamics (Leaky Integrate-and-Fire (LIF)) to directly capture multi-timescale features from event streams. The feature extractor integrates with a structured state-space model with a novel Event-Aware HiPPO (EA-HiPPO) mechanism that dynamically adapts memory retention based on inter-event intervals to understand relationship across varying temporal scales and event sequences. ANormal Plus Low Rank (NPLR) decomposition reduces the computational complexity of state update from O(N2) to O(Nr), where N represents the dimension of the core state vector and r is the rank of a low-rank component (with r N). FLAME demonstrates state-of-the-art accuracy for event-by-event processing on complex event camera datasets.

Test-time adaptation (TTA) aims to adapt a source model to a target domain using only test data. Existing methods predominantly rely on unsupervised entropy minimization or its variants, which suffer from degeneration, leading to trivial solutions with low-entropy but inaccurate predictions. In this work, we identify entropy-deceptive (ED) samples, instances where the model makes highly confident yet incorrect predictions, as the underlying cause of degeneration. Further, we reveal that the gradients of entropy minimization in TTA have an intrinsic lowdimensional structure, driven primarily by entropy-truthful (ET) samples whose gradients are highly correlated. In contrast, ED samples have scattered, less correlated gradients. Leveraging this observation, we show that the detrimental impact of ED samples can be suppressed by constraining model updates within the principal subspace of backward gradients. Building on this insight, we propose LCoTTA, a lifelong continual TTA method that tracks the principal subspace of gradients online and utilizes their projections onto this subspace for adaptation. Further, we provide theoretical analysis to show that the proposed subspace-based method can enhance the robustness against detrimental ED samples. Extensive experiments demonstrate that LCoTTA effectively overcomes degeneration and significantly outperforms existing methods in long-term continual adaptation scenarios.

Current document retrieval-augmented generation (DocRAG) Therefore, the number of female respondents who never listened to theradio is: Number of females who never listened = 2,001 * 0.557 = 1,115 methods remain limited by their text-centric approaches, frequently missing "text12": [ "The table provides a