Memory systems are fundamental to AI agents, yet existing work often lacks adaptability to diverse tasks and overlooks the constructive and task-oriented role of AI agent memory. Drawing from Piaget's theory of cognitive development, we propose PISA, a pragmatic, psych-inspired unified memory system that addresses these limitations by treating memory as a constructive and adaptive process. To enable continuous learning and adaptability, PISA introduces a trimodal adaptation mechanism (i.e., schema updation, schema evolution, and schema creation) that preserves coherent organization while supporting flexible memory updates. Building on these schema-grounded structures, we further design a hybrid memory access architecture that seamlessly integrates symbolic reasoning with neural retrieval, significantly improving retrieval accuracy and efficiency. Our empirical evaluation, conducted on the existing LOCOMO benchmark and our newly proposed AggQA benchmark for data analysis tasks, confirms that PISA sets a new state-of-the-art by significantly enhancing adaptability and long-term knowledge retention.

Modern large reasoning models (LRMs) exhibit impressive multi-step problem-solving via chain-of-thought (CoT) reasoning. However, this iterative thinking mechanism introduces a new vulnerability surface. We present the Deadlock Attack, a resource exhaustion method that hijacks an LRM's generative control flow by training a malicious adversarial embedding to induce perpetual reasoning loops. Specifically, the optimized embedding encourages transitional tokens (e.g., "Wait", "But") after reasoning steps, preventing the model from concluding its answer. A key challenge we identify is the continuous-to-discrete projection gap: naïve projections of adversarial embeddings to token sequences nullify the attack. To overcome this, we introduce a backdoor implantation strategy, enabling reliable activation through specific trigger tokens. Our method achieves a 100% attack success rate across four advanced LRMs (Phi-RM, Nemotron-Nano, R1-Qwen, R1-Llama) and three math reasoning benchmarks, forcing models to generate up to their maximum token limits. The attack is also stealthy (in terms of causing negligible utility loss on benign user inputs) and remains robust against existing strategies trying to mitigate the overthinking issue. Our findings expose a critical and underexplored security vulnerability in LRMs from the perspective of reasoning (in)efficiency.

The adaptation of pre-trained large language models (LLMs) to diverse downstream tasks via fine-tuning is critical for numerous applications. However, the inefficiency of parameter-efficient fine-tuning (PEFT) techniques presents significant challenges in terms of time investments and operational costs. In this paper, we first introduce a nuanced form of sparsity, termed Shadowy Sparsity, which is distinctive in fine-tuning and has not been adequately addressed for acceleration. Under Shadowy Sparsity, we propose Long Exposure, an efficient system to accelerate PEFT for LLMs. Long Exposure comprises three key components: Shadowy-sparsity Exposer employs a prolonged sensing range to capture more sparsity details under shadowy sparsity; Sequence-oriented Predictor provides efficient yet accurate predictions to handle large sequence inputs and constantly-evolving parameters; and Dynamic-aware Operator facilitates more structured computational patterns and coalesced memory accesses, addressing dynamic sparse operations. Extensive evaluations show that Long Exposure outperforms state-of-the-arts with up to a $2.49\times$ speedup in end-to-end fine-tuning, offering promising advancements in accelerating PEFT for LLMs.

Although language model-based chat systems are increasingly used in daily life, most Americans remain non-adopters of chat-based LLMs -- as of June 2025, 66% had never used ChatGPT. At the same time, LLM development and evaluation rely mainly on data from adopters (e.g., logs, preference data), focusing on the needs and tasks for a limited demographic group of adopters in terms of geographic location, education, and gender. In this position paper, we argue that incorporating non-adopter perspectives is essential for developing broadly useful and capable LLMs. We contend that relying on methods that focus primarily on adopters will risk missing a range of tasks and needs prioritized by non-adopters, entrenching inequalities in who benefits from LLMs, and creating oversights in model development and evaluation. To illustrate this claim, we conduct case studies with non-adopters and show: how non-adopter needs diverge from those of current users, how non-adopter needs point us towards novel reasoning tasks, and how to systematically integrate non-adopter needs via human-centered methods.

Parkinson's disease (PD) presents a growing global challenge, affecting over 10 million individuals, with prevalence expected to double by 2040. Early diagnosis remains difficult due to the late emergence of motor symptoms and limitations of traditional clinical assessments. In this study, we propose a novel pipeline that leverages keystroke dynamics as a non-invasive and scalable biomarker for remote PD screening and telemonitoring. Our methodology involves three main stages: (i) preprocessing of data from four distinct datasets, extracting four temporal signals and addressing class imbalance through the comparison of three methods; (ii) pre-training eight state-of-the-art deep-learning architectures on the two largest datasets, optimizing temporal windowing, stride, and other hyperparameters; (iii) fine-tuning on an intermediate-sized dataset and performing external validation on a fourth, independent cohort. Our results demonstrate that hybrid convolutional-recurrent and transformer-based models achieve strong external validation performance, with AUC-ROC scores exceeding 90% and F1-Score over 70%. Notably, a temporal convolutional model attains an AUC-ROC of 91.14% in external validation, outperforming existing methods that rely solely on internal validation. These findings underscore the potential of keystroke dynamics as a reliable digital biomarker for PD, offering a promising avenue for early detection and continuous monitoring.

Large Vision-Language Models (LVLMs) have achieved remarkable progress in multimodal perception and generation, yet their safety alignment remains a critical challenge.Existing defenses and vulnerable to multimodal jailbreaks, as visual inputs introduce new attack surfaces, reasoning chains lack safety supervision, and alignment often degrades under modality fusion.To overcome these limitation, we propose VisuoAlign, a framework for multi-modal safety alignment via prompt-guided tree search.VisuoAlign embeds safety constrains into the reasoning process through visual-textual interactive prompts, employs Monte Carlo Tree Search(MCTS) to systematically construct diverse safety-critical prompt trajectories, and introduces prompt-based scaling to ensure real-time risk detection and compliant responses.Extensive experiments demonstrate that VisuoAlign proactively exposes risks, enables comprehensive dataset generation, and significantly improves the robustness of LVLMs against complex cross-modal threats.

We present Lean Finder, a semantic search engine for Lean and mathlib that understands and aligns with the intents of mathematicians. We further align Lean Finder with mathematicians' preferences using In addition, Lean Finder is compatible with LLM-based theorem provers, bridging retrieval with formal reasoning. Advances in Lean and mathlib (De Moura et al., 2015; Moura & Ullrich, 2021) are turning mathematical discovery into a collaborative and verifiable research workflow. Despite these advances, state-of-the-art LLMs still cannot solve math research problems. Lean's syn tax, gram mar, and tac tics in cur a steep learn ing curve. All experiments and data processing were conducted outside Meta. Figure 1: In the evaluation with user queries, real users preferred Lean Finder in 81.6% of cases, compared with Consider the two queries below. Lean search engines handle (Gao et al., 2024a;b; Ju & Dong, 2025; Asher, 2025): Denote L/K a field extension, x, y in L are algebraic elements over K with the same minimal polynomial. I'm working with algebraic elements over a field extension and I have two elements, say x and y in L. I know x is algebraic over K, and I've shown that y is a root of the minimal polynomial of x. Does this imply that the minimal polynomials of x and y are actually equal? T arget Statement 2: 1 theorem eq_of_root {x y: L} (hx: IsAlgebraic K x) (h_ev: Polynomial.aeval y (minpoly K x) = 0): minpoly K y = minpoly K x):= -- proof omitted for brevity This user latent (motivation, perspective, abstraction) cannot be inferred or encoded by a purely syntactic informalization. Addressing this challenge calls for Lean search engines that can understand a mathematician's intent, not merely We defer a more rigorous analysis in Section 2.2, and ask our core question: Our approach analyzes and clusters public discussions, then synthesizes queries that simulate user intents (Section 3.1).

Existing storage systems lack visibility into workload intent, limiting their ability to adapt to the semantics of modern, large-scale data-intensive applications. This disconnect leads to brittle heuristics and fragmented, siloed optimizations. To address these limitations, we propose Intent-Driven Storage Systems (IDSS), a vision for a new paradigm where large language models (LLMs) infer workload and system intent from unstructured signals to guide adaptive and cross-layer parameter reconfiguration. IDSS provides holistic reasoning for competing demands, synthesizing safe and efficient decisions within policy guardrails. We present four design principles for integrating LLMs into storage control loops and propose a corresponding system architecture. Initial results on FileBench workloads show that IDSS can improve IOPS by up to 2.45X by interpreting intent and generating actionable configurations for storage components such as caching and prefetching. These findings suggest that, when constrained by guardrails and embedded within structured workflows, LLMs can function as high-level semantic optimizers, bridging the gap between application goals and low-level system control. IDSS points toward a future in which storage systems are increasingly adaptive, autonomous, and aligned with dynamic workload demands.

Large Language Models (LLMs) are gaining prominence in various fields, thanks to their ability to generate high- quality content from human instructions. This paper delves into the field of chip design using LLMs, specifically in Power- Performance-Area (PPA) optimization and the generation of accurate Verilog codes for circuit designs. We introduce a novel framework VeriPPA designed to optimize PPA and generate Verilog code using LLMs. Our method includes a two-stage process where the first stage focuses on improving the functional and syntactic correctness of the generated Verilog codes, while the second stage focuses on optimizing the Verilog codes to meet PPA constraints of circuit designs, a crucial element of chip design. Our framework achieves an 81.37% success rate in syntactic correctness and 62.06% in functional correctness for code genera- tion, outperforming current state-of-the-art (SOTA) methods. On the RTLLM dataset. On the VerilogEval dataset, our framework achieves 99.56% syntactic correctness and 43.79% functional correctness, also surpassing SOTA, which stands at 92.11% for syntactic correctness and 33.57% for functional correctness. Furthermore, Our framework able to optimize the PPA of the designs. These results highlight the potential of LLMs in handling complex technical areas and indicate an encouraging development in the automation of chip design processes.

--The unabated growth in AI workload demands is driving the need for concerted advances in compute, memory, and interconnect performance. As traditional semiconductor scaling slows, high-speed interconnects have emerged as the new scaling engine, enabling the creation of larger logical GPUs by linking many GPUs into a single, low-latency, high-bandwidth compute domain. While initial scale-up fabrics leveraged copper interconnects for their power and cost advantages, the maximum reach of passive electrical interconnects (approximately 1 meter) effectively limits the scale-up domain to within a single rack. The advent of 3D-stacked optics and logic offers a transformative, power-efficient scale-up solution for connecting hundreds of GPU packages (thousands of GPUs) across multiple data center racks. This work explores the design tradeoffs of scale-up technologies and demonstrates how frontier LLMs necessitate novel photonic solutions to achieve aggressive power and performance targets. We model the benefits of 3D CPO (Passage) enabled GPUs and switches within the scale-up domain when training Frontier Mixture of Experts (MoE) models exceeding one trillion parameters. Our results show that the substantial increases in bandwidth and radix enabled by 3D CPO allow for an 8X increase in scale-up capability. The race to build larger, more sophisticated AI models is pushing the limits of existing infrastructure. At the chip and package level, GPUs are constrained by shoreline, yields and power. These challenges have led to the development of large high-bandwidth, low-latency scale-up pods. These pods effectively combine hundreds of GPUs into a single logical GPU to facilitate a variety of parallelism strategies (e.g. Approaches like Mixture of Experts (MoE) [1] have pushed scale-up networks to their limits due to copper reach (1 meter), which constrains the number of GPUs that can be connected within a single network hop. With MoEs, an ensemble of specialized sub-networks work together through sparse activations to increase model capacity without significantly increasing computational requirements. The output of the selected experts are combined to create the final result.