Large Language Models (LLMs) face a crucial challenge from fixed context windows and inadequate memory management, leading to a severe shortage of long-term memory capabilities and limited personalization in the interactive experience with AI agents. To overcome this challenge, we innovatively propose a Memory Operating System, i.e., MemoryOS, to achieve comprehensive and efficient memory management for AI agents. Inspired by the memory management principles in operating systems, MemoryOS designs a hierarchical storage architecture and consists of four key modules: Memory Storage, Updating, Retrieval, and Generation. Specifically, the architecture comprises three levels of storage units: short-term memory, mid-term memory, and long-term personal memory. Key operations within MemoryOS include dynamic updates between storage units: short-term to mid-term updates follow a dialogue-chain-based FIFO principle, while mid-term to long-term updates use a segmented page organization strategy. Our pioneering MemoryOS enables hierarchical memory integration and dynamic updating. Extensive experiments on the LoCoMo benchmark show an average improvement of 49.11% on F1 and 46.18% on BLEU-1 over the baselines on GPT-4o-mini, showing contextual coherence and personalized memory retention in long conversations. The implementation code is open-sourced at https://github.com/BAI-LAB/MemoryOS.

We present STARFlow, a scalable generative model based on normalizing flows that achieves strong performance in high-resolution image synthesis. The core of STARFlow is Transformer Autoregressive Flow (TARFlow), which combines the expressive power of normalizing flows with the structured modeling capabilities of Autoregressive Transformers. We first establish the theoretical universality of TARFlow for modeling continuous distributions. Building on this foundation, we introduce several key architectural and algorithmic innovations to significantly enhance scalability: (1) a deep-shallow design, wherein a deep Transformer block captures most of the model representational capacity, complemented by a few shallow Transformer blocks that are computationally efficient yet substantially beneficial; (2) modeling in the latent space of pretrained autoencoders, which proves more effective than direct pixel-level modeling; and (3) a novel guidance algorithm that significantly boosts sample quality. Crucially, our model remains an end-to-end normalizing flow, enabling exact maximum likelihood training in continuous spaces without discretization. STARFlow achieves competitive performance in both class-conditional and text-conditional image generation tasks, approaching state-of-the-art diffusion models in sample quality. To our knowledge, this work is the first successful demonstration of normalizing flows operating effectively at this scale and resolution.

While Foundation Models provide a general tool for rapid content creation, they regularly require task-specific adaptation. Traditionally, this exercise involves careful curation of datasets and repeated fine-tuning of the underlying model. Fine-tuning techniques enable practitioners to adapt foundation models for many new applications but require expensive and lengthy training while being notably sensitive to hyperparameter choices. To overcome these limitations, we introduce Text-to-LoRA (T2L), a model capable of adapting large language models (LLMs) on the fly solely based on a natural language description of the target task. T2L is a hypernetwork trained to construct LoRAs in a single inexpensive forward pass. After training T2L on a suite of 9 pre-trained LoRA adapters (GSM8K, Arc, etc.), we show that the ad-hoc reconstructed LoRA instances match the performance of task-specific adapters across the corresponding test sets. Furthermore, T2L can compress hundreds of LoRA instances and zero-shot generalize to entirely unseen tasks. This approach provides a significant step towards democratizing the specialization of foundation models and enables language-based adaptation with minimal compute requirements. Our code is available at https://github.com/SakanaAI/text-to-lora

LLMs are bound to transform healthcare with advanced decision support and flexible chat assistants. However, LLMs are prone to generate inaccurate medical content. To ground LLMs in high-quality medical knowledge, LLMs have been equipped with external knowledge via RAG, where unstructured medical knowledge is split into small text chunks that can be selectively retrieved and integrated into the LLMs context. Yet, existing RAG pipelines rely on raw, unstructured medical text, which can be noisy, uncurated and difficult for LLMs to effectively leverage. Systematic approaches to organize medical knowledge to best surface it to LLMs are generally lacking. To address these challenges, we introduce MIRIAD, a large-scale, curated corpus of 5,821,948 medical QA pairs, each rephrased from and grounded in a passage from peer-reviewed medical literature using a semi-automated pipeline combining LLM generation, filtering, grounding, and human annotation. Unlike prior medical corpora, which rely on unstructured text, MIRIAD encapsulates web-scale medical knowledge in an operationalized query-response format, which enables more targeted retrieval. Experiments on challenging medical QA benchmarks show that augmenting LLMs with MIRIAD improves accuracy up to 6.7% compared to unstructured RAG baselines with the same source corpus and with the same amount of retrieved text. Moreover, MIRIAD improved the ability of LLMs to detect medical hallucinations by 22.5 to 37% (increase in F1 score). We further introduce MIRIAD-Atlas, an interactive map of MIRIAD spanning 56 medical disciplines, enabling clinical users to visually explore, search, and refine medical knowledge. MIRIAD promises to unlock a wealth of down-stream applications, including medical information retrievers, enhanced RAG applications, and knowledge-grounded chat interfaces, which ultimately enables more reliable LLM applications in healthcare.

Significant investments have been made towards the com-modification of diffusion models for generation of diverse media. Their mass-market adoption is however still hobbled by the intense hardware resource requirements of diffusion model inference. Model quantization strategies tailored specifically towards diffusion models have been useful in easing this burden, yet have generally explored the Uniform Scalar Quantization (USQ) family of quantization methods. In contrast, V ector Quantization (VQ) methods, which operate on groups of multiple related weights as the basic unit of compression, have seen substantial success in Large Language Model (LLM) quantization. In this work, we apply codebook-based additive vector quantization to the problem of diffusion model compression. Our resulting approach achieves a new Pareto frontier for the extremely low-bit weight quantization on the standard class-conditional benchmark of LDM-4 on ImageNet at 20 inference time steps. Notably, we report sFID 1.92 points lower than the full-precision model at W4A8 and the best-reported results for FID, sFID and ISC at W2A8. W e are also able to demonstrate FLOPs savings on arbitrary hardware via an efficient inference kernel, as opposed to savings resulting from small integer operations which may lack broad hardware support.

We identify \textbf{Cross-lingual Collapse}, a systematic drift in which the chain-of-thought (CoT) of a multilingual language model reverts to its dominant pre-training language even when the prompt is expressed in a different language. Recent large language models (LLMs) with reinforcement learning with verifiable reward (RLVR) have achieved strong logical reasoning performances by exposing their intermediate reasoning traces, giving rise to large reasoning models (LRMs). However, the mechanism behind multilingual reasoning in LRMs is not yet fully explored. To investigate the issue, we fine-tune multilingual LRMs with Group-Relative Policy Optimization (GRPO) on translated versions of the GSM$8$K and SimpleRL-Zoo datasets in three different languages: Chinese, Korean, and Ukrainian. During training, we monitor both task accuracy and language consistency of the reasoning chains. Our experiments reveal three key findings: (i) GRPO rapidly amplifies pre-training language imbalances, leading to the erosion of low-resource languages within just a few hundred updates; (ii) language consistency reward mitigates this drift but does so at the expense of an almost 5 - 10 pp drop in accuracy. and (iii) the resulting language collapse is severely damaging and largely irreversible, as subsequent fine-tuning struggles to steer the model back toward its original target-language reasoning capabilities. Together, these findings point to a remarkable conclusion: \textit{not all languages are trained equally for reasoning}. Furthermore, our paper sheds light on the roles of reward shaping, data difficulty, and pre-training priors in eliciting multilingual reasoning.

Event Causality Identification (ECI) aims to detect causal relationships between events in textual contexts. Existing ECI models predominantly rely on supervised methodologies, suffering from dependence on large-scale annotated data. Although Large Language Models (LLMs) enable zero-shot ECI, they are prone to causal hallucination-erroneously establishing spurious causal links. To address these challenges, we propose MEFA, a novel zero-shot framework based on Multi-source Evidence Fuzzy Aggregation. First, we decompose causality reasoning into three main tasks (temporality determination, necessity analysis, and sufficiency verification) complemented by three auxiliary tasks. Second, leveraging meticulously designed prompts, we guide LLMs to generate uncertain responses and deterministic outputs. Finally, we quantify LLM's responses of sub-tasks and employ fuzzy aggregation to integrate these evidence for causality scoring and causality determination. Extensive experiments on three benchmarks demonstrate that MEFA outperforms second-best unsupervised baselines by 6.2% in F1-score and 9.3% in precision, while significantly reducing hallucination-induced errors. In-depth analysis verify the effectiveness of task decomposition and the superiority of fuzzy aggregation.

We aim at designing language agents with greater autonomy for crystal materials discovery. While most of existing studies restrict the agents to perform specific tasks within predefined workflows, we aim to automate workflow planning given high-level goals and scientist intuition. To this end, we propose Materials Agent unifying Planning, Physics, and Scientists, known as MAPPS. MAPPS consists of a Workflow Planner, a Tool Code Generator, and a Scientific Mediator. The Workflow Planner uses large language models (LLMs) to generate structured and multi-step workflows. The Tool Code Generator synthesizes executable Python code for various tasks, including invoking a force field foundation model that encodes physics. The Scientific Mediator coordinates communications, facilitates scientist feedback, and ensures robustness through error reflection and recovery. By unifying planning, physics, and scientists, MAPPS enables flexible and reliable materials discovery with greater autonomy, achieving a five-fold improvement in stability, uniqueness, and novelty rates compared with prior generative models when evaluated on the MP-20 data. We provide extensive experiments across diverse tasks to show that MAPPS is a promising framework for autonomous materials discovery.

Recent advancements in AI reasoning have driven substantial improvements across diverse tasks. A critical open question is whether these improvements also yields better knowledge transfer: the ability of models to communicate reasoning in ways humans can understand, apply, and learn from. To investigate this, we introduce Knowledge Integration and Transfer Evaluation (KITE), a conceptual and experimental framework for Human-AI knowledge transfer capabilities and conduct the first large-scale human study (N=118) explicitly designed to measure it. In our two-phase setup, humans first ideate with an AI on problem-solving strategies, then independently implement solutions, isolating model explanations' influence on human understanding. Our findings reveal that although model benchmark performance correlates with collaborative outcomes, this relationship is notably inconsistent, featuring significant outliers, indicating that knowledge transfer requires dedicated optimization. Our analysis identifies behavioral and strategic factors mediating successful knowledge transfer. We release our code, dataset, and evaluation framework to support future work on communicatively aligned models.

Large Language Model (LLM) safeguards, which implement request refusals, have become a widely adopted mitigation strategy against misuse. At the intersection of adversarial machine learning and AI safety, safeguard red teaming has effectively identified critical vulnerabilities in state-of-the-art refusal-trained LLMs. However, in our view the many conference submissions on LLM red teaming do not, in aggregate, prioritize the right research problems. First, testing against clear product safety specifications should take a higher priority than abstract social biases or ethical principles. Second, red teaming should prioritize realistic threat models that represent the expanding risk landscape and what real attackers might do. Finally, we contend that system-level safety is a necessary step to move red teaming research forward, as AI models present new threats as well as affordances for threat mitigation (e.g., detection and banning of malicious users) once placed in a deployment context. Adopting these priorities will be necessary in order for red teaming research to adequately address the slate of new threats that rapid AI advances present today and will present in the very near future.