arXiv:2602.12316v1 Announce Type: new Abstract: Frontier AI systems are increasingly capable and deployed in high-stakes multi-agent environments. However, existing AI safety benchmarks largely evaluate single agents, leaving multi-agent risks such as coordination failure and conflict poorly understood. We introduce GT-HarmBench, a benchmark of 2,009 high-stakes scenarios spanning game-theoretic structures such as the Prisoner’s Dilemma, Stag Hunt and Chicken. Scenarios are drawn from realistic AI risk contexts in the MIT AI Risk Repository. Across 15 frontier models, agents choose […]

Ver mais

arXiv:2601.09723v1 Announce Type: new Abstract: Large Language Models (LLMs) have shown significant progress, but understanding long and complex documents remains challenging. Many long-context benchmarks have been proposed, but they face several limitations, including task realism, data scalability, and data quality. To this end, we introduce SagaScale, a realistic, scalable, and high-quality long-context benchmark built from full-length novels. The entire benchmark is constructed using an automated data collection pipeline that utilizes external resources (e.g., Wikipedia pages) to curate question-answer […]

Ver mais

arXiv:2601.04216v1 Announce Type: new Abstract: This paper tackles practical challenges in governing child centered artificial intelligence: policy texts state principles and requirements but often lack reproducible evidence anchors, explicit causal pathways, executable governance toolchains, and computable audit metrics. We propose Graph-GAP, a methodology that decomposes requirements from authoritative policy texts into a four layer graph of evidence, mechanism, governance, and indicator, and that computes two metrics, GAP score and mitigation readiness, to identify governance gaps and prioritise actions. […]

Ver mais

The landscape of multimodal large language models (MLLMs) has shifted from experimental ‘wrappers’—where separate vision or audio encoders are stitched onto a text-based backbone—to native, end-to-end ‘omnimodal’ architectures. Alibaba Qwen team latest release, Qwen3.5-Omni, represents a significant milestone in this evolution. Designed as a direct competitor to flagship models like Gemini 3.1 Pro, the Qwen3.5-Omni series introduces a unified framework capable of processing text, images, audio, and video simultaneously within a single computational pipeline. The technical significance of […]

Ver mais

Machine learning models have recently emerged to predict whether hypothetical solid-state materials can be synthesized. These models aim to circumvent direct first-principles modeling of solid-state phase transformations, instead learning from large databases of successfully synthesized materials. Here, we assess the alignment of several recently introduced synthesis prediction models with material and reaction thermodynamics, quantified by the energy with respect to the convex hull and a metric accounting for thermodynamic selectivity of enumerated synthesis reactions. A dataset of successful […]

Ver mais

World models learned from high-dimensional visual observations allow agents to make decisions and plan directly in latent space, avoiding pixel-level reconstruction. However, recent latent predictive architectures (JEPAs), including the DINO world model (DINO-WM), display a degradation in test time robustness due to their sensitivity to "slow features". These include visual variations such as background changes and distractors that are irrelevant to the task being solved. We address this limitation by augmenting the predictive objective with a bisimulation encoder […]

Ver mais

Predictive Coding (PC) is a biologically-inspired learning framework characterised by local, parallelisable operations, properties that enable energy-efficient implementation on neuromorphic hardware. Despite this, extending PC effectively to recurrent neural networks (RNNs) has been challenging, particularly for tasks involving long-range temporal dependencies. Backpropagation Through Time (BPTT) remains the dominant method for training RNNs, but its non-local computation, lack of spatial parallelism, and requirement to store extensive activation histories results in significant energy consumption. This work introduces a novel method […]

Ver mais

Transformer-based large language models exhibit in-context learning, enabling adaptation to downstream tasks via few-shot prompting with demonstrations. In practice, such models are often fine-tuned to improve zero-shot performance on downstream tasks, allowing them to solve tasks without examples and thereby reducing inference costs. However, fine-tuning can degrade in-context learning, limiting the performance of fine-tuned models on tasks not seen during fine-tuning. Using linear attention models, we provide a theoretical analysis that characterizes how fine-tuning objectives modify attention parameters […]

Ver mais

Model merging enables multiple large language models (LLMs) to be combined into a single model while preserving performance. This makes it a valuable tool in LLM development, offering a competitive alternative to multi-task training. However, merging can be difficult at scale, as successful merging requires choosing the right merge operator, selecting the right models, and merging them in the right order. This often leads researchers to run expensive merge-and-evaluate searches to select the best merge. In this work, […]

Ver mais