Archives AI News

arXiv:2508.21058v1 Announce Type: cross Abstract: Long video generation is fundamentally a long context memory problem: models must retain and retrieve salient events across a long range without collapsing or drifting. However, scaling diffusion transformers to generate long-context videos is fundamentally limited by the quadratic cost of self-attention, which makes memory and computation intractable and difficult to optimize for long sequences. We recast long-context video generation as an internal information retrieval task and propose a simple, learnable sparse attention routing module, Mixture of Contexts (MoC), as an effective long-term memory retrieval engine. In MoC, each query dynamically selects a few informative chunks plus mandatory anchors (caption, local windows) to attend to, with causal routing that prevents loop closures. As we scale the data and gradually sparsify the routing, the model allocates compute to salient history, preserving identities, actions, and scenes over minutes of content. Efficiency follows as a byproduct of retrieval (near-linear scaling), which enables practical training and synthesis, and the emergence of memory and consistency at the scale of minutes.

arXiv:2508.20371v1 Announce Type: new Abstract: Machine-learning models are increasingly driving decisions in high-stakes settings, such as finance, law, and hiring, thus, highlighting the need for transparency. However, the key challenge is to balance transparency -- clarifying `why' a decision was made -- with recourse: providing actionable steps on `how' to achieve a favourable outcome from an unfavourable outcome. Counterfactual explanations reveal `why' an undesired outcome occurred and `how' to reverse it through targeted feature changes (interventions). Current counterfactual approaches have limitations: 1) they often ignore causal dependencies between features, and 2) they typically assume all interventions can happen simultaneously, an unrealistic assumption in practical scenarios where actions are typically taken in a sequence. As a result, these counterfactuals are often not achievable in the real world. We present P2C (Path-to-Counterfactuals), a model-agnostic framework that produces a plan (ordered sequence of actions) converting an unfavourable outcome to a causally consistent favourable outcome. P2C addresses both limitations by 1) Explicitly modelling causal relationships between features and 2) Ensuring that each intermediate state in the plan is feasible and causally valid. P2C uses the goal-directed Answer Set Programming system s(CASP) to generate the plan accounting for feature changes that happen automatically due to causal dependencies. Furthermore, P2C refines cost (effort) computation by only counting changes actively made by the user, resulting in realistic cost estimates. Finally, P2C highlights how its causal planner outperforms standard planners, which lack causal knowledge and thus can generate illegal actions.

arXiv:2508.08882v2 Announce Type: replace Abstract: Recent advances in multi-agent systems highlight the potential of specialized small agents that collaborate via division of labor. Existing tool-integrated reasoning systems, however, often follow a single-agent paradigm in which one large model interleaves long-horizon reasoning with precise tool operations, leading to cognitive-load interference and unstable coordination. We present MSARL, a Multi-Small-Agent Reinforcement Learning framework that explicitly decouples reasoning from tool use. In MSARL, a Reasoning Agent decomposes problems and plans tool invocations, while multiple Tool Agents specialize in specific external tools, each trained via a combination of imitation learning and reinforcement learning with role-specific rewards. On mathematical problem solving with code execution, MSARL significantly improves reasoning stability and final-answer accuracy over single-agent baselines. Moreover, the architecture generalizes to diverse tool-use tasks, demonstrating that cognitive-role decoupling with small agents is a scalable blueprint for multi-agent AI design.

arXiv:2508.20374v1 Announce Type: new Abstract: Diverse instruction data is vital for effective instruction tuning of large language models, as it enables the model to generalize across different types of inputs . Building such diversified instruction dataset is an essential step in this process. Existing approaches often leverage large language models to automatically explore and generate diverse instructions, ensuring both data diversity and quality. However, they tend to overlook an important factor in real-world applications: on-task relevance. In practice, only a few real-world applications require a truly general-purpose model; most benefit from task-specific knowledge tailored to their particular use case. Therefore, it is vital to develop instruction augmentation methods that not only maintain diversity but are also optimized for specific, real-world scenarios. We thus introduce Task Centric Instruction Augmentation (TCIA), a framework that systematically expands instructions while preserving both diversity and task alignment. By representing instructions in a discrete query-constraints space, TCIA creates a rich set of task-relevant instructions and enables models to generalize to these task-specific instructions without sacrificing overall performance. Experiments show that TCIA improves open-source LLMs' performance by an average of 8.7% across four real-world, task-specific applications, and in some cases outperforming leading closed-source models. These improvements do not compromise general instruction-following ability, making TCIA a scalable and efficient solution for adapting LLMs to real-world, task-focused applications.

arXiv:2411.14870v2 Announce Type: replace-cross Abstract: Context: With artificial intelligence (AI) being well established within the daily lives of research communities, we turn our gaze toward formal methods (FM). FM aim to provide sound and verifiable reasoning about problems in computer science. Objective: We conduct a systematic mapping study to overview the current landscape of research publications that apply AI to FM. We aim to identify how FM can benefit from AI techniques and highlight areas for further research. Our focus lies on the previous five years (2019-2023) of research. Method: Following the proposed guidelines for systematic mapping studies, we searched for relevant publications in four major databases, defined inclusion and exclusion criteria, and applied extensive snowballing to uncover potential additional sources. Results: This investigation results in 189 entries which we explored to find current trends and highlight research gaps. We find a strong focus on AI in the area of theorem proving while other subfields of FM are less represented. Conclusions: The mapping study provides a quantitative overview of the modern state of AI application in FM. The current trend of the field is yet to mature. Many primary studies focus on practical application, yet we identify a lack of theoretical groundwork, standard benchmarks, or case studies. Further, we identify issues regarding shared training data sets and standard benchmarks.

arXiv:2508.20384v1 Announce Type: new Abstract: In this work, we introduce Entropy Area Score (EAS), a simple yet effective metric to quantify uncertainty in the answer generation process of reasoning large language models (LLMs). EAS requires neither external models nor repeated sampling, it integrates token-level predictive entropy from the model itself to capture the evolution of uncertainty during generation. Empirical results show that EAS is strongly correlated with answer entropy across models and datasets. In training data selection, EAS identifies high-potential samples and consistently outperforms Pass Rate filtering under equal sample budgets, improving student model accuracy on math benchmarks. EAS is both efficient and interpretable, offering a practical tool for uncertainty modeling and data quality assessment in LLM training.

arXiv:2507.06056v2 Announce Type: replace-cross Abstract: Large Language Models (LLMs) are known to memorize portions of their training data, sometimes reproducing content verbatim when prompted appropriately. In this work, we investigate a fundamental yet under-explored question in the domain of memorization: How to characterize memorization difficulty of training data in LLMs? Through empirical experiments on OLMo, a family of open models, we present the Entropy-Memorization Law. It suggests that data entropy is linearly correlated with memorization score. Moreover, in a case study of memorizing highly randomized strings, or "gibberish", we observe that such sequences, despite their apparent randomness, exhibit unexpectedly low empirical entropy compared to the broader training corpus. Adopting the same strategy to discover Entropy-Memorization Law, we derive a simple yet effective approach to distinguish training and testing data, enabling Dataset Inference (DI).

arXiv:2508.20404v1 Announce Type: new Abstract: The learning from practice paradigm is crucial for developing capable Agentic AI systems, yet it is severely hampered by inefficient experience generation, a bottleneck especially pronounced in complex benchmarks like GAIA. To address this, we introduce AWorld, an open-source system engineered for large-scale agent-environment interaction. By distributing tasks across a cluster, AWorld accelerates experience collection by 14.6x compared to standard single-node, sequential execution. This critical speedup makes extensive reinforcement learning practical and scalable. Leveraging this capability, we trained a Qwen3-32B-based agent that significantly outperforms its base model, increasing its overall GAIA accuracy from 21.59% to 32.23%. On the benchmark's most challenging levels, our agent achieves a score of 16.33%, surpassing the performance of leading proprietary models. Our open-source system and resulting agent provide a practical blueprint for a complete agentic AI training pipeline, from efficient interaction to demonstrable model improvement.

arXiv:2508.19011v2 Announce Type: replace-cross Abstract: Most deep learning methods for imputing missing values treat the task as completing patterns within a fixed time window. This assumption often fails in industrial systems, where dynamics are driven by control actions, are highly non-stationary, and can experience long, uninterrupted gaps. We propose STDiff, which reframes imputation as learning how the system evolves from one state to the next. STDiff uses a conditional denoising diffusion model with a causal bias aligned to control theory, generating missing values step-by-step based on the most recent known state and relevant control or environmental inputs. On a public wastewater treatment dataset with simulated missing blocks, STDiff consistently achieves the lowest errors, with its advantage increasing for longer gaps. On a raw industrial dataset with substantial real gaps, it produces trajectories that remain dynamically plausible, in contrast to window-based models that tend to flatten or over-smooth. These results support dynamics-aware, explicitly conditioned imputation as a robust approach for industrial time series, and we discuss computational trade-offs and extensions to broader domains.

arXiv:2508.20411v1 Announce Type: new Abstract: As AI rapidly advances, the security risks posed by AI are becoming increasingly severe, especially in critical scenarios, including those posing existential risks. If AI becomes uncontrollable, manipulated, or actively evades safety mechanisms, it could trigger systemic disasters. Existing AI safety approaches-such as model enhancement, value alignment, and human intervention-suffer from fundamental, in-principle limitations when facing AI with extreme motivations and unlimited intelligence, and cannot guarantee security. To address this challenge, we propose a Governable AI (GAI) framework that shifts from traditional internal constraints to externally enforced structural compliance based on cryptographic mechanisms that are computationally infeasible to break, even for future AI, under the defined threat model and well-established cryptographic assumptions.The GAI framework is composed of a simple yet reliable, fully deterministic, powerful, flexible, and general-purpose rule enforcement module (REM); governance rules; and a governable secure super-platform (GSSP) that offers end-to-end protection against compromise or subversion by AI. The decoupling of the governance rules and the technical platform further enables a feasible and generalizable technical pathway for the safety governance of AI. REM enforces the bottom line defined by governance rules, while GSSP ensures non-bypassability, tamper-resistance, and unforgeability to eliminate all identified attack vectors. This paper also presents a rigorous formal proof of the security properties of this mechanism and demonstrates its effectiveness through a prototype implementation evaluated in representative high-stakes scenarios.