Ensuring that large language models (LLMs) comply with safety requirements is a central challenge in AI deployment. Existing alignment approaches primarily operate during training, such as through fine-tuning or reinforcement learning from human feedback, but these methods are costly and inflexible, requiring retraining whenever new requirements arise. Recent efforts toward inference-time alignment mitigate some of these limitations but still assume access to model internals, which is impractical, and not suitable for third party stakeholders who do not have access to the models. In this work, we propose a model-independent, black-box framework for safety alignment that does not require retraining or access to the underlying LLM architecture. As a proof of concept, we address the problem of trading off between generating safe but uninformative answers versus helpful yet potentially risky ones. We formulate this dilemma as a two-player zero-sum game whose minimax equilibrium captures the optimal balance between safety and helpfulness. LLM agents operationalize this framework by leveraging a linear programming solver at inference time to compute equilibrium strategies. Our results demonstrate the feasibility of black-box safety alignment, offering a scalable and accessible pathway for stakeholders, including smaller organizations and entities in resource-constrained settings, to enforce safety across rapidly evolving LLM ecosystems.

Scaling unsupervised skill discovery algorithms to high-DoF agents remains challenging. As dimensionality increases, the exploration space grows exponentially, while the manifold of meaningful skills remains limited. Therefore, semantic meaningfulness becomes essential to effectively guide exploration in high-dimensional spaces. In this work, we present **Reference-Grounded Skill Discovery (RGSD)**, a novel algorithm that grounds skill discovery in a semantically meaningful latent space using reference data. RGSD first performs contrastive pretraining to embed motions on a unit hypersphere, clustering each reference trajectory into a distinct direction. This grounding enables skill discovery to simultaneously involve both imitation of reference behaviors and the discovery of semantically related diverse behaviors. On a simulated SMPL humanoid with $359$-D observations and $69$-D actions, RGSD successfully imitates skills such as walking, running, punching, and sidestepping, while also discover variations of these behaviors. In downstream locomotion tasks, RGSD leverages the discovered skills to faithfully satisfy user-specified style commands and outperforms imitation-learning baselines, which often fail to maintain the commanded style. Overall, our results suggest that lightweight reference-grounding offers a practical path to discovering semantically rich and structured skills in high-DoF systems.

Large Multimodal Models (LMMs) are inherently modular, consisting of vision and audio encoders, projectors, and large language models. Yet, they are almost always executed monolithically, which underutilizes the heterogeneous accelerators (NPUs, GPUs, DSPs) in modern SoCs and leads to high end-to-end latency. In this paper, we present NANOMIND, a hardware--software co-design inference framework for Large Multimodal Models (LMMs) that breaks large models into modular ``bricks'' (vision, language, audio, etc.) and maps each to its ideal accelerator. The key insight is that large models can be broken into modular components and scheduled to run on the most appropriate compute units. It performs module-level dynamic offloading across accelerators on unified-memory SoCs. By combining customized hardware design, system-level scheduling, and optimized low-bit computation kernels, we demonstrate our framework with a compact, battery-powered device capable of running LMMs entirely on device. This prototype functions as a self-contained intelligent assistant that requires no network connectivity, while achieving higher throughput and superior power efficiency under strict resource constraints. The design further bypasses CPU bottlenecks and reduces redundant memory usage through token-aware buffer management and module-level coordination. Our system outperforms existing implementations in resource efficiency, cutting energy consumption by 42.3\% and GPU memory usage by 11.2\%. This enables a battery-powered device to run LLaVA-OneVision with a camera for nearly 20.8 hours.

Recent large language models (LLMs) have shown strong reasoning capabilities. However, a critical question remains: do these models possess genuine reasoning skills particularly complex strategic reasoning or are they primarily excelling at sophisticated pattern recognition within their training data? To address this question, this paper presents a chess testbed, ChessArena, to evaluate the strategic reasoning capabilities of LLMs. Chess requires complex strategic reasoning capabilities including long-term planning, strict rule comprehension, and multi-turn conversation memorization. Specifically, ChessArena is a competitive framework where LLMs play against each other, under four different play modes. The testbed is equipped with a ranking algorithm and a leaderboard. The testbed can also evaluate fine-grained capabilities including basic understanding, move selection, and puzzle solving. Over 13 LLMs with different modes are evaluated in ChessArena, playing over 800 games. The results reveal significant shortcomings in current LLMs: no model can beat Maia-1100 (a chess engine at human amateur level), while some even failed to defeat a random player that selects moves arbitrarily. We also present a strong baseline to the testbed: our fine-tuned Qwen3-8B substantially improved performance, approaching much larger state-of-the-art reasoning models.

Progress in large language models is constrained by an evaluation bottleneck: build a benchmark, run models, then iterate. We ask a question: can we forecast outcomes before running any experiments to inform earlier study design? For example, a team building an AI assistant for a certain task can estimate whether expected performance is around 50 or closer to 80, evidence that supports whether to proceed to a pilot study, how to scope it, and how to allocate resources. We study text-only performance forecasting, where a model predicts a score from a redacted task description and intended configuration, with no access to dataset instances. To support systematic study, we curate PRECOG, a corpus of redacted description-performance pairs spanning diverse tasks, domains, and metrics. We scrape task and configuration descriptions from arXiv, yielding 2,290 instances covering 1,519 papers, and construct a leakage free test split using papers published after the knowledge cutoff of the evaluated models. Experiments show the task is challenging but feasible: reasoning models achieve moderate prediction performance with well calibrated uncertainty, reaching mean absolute error as low as 9.9 at high confidence thresholds. We further test a zero-leakage setting, forecasting on newly released datasets or experiments before their papers are indexed, where GPT5 with built in web search still attains nontrivial prediction accuracy. Overall, our corpus and analyses offer an initial step toward open ended anticipatory evaluation, supporting difficulty estimation and smarter experiment prioritization.

Data scarcity remains one of the most limiting factors in driving progress in robotics. However, the amount of available robotics data in the wild is growing exponentially, creating new opportunities for large-scale data utilization. Reliable temporal task completion prediction could help automatically annotate and curate this data at scale. The Generative Value Learning (GVL) approach was recently proposed, leveraging the knowledge embedded in vision-language models (VLMs) to predict task progress from visual observations. Building upon GVL, we propose OpenGVL, a comprehensive benchmark for estimating task progress across diverse challenging manipulation tasks involving both robotic and human embodiments. We evaluate the capabilities of publicly available open-source foundation models, showing that open-source model families significantly underperform closed-source counterparts, achieving only approximately $70\%$ of their performance on temporal progress prediction tasks. Furthermore, we demonstrate how OpenGVL can serve as a practical tool for automated data curation and filtering, enabling efficient quality assessment of large-scale robotics datasets. We release the benchmark along with the complete codebase at \href{github.com/budzianowski/opengvl}{OpenGVL}.

We present ParlAI Vote, an interactive web platform for exploring European Parliament debates and votes, and for testing LLMs on vote prediction and bias analysis. This web system connects debate topics, speeches, and roll-call outcomes, and includes rich demographic data such as gender, age, country, and political group. Users can browse debates, inspect linked speeches, compare real voting outcomes with predictions from frontier LLMs, and view error breakdowns by demographic group. Visualizing the EuroParlVote benchmark and its core tasks of gender classification and vote prediction, ParlAI Vote highlights systematic performance bias in state-of-the-art LLMs. It unifies data, models, and visual analytics in a single interface, lowering the barrier for reproducing findings, auditing behavior, and running counterfactual scenarios. This web platform also shows model reasoning, helping users see why errors occur and what cues the models rely on. It supports research, education, and public engagement with legislative decision-making, while making clear both the strengths and the limitations of current LLMs in political analysis.

We present Apertus, a fully open suite of large language models (LLMs) designed to address two systemic shortcomings in today's open model ecosystem: data compliance and multilingual representation. Unlike many prior models that release weights without reproducible data pipelines or regard for content-owner rights, Apertus models are pretrained exclusively on openly available data, retroactively respecting `robots.txt` exclusions and filtering for non-permissive, toxic, and personally identifiable content. To mitigate risks of memorization, we adopt the Goldfish objective during pretraining, strongly suppressing verbatim recall of data while retaining downstream task performance. The Apertus models also expand multilingual coverage, training on 15T tokens from over 1800 languages, with ~40% of pretraining data allocated to non-English content. Released at 8B and 70B scales, Apertus approaches state-of-the-art results among fully open models on multilingual benchmarks, rivalling or surpassing open-weight counterparts. Beyond model weights, we release all scientific artifacts from our development cycle with a permissive license, including data preparation scripts, checkpoints, evaluation suites, and training code, enabling transparent audit and extension.

Large language models demonstrate impressive results across diverse tasks but are still known to hallucinate, generating linguistically plausible but incorrect answers to questions. Uncertainty quantification has been proposed as a strategy for hallucination detection, requiring estimates for both global uncertainty (attributed to a batch of responses) and local uncertainty (attributed to individual responses). While recent black-box approaches have shown some success, they often rely on disjoint heuristics or graph-theoretic approximations that lack a unified geometric interpretation. We introduce a geometric framework to address this, based on archetypal analysis of batches of responses sampled with only black-box model access. At the global level, we propose Geometric V olume, which measures the convex hull volume of archetypes derived from response embeddings. At the local level, we propose Geometric Suspicion, which leverages the spatial relationship between responses and these archetypes to rank reliability, enabling hallucination reduction through preferential response selection. Unlike prior methods that rely on discrete pairwise comparisons, our approach provides continuous semantic boundary points which have utility for attributing reliability to individual responses. Experiments show that our framework performs comparably to or better than prior methods on short form question-answering datasets, and achieves superior results on medical datasets where hallucinations carry particularly critical risks. We also provide theoretical justification by proving a link between convex hull volume and entropy. Large language models (LLMs) have achieved remarkable performance across diverse natural language processing tasks (Guo et al., 2025; Anthropic, 2025; Gemini Team, Google DeepMind, 2025; OpenAI, 2025) and are increasingly applied in areas such as medical diagnosis, law, and financial advice (Y ang et al., 2025; Chen et al., 2024; Kong et al., 2024). Hallucinations, however, where models generate plausible but false or fabricated content, pose significant risks for adoption in high-stakes applications (Farquhar et al., 2024). Recent work, for example, finds GPT -4 hallucinating in 28.6% of reference generation tasks (Chelli et al., 2024).

Unmanned Aerial Vehicles (UAVs) are increasingly used in defense, surveillance, and disaster response, yet most systems still operate at SAE Level 2 to 3 autonomy. Their dependence on rule-based control and narrow AI limits adaptability in dynamic and uncertain missions. Current UAV architectures lack context-aware reasoning, autonomous decision-making, and integration with external systems. Importantly, none make use of Large Language Model (LLM) agents with tool-calling for real-time knowledge access. This paper introduces the Agentic UAVs framework, a five-layer architecture consisting of Perception, Reasoning, Action, Integration, and Learning. The framework enhances UAV autonomy through LLM-driven reasoning, database querying, and interaction with third-party systems. A prototype built with ROS 2 and Gazebo combines YOLOv11 for object detection with GPT-4 for reasoning and a locally deployed Gemma 3 model. In simulated search-and-rescue scenarios, agentic UAVs achieved higher detection confidence (0.79 compared to 0.72), improved person detection rates (91% compared to 75%), and a major increase in correct action recommendations (92% compared to 4.5%). These results show that modest computational overhead can enable significantly higher levels of autonomy and system-level integration.