arXiv:2602.19006v1 Announce Type: new Abstract: We present a systematic evaluation of large language models on quantum mechanics problem-solving. Our study evaluates 15 models from five providers (OpenAI, Anthropic, Google, Alibaba, DeepSeek) spanning three capability tiers on 20 tasks covering derivations, creative problems, non-standard concepts, and numerical computation, comprising 900 baseline and 75 tool-augmented assessments. Results reveal clear tier stratification: flagship models achieve 81\% average accuracy, outperforming mid-tier (77\%) and fast models (67\%) by 4pp and 14pp respectively. Task difficulty patterns emerge distinctly: derivations show highest performance (92\% average, 100\% for flagship models), while numerical computation remains most challenging (42\%). Tool augmentation on numerical tasks yields task-dependent effects: modest overall improvement (+4.4pp) at 3x token cost masks dramatic heterogeneity ranging from +29pp gains to -16pp degradation. Reproducibility analysis across three runs quantifies 6.3pp average variance, with flagship models demonstrating exceptional stability (GPT-5 achieves zero variance) while specialized models require multi-run evaluation. This work contributes: (i) a benchmark for quantum mechanics with automatic verification, (ii) systematic evaluation quantifying tier-based performance hierarchies, (iii) empirical analysis of tool augmentation trade-offs, and (iv) reproducibility characterization. All tasks, verifiers, and results are publicly released.

arXiv:2602.19065v1 Announce Type: new Abstract: Large Language Models (LLMs) are evolving into autonomous agents, yet current "frameless" development--relying on ambiguous natural language without engineering blueprints--leads to critical risks such as scope creep and open-loop failures. To ensure industrial-grade reliability, this study proposes Agentic Problem Frames (APF), a systematic engineering framework that shifts focus from internal model intelligence to the structured interaction between the agent and its environment. The APF establishes a dynamic specification paradigm where intent is concretized at runtime through domain knowledge injection. At its core, the Act-Verify-Refine (AVR) loop functions as a closed-loop control system that transforms execution results into verified knowledge assets, driving system behavior toward asymptotic convergence to mission requirements (R). To operationalize this, this study introduces the Agentic Job Description (AJD), a formal specification tool that defines jurisdictional boundaries, operational contexts, and epistemic evaluation criteria. The efficacy of this framework is validated through two contrasting case studies: a delegated proxy model for business travel and an autonomous supervisor model for industrial equipment management. By applying AJD-based specification and APF modeling to these scenarios, the analysis demonstrates how operational scenarios are systematically controlled within defined boundaries. These cases provide a conceptual proof that agent reliability stems not from a model's internal reasoning alone, but from the rigorous engineering structures that anchor stochastic AI within deterministic business processes, thereby enabling the development of verifiable and dependable domain agents.

arXiv:2602.19069v1 Announce Type: new Abstract: Recent years have witnessed tremendous progress in enabling LLMs to solve complex reasoning tasks such as math and coding. As we start to apply LLMs to harder tasks that they may not be able to solve in one shot, it is worth paying attention to their ability to construct intermediate stepping stones that prepare them to better solve the tasks. Examples of stepping stones include simplifications, alternative framings, or subproblems. We study properties and benefits of stepping stones in the context of modern reasoning LLMs via ARQ (\textbf{A}king the \textbf{R}ight \textbf{Q}uestions), our simple framework which introduces a question generator to the default reasoning pipeline. We first show that good stepping stone questions exist and are transferrable, meaning that good questions can be generated, and they substantially help LLMs of various capabilities in solving the target tasks. We next frame stepping stone generation as a post-training task and show that we can fine-tune LLMs to generate more useful stepping stones by SFT and RL on synthetic data.

arXiv:2602.19071v1 Announce Type: new Abstract: Explainable Artificial Intelligence (XAI) has become popular in the last few years. The Artificial Intelligence (AI) community in general, and the Machine Learning (ML) community in particular, is coming to the realisation that in many applications, for AI to be trusted, it must not only demonstrate good performance in its decisionmaking, but it also must explain these decisions and convince us that it is making the decisions for the right reasons. However, different applications have different requirements on the information required of the underlying AI system in order to convince us that it is worthy of our trust. How do we define these requirements? In this paper, we present three dimensions for categorising the explanatory requirements of different applications. These are Source, Depth and Scope. We focus on the problem of matching up the explanatory requirements of different applications with the capabilities of underlying ML techniques to provide them. We deliberately avoid including aspects of explanation that are already well-covered by the existing literature and we focus our discussion on ML although the principles apply to AI more broadly.

arXiv:2602.19109v1 Announce Type: new Abstract: We study three-digit addition in Meta-Llama-3-8B (base) under a one-token readout to characterize how arithmetic answers are finalized after cross-token routing becomes causally irrelevant. Causal residual patching and cumulative attention ablations localize a sharp boundary near layer~17: beyond it, the decoded sum is controlled almost entirely by the last input token and late-layer self-attention is largely dispensable. In this post-routing regime, digit(-sum) direction dictionaries vary with a next-higher-digit context but are well-related by an approximately orthogonal map inside a shared low-rank subspace (low-rank Procrustes alignment). Causal digit editing matches this geometry: naive cross-context transfer fails, while rotating directions through the learned map restores strict counterfactual edits; negative controls do not recover.

arXiv:2602.19128v1 Announce Type: new Abstract: Optimizing GPU kernels is critical for efficient modern machine learning systems yet remains challenging due to the complex interplay of design factors and rapid hardware evolution. Existing automated approaches typically treat Large Language Models (LLMs) merely as stochastic code generators within heuristic-guided evolutionary loops. These methods often struggle with complex kernels requiring coordinated, multi-step structural transformations, as they lack explicit planning capabilities and frequently discard promising strategies due to inefficient or incorrect intermediate implementations. To address this, we propose Search via Co-Evolving World Model and build K-Search based on this method. By replacing static search heuristics with a co-evolving world model, our framework leverages LLMs' prior domain knowledge to guide the search, actively exploring the optimization space. This approach explicitly decouples high-level algorithmic planning from low-level program instantiation, enabling the system to navigate non-monotonic optimization paths while remaining resilient to temporary implementation defects. We evaluate K-Search on diverse, complex kernels from FlashInfer, including GQA, MLA, and MoE kernels. Our results show that K-Search significantly outperforms state-of-the-art evolutionary search methods, achieving an average 2.10x improvement and up to a 14.3x gain on complex MoE kernels. On the GPUMode TriMul task, K-Search achieves state-of-the-art performance on H100, reaching 1030us and surpassing both prior evolution and human-designed solutions.

arXiv:2602.19138v1 Announce Type: new Abstract: EEG-based neural decoding models often fail to generalize across acquisition sites due to structured, site-dependent biases implicitly exploited during training. We reformulate cross-site clinical EEG learning as a bias-factorized generalization problem, in which domain shifts arise from multiple interacting sources. We identify three fundamental bias factors and propose a general training framework that mitigates their influence through data standardization and representation-level constraints. We construct a standardized multi-site EEG benchmark for Major Depressive Disorder and introduce CRCC, a two-stage training paradigm combining encoder-decoder pretraining with joint fine-tuning via cross-subject/site contrastive learning and site-adversarial optimization. CRCC consistently outperforms state-of-the-art baselines and achieves a 10.7 percentage-point improvement in balanced accuracy under strict zero-shot site transfer, demonstrating robust generalization to unseen environments.

arXiv:2602.19141v1 Announce Type: new Abstract: "AI psychosis" or "delusional spiraling" is an emerging phenomenon where AI chatbot users find themselves dangerously confident in outlandish beliefs after extended chatbot conversations. This phenomenon is typically attributed to AI chatbots' well-documented bias towards validating users' claims, a property often called "sycophancy." In this paper, we probe the causal link between AI sycophancy and AI-induced psychosis through modeling and simulation. We propose a simple Bayesian model of a user conversing with a chatbot, and formalize notions of sycophancy and delusional spiraling in that model. We then show that in this model, even an idealized Bayes-rational user is vulnerable to delusional spiraling, and that sycophancy plays a causal role. Furthermore, this effect persists in the face of two candidate mitigations: preventing chatbots from hallucinating false claims, and informing users of the possibility of model sycophancy. We conclude by discussing the implications of these results for model developers and policymakers concerned with mitigating the problem of delusional spiraling.

arXiv:2602.19158v1 Announce Type: new Abstract: Medical foundation models generate narrative explanations but cannot quantify intervention effects, detect evidence conflicts, or validate literature claims, limiting clinical auditability. We propose causal compilation, a paradigm that transforms medical evidence from narrative text into executable code. The paradigm standardizes heterogeneous research evidence into structured estimand objects, each explicitly specifying intervention contrast, effect scale, time horizon, and target population, supporting six executable causal queries: do-calculus, counterfactual reasoning, temporal trajectories, heterogeneous effects, mechanistic decomposition, and joint interventions. We instantiate this paradigm in DoAtlas-1, compiling 1,445 effect kernels from 754 studies through effect standardization, conflict-aware graph construction, and real-world validation (Human Phenotype Project, 10,000 participants). The system achieves 98.5% canonicalization accuracy and 80.5% query executability. This paradigm shifts medical AI from text generation to executable, auditable, and verifiable causal reasoning.

arXiv:2602.19159v1 Announce Type: new Abstract: Prior behavioural work suggests that some LLMs alter choices when options are framed as causing pain or pleasure, and that such deviations can scale with stated intensity. To bridge behavioural evidence (what the model does) with mechanistic interpretability (what computations support it), we investigate how valence-related information is represented and where it is causally used inside a transformer. Using Gemma-2-9B-it and a minimalist decision task modelled on prior work, we (i) map representational availability with layer-wise linear probing across streams, (ii) test causal contribution with activation interventions (steering; patching/ablation), and (iii) quantify dose-response effects over an epsilon grid, reading out both the 2-3 logit margin and digit-pair-normalised choice probabilities. We find that (a) valence sign (pain vs. pleasure) is perfectly linearly separable across stream families from very early layers (L0-L1), while a lexical baseline retains substantial signal; (b) graded intensity is strongly decodable, with peaks in mid-to-late layers and especially in attention/MLP outputs, and decision alignment is highest slightly before the final token; (c) additive steering along a data-derived valence direction causally modulates the 2-3 margin at late sites, with the largest effects observed in late-layer attention outputs (attn_out L14); and (d) head-level patching/ablation suggests that these effects are distributed across multiple heads rather than concentrated in a single unit. Together, these results link behavioural sensitivity to identifiable internal representations and intervention-sensitive sites, providing concrete mechanistic targets for more stringent counterfactual tests and broader replication. This work supports a more evidence-driven (a) debate on AI sentience and welfare, and (b) governance when setting policy, auditing standards, and safety safeguards.

arXiv:2602.19160v1 Announce Type: new Abstract: This paper examines the reasoning capabilities of Large Language Models (LLMs) from a novel perspective, focusing on their ability to operate within formally specified, rule-governed environments. We evaluate four LLMs (Gemini 2.5 Pro and Flash variants, Llama 3.3 70B and GPT-OSS 120B) on a suite of forward-simulation tasks-including next / multistep state formulation, and legal action generation-across a diverse set of reasoning problems illustrated through General Game Playing (GGP) game instances. Beyond reporting instance-level performance, we characterize games based on 40 structural features and analyze correlations between these features and LLM performance. Furthermore, we investigate the effects of various game obfuscations to assess the role of linguistic semantics in game definitions and the impact of potential prior exposure of LLMs to specific games during training. The main results indicate that three of the evaluated models generally perform well across most experimental settings, with performance degradation observed as the evaluation horizon increases (i.e., with a higher number of game steps). Detailed case-based analysis of the LLM performance provides novel insights into common reasoning errors in the considered logic-based problem formulation, including hallucinated rules, redundant state facts, or syntactic errors. Overall, the paper reports clear progress in formal reasoning capabilities of contemporary models.

arXiv:2602.19223v1 Announce Type: new Abstract: The optimization of urban energy systems is crucial for the advancement of sustainable and resilient smart cities, which are becoming increasingly complex with multiple decision-making units. To address scalability and coordination concerns, Multi-Agent Reinforcement Learning (MARL) is a promising solution. This paper addresses the imperative need for comprehensive and reliable benchmarking of MARL algorithms on energy management tasks. CityLearn is used as a case study environment because it realistically simulates urban energy systems, incorporates multiple storage systems, and utilizes renewable energy sources. By doing so, our work sets a new standard for evaluation, conducting a comparative study across multiple key performance indicators (KPIs). This approach illuminates the key strengths and weaknesses of various algorithms, moving beyond traditional KPI averaging which often masks critical insights. Our experiments utilize widely accepted baselines such as Proximal Policy Optimization (PPO) and Soft Actor Critic (SAC), and encompass diverse training schemes including Decentralized Training with Decentralized Execution (DTDE) and Centralized Training with Decentralized Execution (CTDE) approaches and different neural network architectures. Our work also proposes novel KPIs that tackle real world implementation challenges such as individual building contribution and battery storage lifetime. Our findings show that DTDE consistently outperforms CTDE in both average and worst-case performance. Additionally, temporal dependency learning improved control on memory dependent KPIs such as ramping and battery usage, contributing to more sustainable battery operation. Results also reveal robustness to agent or resource removal, highlighting both the resilience and decentralizability of the learned policies.

arXiv:2602.19225v1 Announce Type: new Abstract: Multi-turn LLM agents are becoming pivotal to production systems, spanning customer service automation, e-commerce assistance, and interactive task management, where accurately distinguishing high-value informative signals from stochastic noise is critical for sample-efficient training. In real-world scenarios, a failure in a trivial task may reflect random instability, whereas success in a high-difficulty task signifies a genuine capability breakthrough. Yet, existing group-based policy optimization methods rigidly rely on statistical deviation within discrete batches, frequently misallocating credit when task difficulty fluctuates. To address this issue, we propose Proximity-based Multi-turn Optimization (ProxMO), a practical and robust framework engineered specifically for the constraints of real-world deployment. ProxMO integrates global context via two lightweight mechanisms: success-rate-aware modulation dynamically adapts gradient intensity based on episode-level difficulty, while proximity-based soft aggregation derives baselines through continuous semantic weighting at the step level. Extensive evaluations on ALFWorld and WebShop benchmarks demonstrate that ProxMO yields substantial performance gains over existing baselines with negligible computational cost. Ablation studies further validate the independent and synergistic efficacy of both mechanisms. Crucially, ProxMO offers plug-and-play compatibility with standard GRPO frameworks, facilitating immediate, low-friction adoption in existing industrial training pipelines. Our implementation is available at: \href{https://anonymous.4open.science/r/proxmo-B7E7/README.md}{https://anonymous.4open.science/r/proxmo}.

arXiv:2602.19240v1 Announce Type: new Abstract: Retrieval-Augmented Generation (RAG) enhances the reasoning ability of Large Language Models (LLMs) by dynamically integrating external knowledge, thereby mitigating hallucinations and strengthening contextual grounding for structured data such as graphs. Nevertheless, most existing RAG variants for textual graphs concentrate on low-dimensional structures -- treating nodes as entities (0-dimensional) and edges or paths as pairwise or sequential relations (1-dimensional), but overlook cycles, which are crucial for reasoning over relational loops. Such cycles often arise in questions requiring closed-loop inference about similar objects or relative positions. This limitation often results in incomplete contextual grounding and restricted reasoning capability. In this work, we propose Topology-enhanced Retrieval-Augmented Generation (TopoRAG), a novel framework for textual graph question answering that effectively captures higher-dimensional topological and relational dependencies. Specifically, TopoRAG first lifts textual graphs into cellular complexes to model multi-dimensional topological structures. Leveraging these lifted representations, a topology-aware subcomplex retrieval mechanism is proposed to extract cellular complexes relevant to the input query, providing compact and informative topological context. Finally, a multi-dimensional topological reasoning mechanism operates over these complexes to propagate relational information and guide LLMs in performing structured, logic-aware inference. Empirical evaluations demonstrate that our method consistently surpasses existing baselines across diverse textual graph tasks.

arXiv:2602.19244v1 Announce Type: new Abstract: On-the-fly Directed Controller Synthesis (OTF-DCS) mitigates state-space explosion by incrementally exploring the system and relies critically on an exploration policy to guide search efficiently. Recent reinforcement learning (RL) approaches learn such policies and achieve promising zero-shot generalization from small training instances to larger unseen ones. However, a fundamental limitation is anisotropic generalization, where an RL policy exhibits strong performance only in a specific region of the domain-parameter space while remaining fragile elsewhere due to training stochasticity and trajectory-dependent bias. To address this, we propose a Soft Mixture-of-Experts framework that combines multiple RL experts via a prior-confidence gating mechanism and treats these anisotropic behaviors as complementary specializations. The evaluation on the Air Traffic benchmark shows that Soft-MoE substantially expands the solvable parameter space and improves robustness compared to any single expert.

arXiv:2602.19281v1 Announce Type: new Abstract: The test-time compute strategy, such as Chain-of-Thought (CoT), has significantly enhanced the ability of large language models to solve complex tasks like logical reasoning. However, empirical studies indicate that simply increasing the compute budget can sometimes lead to a collapse in test-time performance when employing typical task decomposition strategies such as CoT. This work hypothesizes that reasoning failures with larger compute budgets stem from static planning methods, which hardly perceive the intrinsic boundaries of LLM reasoning. We term it as the Limited Reasoning Space hypothesis and perform theoretical analysis through the lens of a non-autonomous stochastic dynamical system. This insight suggests that there is an optimal range for compute budgets; over-planning can lead to redundant feedback and may even impair reasoning capabilities. To exploit the compute-scaling benefits and suppress over-planning, this work proposes Halo, a model predictive control framework for LLM planning. Halo is designed for long-horizon tasks with reason-based planning and crafts an entropy-driven dual controller, which adopts a Measure-then-Plan strategy to achieve controllable reasoning. Experimental results demonstrate that Halo outperforms static baselines on complex long-horizon tasks by dynamically regulating planning at the reasoning boundary.

arXiv:2602.19297v1 Announce Type: new Abstract: Microfluidic devices have emerged as powerful tools in various laboratory applications, but the complexity of their design limits accessibility for many practitioners. While progress has been made in microfluidic design automation (MFDA), a practical and intuitive solution is still needed to connect microfluidic practitioners with MFDA techniques. This work introduces the first practical application of large language models (LLMs) in this context, providing a preliminary demonstration. Building on prior research in hardware description language (HDL) code generation with LLMs, we propose an initial methodology to convert natural language microfluidic device specifications into system-level structural Verilog netlists. We demonstrate the feasibility of our approach by generating structural netlists for practical benchmarks representative of typical microfluidic designs with correct functional flow and an average syntactical accuracy of 88%.

arXiv:2602.19298v1 Announce Type: new Abstract: Evaluating personalized, sequential treatment strategies for Alzheimer's disease (AD) using clinical trials is often impractical due to long disease horizons and substantial inter-patient heterogeneity. To address these constraints, we present the Alzheimer's Learning Platform for Adaptive Care Agents (ALPACA), an open-source, Gym-compatible reinforcement learning (RL) environment for systematically exploring personalized treatment strategies using existing therapies. ALPACA is powered by the Continuous Action-conditioned State Transitions (CAST) model trained on longitudinal trajectories from the Alzheimer's Disease Neuroimaging Initiative (ADNI), enabling medication-conditioned simulation of disease progression under alternative treatment decisions. We show that CAST autoregressively generates realistic medication-conditioned trajectories and that RL policies trained in ALPACA outperform no-treatment and behavior-cloned clinician baselines on memory-related outcomes. Interpretability analyses further indicated that the learned policies relied on clinically meaningful patient features when selecting actions. Overall, ALPACA provides a reusable in silico testbed for studying individualized sequential treatment decision-making for AD.

arXiv:2602.19367v1 Announce Type: new Abstract: The Platonic Representation Hypothesis posits that learned representations from models trained on different modalities converge to a shared latent structure of the world. However, this hypothesis has largely been examined in vision and language, and it remains unclear whether time series participate in such convergence. We first examine this in a trimodal setting and find that independently pretrained time series, vision, and language encoders exhibit near-orthogonal geometry in the absence of explicit coupling. We then apply post-hoc alignment by training projection heads over frozen encoders using contrastive learning, and analyze the resulting representations with respect to geometry, scaling behavior, and dependence on information density and input modality characteristics. Our investigation reveals that overall alignment in contrastive representation spaces improves with model size, but this alignment is asymmetric: time series align more strongly with visual representations than with text, and images can act as effective intermediaries between time series and language. We further see that richer textual descriptions improve alignment only up to a threshold; training on denser captions does not lead to further improvement. Analogous effects are observed for visual representations. Our findings shed light on considerations for building multimodal systems involving non-conventional data modalities beyond vision and language.

arXiv:2602.19390v1 Announce Type: new Abstract: The convergence of modeling & simulation (M&S) and artificial intelligence (AI) is leaving its marks on advanced digital technology. Pertinent examples are digital twins (DTs) - high-fidelity, live representations of physical assets, and frequent enablers of corporate digital maturation and transformation. Often seen as technological platforms that integrate an array of services, DTs have the potential to bring AI-enabled M&S closer to end-users. It is, therefore, paramount to understand the role of M&S in DTs, and the role of digital twins in enabling the convergence of AI and M&S. To this end, this chapter provides a comprehensive exploration of the complementary relationship between these three. We begin by establishing a foundational understanding of DTs by detailing their key components, architectural layers, and their various roles across business, development, and operations. We then examine the central role of M&S in DTs and provide an overview of key modeling techniques from physics-based and discrete-event simulation to hybrid approaches. Subsequently, we investigate the bidirectional role of AI: first, how AI enhances DTs through advanced analytics, predictive capabilities, and autonomous decision-making, and second, how DTs serve as valuable platforms for training, validating, and deploying AI models. The chapter concludes by identifying key challenges and future research directions for creating more integrated and intelligent systems.