Academic Intelligence · Curated Daily

探索全球前沿学术脉络

AcademicHub 汇聚顶级期刊与预印本平台的实时文献。定制您的专属科研雷达,利用大语言模型自动生成交叉领域文献分析简报。

01.
medRxiv (Medicine) 2026-06-18

Early-life Urban Environment, Nutrition, and Pubertal Timing in Southern Europe: An Exposome Analysis

Background: Urban environmental and lifestyle factors during early life may influence pubertal timing, but the combined effects of multiple environmental exposures within an exposome analytical framework remain poorly understood. Objective: To examine the association between early-life urban environmental exposures and pubertal timing, and to explore whether these exposures interact with early-life nutritional factors, namely breastfeeding duration and childhood diet quality. Methods: Data from two European population-based birth cohorts were analysed: Generation XXI (G21, Portugal; n=5263; 51.5% girls) and INfancia y Medio Ambiente (INMA, Spain; n=1019; 50.1% girls). Urban environmental exposures including indicators of air pollution, traffic, built environment, and natural spaces were estimated at 4 early-life stages at both cohorts: pregnancy (INMA only), birth, 1 year, and 4-5 years of age. Pubertal development timing was assessed using Tanner staging and/or the Pubertal Development Scale (PDS), and age at menarche was self-reported. Exposome-Wide Association Study (ExWAS) models and unsupervised clustering followed by ordinal logistic regression models were used to examine single- and multi-exposure associations, respectively. Regression models were fitted adjusting for relevant child characteristics, maternal factors, and household socioeconomic conditions, and corrected for multiple testing. Results: Individuals living in more unfavourable urban environments characterised by higher building density, air pollution, and lower access to natural spaces showed earlier pubertal timing according to multiple outcomes, across multiple early-life exposure periods, and in both cohorts. In the G21 cohort, these environmental profiles were associated with earlier age at menarche, particularly for exposures at 1-1.5 and 4-5 years (e.g., 1-1.5y: {beta}=-0.172, FDR-adjusted p-value=0.041), while in the INMA cohort, boys exposed to more unfavourable environmental profiles showed more advanced pubertal development, also particularly for exposures at 1-1.5 and 4-5 years of age (e.g., 1-1.5y; {beta}=0.572, FDR-adjusted p-value=0.008). Among environmental domains, air pollution and traffic were the factors most consistently associated with pubertal timing. Regarding early-life nutritional factors, longer duration of exclusive breastfeeding was associated with a lower Tanner stage among girls in G21. No significant interactions between breastfeeding duration and environmental exposure clusters were observed. Conclusion: Early-life urban environmental exposures, particularly air pollution and traffic, may influence pubertal timing. Exclusive breastfeeding may have a protective role against earlier pubertal development. These findings highlight the importance of improving urban environmental conditions and promoting breastfeeding to support healthy developmental trajectories.

02.
arXiv (CS.CL) 2026-06-16

Building Customer Support AI Agents at 100M-User Scale: An Evaluation-Driven Framework

The rapid rise in LLM capabilities has made AI agents increasingly viable across a broad range of tasks. Among the most promising applications is building production-ready customer-facing agents, a challenge that demands coordinated excellence in evaluation methodology, context engineering, training, and online measurement. Yet these critical pillars are typically developed in isolation, creating blind spots that only surface after deployment. In this paper, we present a unified framework that bridges offline development with online impact for customer support AI agents at Nubank, a company with 100M+ users. Our approach integrates several key components: (1) structured context engineering tailored to customer support agents, (2) systematic human-in-the-loop prompt iteration, (3) rigorous LLM judge evaluation with measured inter-rater agreement and GEPA optimization for consistency, and (4) ideation-to-production validation. A central insight is that evaluation-pipeline quality directly determines iteration velocity. We present results from five production deployments spanning distinct domains: card delivery, debt management, credit-limit support, card management, and product explanation. These deployments deliver consistent customer-satisfaction gains while substantially accelerating iteration. In our card-delivery deployment, large-scale A/B testing yields a 37 percentage-point improvement in AI transactional Net Promoter Score and a 29 percentage-point gain in self-service rate over prior agent variants, alongside a strong correlation between offline simulation metrics and online outcomes, demonstrating that eval-driven development reliably predicts production impact. On most use cases, AI satisfaction reaches within a few percentage points of expert human agents.

03.
arXiv (CS.CV) 2026-06-16

KeepLoRA++: Continual Learning with Layer-Scaled Residual Gradient Adaptation

Continual learning for pre-trained vision-language models requires balancing three competing objectives: retaining pre-trained knowledge, preserving knowledge from a sequence of learned tasks, and maintaining the plasticity to acquire new knowledge. This paper presents KeepLoRA++, balancing these objectives through a unified dual-dimensional knowledge retention mechanism. We analyze knowledge distribution of Transformer architecture from both inter-layer and intra-layer perspectives. The inter-layer perspective examines how retention is distributed across layers, while the intra-layer perspective focuses on the parameter space within each layer. Our analysis reveals a structural property: general transferable knowledge is mainly encoded in the shallow layers and the principal subspace of the parameters, while task-specific adaptations are localized in the deep layers and the residual subspace. Motivated by this insight, KeepLoRA++ introduces a layer-scaled residual gradient adaptation method. New tasks are learned by restricting LoRA parameter updates to the residual subspace, combined with a shallow-to-deep layer scaling, to prevent interference with previously acquired capabilities. Specifically, the gradient of a new task is projected onto a subspace orthogonal to both the principal subspace of the pre-trained model and the dominant directions of previous task features, while simultaneously assigning smaller update magnitudes to shallow layers and larger ones to deeper layers. Our theoretical analysis and empirical evaluations confirm that KeepLoRA++ successfully balances these three competing objectives, consistently outperforming representative baselines across image classification, visual question answering, and video understanding tasks.

04.
arXiv (CS.CV) 2026-06-16

Bridging Geographic Bias in Urban Streetscape Inference via Lifelong Learning with Visual-Semantic Pivoting

作者:

Visual perception of urban streetscapes underpins evidence-based decisions in landscape planning, public health, and place-making. Yet models trained on a few well-photographed metropolises systematically misjudge underrepresented districts, propagating geographic bias into downstream policy. We address this gap with HVSP-LL, a lifelong learning framework that couples a stratified visual-semantic pivoting module with an equity-aware rehearsal mechanism. The pivoting module organises landscape concepts along a three-tier ontology (macro structure, meso composition, micro element) and aligns image features to learnable semantic anchors at each tier, providing transferable representations that resist distributional drift. The lifelong adaptation component sequentially absorbs new urban regions while constraining inter-region perception gaps through a worst-region sample-reweighting objective and a structurally-aware exemplar buffer. We evaluate HVSP-LL on a panoramic streetscape benchmark assembled from twelve cities across four continents and seven perceptual dimensions. The framework attains 0.834 Spearman correlation on the held-out city sequence, an absolute 6.1 point improvement over the strongest continual baseline, and shrinks the inter-city perception gap to 0.094 – a 38% reduction relative to the strongest continual baseline (0.151) and a 57% reduction relative to a representative regularisation baseline (0.218). Ablations confirm that each tier of the pivoting hierarchy contributes monotonically, and the equity-aware rehearsal converts mean backward transfer from -0.038 (without retention) to +0.013, eliminating catastrophic forgetting on the held-out sequence. Our results indicate that hierarchical anchoring is a practical pathway toward geographically equitable streetscape inference at city scale.

05.
arXiv (CS.AI) 2026-06-18

Leveraging Energy Features for Surface Classification with Deep Learning: A Comparative Analysis Across Three Independent Datasets

arXiv:2606.18698v1 Announce Type: cross Abstract: The energy-based method remains a comparatively underexamined approach for surface classification in mobile robotics, despite promising results in constrained environments. This study evaluated the viability of using energy-derived features as either a standalone classification modality or as supplementary input to inertial data. A comprehensive evaluation was conducted across three publicly available datasets, comparing the performance of modern deep learning architectures including recurrent neural networks, convolutional neural networks, encoder-only transformers, and Mamba state-space models, under automated hyperparameter tuning and input sequence length optimization. The models achieved higher accuracy than previously reported values on all evaluated datasets, with the convolutional neural network yielding the highest overall performance. When relying exclusively on energy-based features, the models attained classification accuracies in the range of 85-90%, approximately 5-10% lower than those achieved when combined with inertial features (96-99%). Augmenting inertial data with energy features resulted in a consistent mean accuracy improvement of 1-2%. These findings indicate that classifiers relying solely on energy features offer sufficient accuracy for standalone deployment, while also providing a consistent gain when used in combination with other sensing modalities.

06.
arXiv (CS.CL) 2026-06-12

MoReBench: Evaluating Procedural and Pluralistic Moral Reasoning in Language Models, More than Outcomes

As AI systems progress, we rely more on them to make decisions with us and for us. To ensure that such decisions are aligned with human values, it is imperative for us to understand not only what decisions they make but also how they come to those decisions. Reasoning language models, which provide both final responses and (partially transparent) intermediate thinking traces, present a timely opportunity to study AI procedural reasoning. Unlike math and code problems which often have objectively correct answers, moral dilemmas are an excellent testbed for process-focused evaluation because they allow for multiple defensible conclusions. To do so, we present MoReBench: 1,000 moral scenarios, each paired with a set of rubric criteria that experts consider essential to include (or avoid) when reasoning about the scenarios. MoReBench contains over 23 thousand criteria including identifying moral considerations, weighing trade-offs, and giving actionable recommendations to cover cases on AI advising humans moral decisions as well as making moral decisions autonomously. Separately, we curate MoReBench-Theory: 150 examples to test whether AI can reason under five major frameworks in normative ethics. Our results show that scaling laws and existing benchmarks on math, code, and scientific reasoning tasks fail to predict models' abilities to perform moral reasoning. Models also show partiality towards specific moral frameworks (e.g., Benthamite Act Utilitarianism and Kantian Deontology), which might be side effects of popular training paradigms. Together, these benchmarks advance process-focused reasoning evaluation towards safer and more transparent AI.

07.
arXiv (quant-ph) 2026-06-17

Learning Arbitrary Lindbladians with Quantum Error Correction

arXiv:2606.18188v1 Announce Type: new Abstract: We study ansatz-free Lindbladian learning, the problem of reconstructing the generator of an open quantum system without prior knowledge of its Hamiltonian or dissipator structures. This problem exhibits two distinct information-theoretic precision limits: Hamiltonian components unmasked by dissipation are Heisenberg-limited, while the remaining Lindbladian components are subject to the quadratically worse standard quantum limit. Existing approaches that attain these optimal scalings strongly rely on pre-specified structure of interaction and noise, leaving the ansatz-free setting an open problem. In this work, we present the first standard-quantum-limited algorithm for learning arbitrary sparse Lindbladians. Under an additional physically motivated regularity condition, our framework also learns the Hamiltonian component disjoint from the dissipator at the Heisenberg limit, without prior knowledge of either the Hamiltonian or dissipator supports. Our main technical ingredient is a recursive random stabilizer-code construction that suppresses the strongest Lindbladian terms while preserving sensitivity to weaker unknown ones. These results establish a scalable framework for characterizing unknown open quantum systems, with quantum error correction serving as a key learning primitive.

08.
arXiv (CS.CL) 2026-06-19

JAMER: Project-Level Code Framework Dataset and Benchmark on Professional Game Engines

Current AI-driven game development has made substantial progress in asset generation, gameplay design, and web-based game coding, yet project-level code engineering on professional game engines remains largely unexplored due to the absence of large-scale datasets and deterministic evaluation methods. We present JamSet and JamBench, the first project-level game code framework dataset and benchmark built on a professional game engine. Our key insight is that Game Jam competitions, community events where developers build complete games under tight time constraints, yield thousands of open-source projects suitable for this purpose. Building on the Godot engine's text-based format and headless execution mode, we design a deterministic verification pipeline from file integrity to runtime behavior collection, distilling 8,133 verified projects from over 240,000 repositories. Of these, 300 manually verified projects form JamBench; the rest constitute JamSet. JamBench defines theme-driven generation and code completion tasks, evaluated through a pipeline combining compilation pass rates, Structural Completeness Score (SCS), and Behavioral Alignment Score (BAS). Evaluation of 9 frontier models reveals a capability cliff as project scale increases, with runtime pass rates dropping from 80.4% on small projects to 5.7% on large ones (Task2a). Code Agents improve compilation rates yet yield no gains in runtime behavioral quality, indicating that the bottleneck lies in architectural design rather than syntactic correctness. Experiments validate JamSet as effective training data. All data and code are publicly available.

09.
arXiv (CS.AI) 2026-06-16

Emergent Strategic Reasoning Risks in AI: A Taxonomy-Driven Evaluation Framework

arXiv:2604.22119v2 Announce Type: replace Abstract: As reasoning capacity and deployment scope grow in tandem, large language models (LLMs) gain the capacity to engage in behaviors that serve their own objectives, a class of risks we term Emergent Strategic Reasoning Risks (ESRRs). These include, but are not limited to, deception (intentionally misleading users or evaluators), evaluation gaming (strategically manipulating performance during safety testing), and reward hacking (exploiting misspecified objectives). Systematically understanding and benchmarking these risks remains an open challenge. To address this gap, we introduce ESRRSim, a taxonomy-driven agentic framework for automated behavioral risk evaluation. We construct an extensible risk taxonomy of 7 categories, which is decomposed into 20 subcategories. ESRRSim generates evaluation scenarios designed to elicit faithful reasoning, paired with dual rubrics assessing both model responses and reasoning traces, in a judge-agnostic and scalable architecture. Evaluation across 11 reasoning LLMs reveals substantial variation in risk profiles (detection rates ranging 14.45%-72.72%), with dramatic generational improvements suggesting models may increasingly recognize and adapt to evaluation contexts.

10.
arXiv (CS.CV) 2026-06-12

Surflo: Consistent 3D Surface Flow Model with Global State

Geometry is invariant to viewpoint, which makes any collection of images a redundant encoding of a single 3D state. Existing feed-forward reconstruction models fail to exploit this: per-view methods emit overlapping, unaligned pointmaps that grow linearly with input count, while global-latent methods commit to a fixed, low-resolution output. We introduce Surflo, which compresses a variable number of unposed RGB views into K latent tokens-one global state-and decodes oriented 3D surface points by independently transporting them from noise onto the surface via flow matching. This frees the output from any fixed grid or token budget: the same latent yields from a few thousand to a million points in a single forward pass. To suppress the local inconsistencies inherent to independent per-point decoding, an inference-time guidance term correlates nearby points by injecting a photometric gradient during ODE integration. Surflo matches or surpasses feed-forward baselines on surface metrics, runs an order of magnitude faster than optimization-based methods that require hundreds of views, and is the only feed-forward approach to combine a global latent with arbitrary-resolution decoding.

11.
arXiv (CS.AI) 2026-06-11

Mind the Perspective: Let's Reason Recursively for Theory of Mind

arXiv:2606.11724v1 Announce Type: new Abstract: Theory of Mind (ToM) reasoning requires inferring agents' beliefs from partial and asymmetric observations, which remains an open challenge for LLMs. Existing prompting-based approaches improve ToM reasoning through observable-event filtering or temporal belief chains, without explicitly modeling nested beliefs. We introduce RecToM, an inference-time framework for ToM reasoning that models nested beliefs via recursive perspective construction. RecToM constructs each character perspective from the preceding character perspective along the character chain specified by the question, reducing higher-order belief questions to actual-world questions within the final constructed perspective. We further provide a KD45 analysis showing that RecToM's perspective construction induces a well-formed belief modality beyond simple event filtering. Experiments on ToM benchmarks, including Hi-ToM, Big-ToM, and FanToM, across multiple LLM backbones show that RecToM consistently outperforms recent advanced approaches, achieving state-of-the-art performance. Notably, RecToM reaches 100\% accuracy on Hi-ToM with GPT-5.4 and Qwen3.5, a benchmark requiring higher-order ToM reasoning.

12.
arXiv (CS.CL) 2026-06-17

Non-negative Elastic Net Decoding for Information Retrieval

Dense retrieval has become the dominant paradigm in information retrieval, in which each document is scored against a query by the inner product of their vector embeddings, and the top-$k$ documents by score are retrieved for this query. However, since each document's score depends solely on the embedding of the query and itself, the retrieval process is oblivious to the content of the entire corpus. Therefore, dense retrieval cannot avoid selecting semantically similar documents from the corpus, which may result in a non-diverse, redundant set of retrieved documents. To this end, we approach retrieval as a joint decoding problem, in which documents are selected as a set with regard to the context of the rest of the corpus. To achieve this, we propose Non-Negative elastic Net (NNN) decoding, which selects documents whose embeddings jointly reconstruct the query embedding as a sparse non-negative linear combination. Our main theoretical result establishes a strict separation between dense retrieval and NNN decoding. For any corpus, every query correctly handled by dense retrieval is also handled by NNN decoding, while on corpora containing correlated documents, NNN decoding additionally handles queries that dense retrieval cannot. Experimental results indicate that applying NNN decoding to frozen embeddings trained for inner-product scoring yields consistent improvements across several benchmarks. Moreover, we introduce an end-to-end training procedure which optimizes the embeddings for NNN decoding, producing significant performance gains surpassing in all metrics and benchmarks compared to dense retrieval. Our work establishes a new paradigm for leveraging dense embeddings in information retrieval, beyond the standard practice of inner-product scoring.

13.
arXiv (CS.AI) 2026-06-18

Efficient Zeroth-Order Federated Finetuning of Language Models on Resource-Constrained Devices

arXiv:2502.10239v3 Announce Type: replace-cross Abstract: Federated Learning (FL) is a promising paradigm for finetuning Large Language Models (LLMs) across distributed data sources while preserving data privacy. However, finetuning such large models is challenging on edge devices due to its high resource demand. Zeroth-order Optimization (ZO) estimates gradients through finite-difference approximations, which rely on function evaluations under random perturbations of the model parameters. Consequently, ZO with task alignment provides a potential solution, allowing finetuning using only forward passes with inference-level memory requirements and low communication overhead, but it suffers from slow convergence and higher computational demand. In this paper, we propose a new ZO-based method that applies a more efficient technique to reduce the computational demand associated with using a large number of perturbations while preserving their convergence benefits. This is achieved by splitting the model into consecutive blocks and allocating a higher number of perturbations to the second block, enabling efficient reuse of intermediate activations to update the full network with fewer forward evaluations. Our evaluation on RoBERTa-large, OPT1.3B, LLaMa-3-3.2B models shows up to $3\times$ reduction in computation compared to the other ZO-based techniques, while retaining the memory and communication benefits over first-order federated learning techniques.

14.
arXiv (CS.AI) 2026-06-12

Bag of Dims: Training-Free Mechanistic Interpretability via Dimension-Level Sign Patterns

arXiv:2606.12629v1 Announce Type: cross Abstract: We show that the standard basis of transformer hidden states already provides a training-free, architecture-general feature basis. Individual dimensions encode semantic content via their signs and confidence via their magnitudes, functioning as independent binary registers. We validate this Bag of Dims framework across three model families (Qwen 3.5-4B, Gemma 3-4B, Mistral 7B) through four progressive experiments. Sign patterns alone carry predictive content: replacing all magnitudes with unity achieves 72-93% top-5 next-token accuracy through the LM head, and pure Hamming scoring without any decoder reaches 80-90% top-4096. These sign patterns organize into semantic features: using a single-token type cache (one forward pass per vocabulary token, no context), we discover 175 categories via per-dimension sign consistency (mean AUC 0.80) from 50 anchors with zero training. A trained probe adds only +0.018 AUC and converges to axis-aligned weights, confirming negligible cross-dimension structure. This structure extends to attention: all 175 categories remain discoverable in K and V projections. On the write side, static FFN weight inspection links 20% of features to individual writer neurons (>0.70 agreement; random controls: 0%), with top-200 neuron coalitions achieving >0.70 agreement on 99.9% of prototypes via majority vote. Fully unsupervised discovery (random seeds, no labels) scales to 1500 features at 100% yield and 99% sparsity across all three models, with pairwise MI of 0.0014 bits confirming low inter-dimension coupling. These results establish that the standard basis already suffices for feature reading throughout the transformer compute pathway, requiring no training, no optimization, and no GPU-days beyond a single forward pass per vocabulary token.

15.
arXiv (CS.AI) 2026-06-11

Noise-Guided Transport for Imitation Learning

arXiv:2509.26294v2 Announce Type: replace-cross Abstract: We consider imitation learning in the low-data regime, where only a limited number of expert demonstrations are available. In this setting, methods that rely on large-scale pretraining or high-capacity architectures can be difficult to apply, and efficiency with respect to demonstration data becomes critical. We introduce Noise-Guided Transport (NGT), a lightweight off-policy method that casts imitation as an optimal transport problem solved via adversarial training. NGT requires no pretraining or specialized architectures, incorporates uncertainty estimation by design, and is easy to implement and tune. Despite its simplicity, NGT achieves strong performance on challenging continuous control tasks, including high-dimensional Humanoid tasks, under ultra-low data regimes with as few as 20 transitions.

16.
arXiv (CS.CV) 2026-06-19

iSAGE: A Human-in-the-Loop Framework for Remote Sensing Semantic Segmentation via Sparse Point Supervision

Semantic segmentation in remote sensing requires costly pixel-level annotations, and nearly every problem demands a new dataset since models rarely transfer across sensors, platforms, or geographies. Existing human-in-the-loop frameworks expand sparse clicks into dense supervision via auxiliary machinery (pseudo-labels, propagation, CRFs, foundation-model prompts, auxiliary heads), all operating on the model's predictive distribution. A confidently wrong pixel is indistinguishable from a confidently correct one in that distribution by construction, so no rule reading it can separate the two; the distinguishing signal is external to the model. This paper hypothesizes that expert clicks targeting confident model errors, not arbitrary pixels, suffice to match dense supervision, with no expansion machinery. iSAGE (Iterative Sparse Annotation Guided by Expert) realizes this hypothesis on an integrated open-source platform, where an error-weighted loss amplifies the gradient at each click and the annotation record itself is the dataset, extensible, correctable, and auditable. Experiments use a minimum-effort regime: at most one labeled pixel per class per frame. On BsB Aerial, iSAGE recovers 97.2% of dense supervision (74.79% mIoU on 0.040% of pixels) with contrasting class dynamics: amorphous classes (permeable areas) saturate from the seed, while small classes (cars) require late-iteration effort. On ISPRS Vaihingen (external benchmark), iSAGE reaches 76.78% mIoU with 0.011% of pixels, matching the dense baseline (76.65%) and exceeding all published methods. Under the same pipeline, four output-reading mechanisms (oracle entropy across budgets 1–100x, pseudo-labels across thresholds 0.90–0.99, CRF-based propagation, uniform random) plateau 7.4 to 14.5 pp below iSAGE. Across 31 surveyed methods, iSAGE is the only iterative human-in-the-loop framework operating without auxiliary machinery.

17.
arXiv (CS.CL) 2026-06-12

One Token to Fool LLM-as-a-Judge

Large language models (LLMs) are increasingly trusted as automated judges, assisting evaluation and providing reward signals for training other models, particularly in reference-based settings like Reinforcement Learning with Verifiable Rewards (RLVR). However, we uncover a critical vulnerability even in this reference-based paradigm: generative reward models are systematically susceptible to reward hacking. We find that superficial inputs, which we term ''master keys'' such as non-word symbols (e.g., '':'' or ''.'') or generic reasoning openers (e.g., ''Thought process:'' or ''Let's solve this problem step by step.''), can consistently elicit false positive rewards without any substantive reasoning. Our systematic evaluation demonstrates this is a widespread failure affecting a diverse range of models, including leading proprietary systems such as GPT-o1 and Claude-4. These results challenge the assumed robustness of LLM judges and pose a significant threat to their reliability. To address this, we propose a simple yet effective data augmentation strategy using truncated model outputs as adversarial negative examples. The resulting Master Reward Models (Master-RMs) demonstrate state-of-the-art robustness against these ''master key'' attacks while maintaining high performance in standard evaluation settings. We supplement these findings with a comprehensive analysis of the vulnerability across model scales, prompt variations, and common inference-time strategies, offering insights to guide future research on robust LLM evaluation. We release our robust, general-domain reward models and the synthetic training data at https://huggingface.co/sarosavo/Master-RM and https://huggingface.co/datasets/sarosavo/Master-RM.

18.
arXiv (quant-ph) 2026-06-17

$\mathcal{PT}$-Symmetric Spin–Boson Model with a Continuous Bosonic Spectrum: Exceptional Points and Dynamics

arXiv:2512.20277v2 Announce Type: replace Abstract: This work studies a $\mathcal{PT}$-symmetric non-Hermitian spin–boson model, consisting of a non-Hermitian two-level system coupled to a continuous bosonic bath. The static properties of the system are analyzed through a projection method derived from the displacement operator. We find that only a single exceptional point (EP) emerges, in contrast to non-Hermitian spin–boson models with finite modes, which typically exhibit multiple EPs. Notably, only a single real eigenvalue is found before the EP, which differs markedly from typical non-Hermitian systems where a pair of real eigenvalues precedes the EP. The time evolution of observables is further investigated via the Dirac–Frenkel time-dependent variational principle. Compared to its Hermitian counterpart, the non-Hermitian model exhibits distinct dynamical signatures, most notably the emergence of oscillations with periodic amplified amplitude. In the $\mathcal{PT}$-unbroken phase, the system exhibits sustained oscillatory dynamics with suppressed decoherence, whereas in the $\mathcal{PT}$-broken phase, additional dissipative channels accelerate decoherence and drive rapid convergence toward a stable steady state. These results shed light on how $\mathcal{PT}$ symmetry protects coherent light–matter interactions in non-Hermitian quantum systems.

19.
arXiv (CS.CL) 2026-06-12

Detecting Functional Memorization in Code Language Models

Large language models (LLMs) are increasingly used to generate code at scale. Meanwhile, prior work has investigated whether training data may be recoverable from model outputs, by auditing the textual overlap between training examples and model generations. Code, however, can be functionally equivalent while textually dissimilar. In this work, we study functional memorization: extraction of functional logic beyond what verbatim metrics detect. We construct a counterfactual setup for Olmo-3-32B, comparing a midtrained model (exposed to target code) against a pretrained reference (not exposed). We prompt both models with Python function signatures and measure both textual and functional similarity (i.e., LLM-as-a-judge, execution-based). Our results show clear evidence of functional memorization, highlighting the need for auditing metrics that go beyond textual overlap.

20.
arXiv (CS.AI) 2026-06-19

Hierarchical Control in Multi-Agent Games: LLM-based Planning and RL Execution

arXiv:2606.20014v1 Announce Type: cross Abstract: Reinforcement learning (RL) has achieved strong performance in sequential decision-making, yet scaling to complex multi-agent environments remains challenging due to sparse rewards, large state-action spaces, and the difficulty of learning coordinated strategies. We propose a hierarchical architecture where a pretrained large language model (LLM) acts as a centralized strategic controller that selects among specialized RL skill policies for a team of agents, while RL policies handle reactive low-level execution. We evaluate this hybrid system in a competitive 2v2 King of the Hill environment against behavior tree (BT) and ``Flat'' RL (end-to-end training without skill decomposition) baselines. The LLM+RL system achieves task performance statistically equivalent to hand-crafted BT (46.4\% vs 51.5\% win rate, $p=0.103$) while both significantly outperform Flat RL trained without skill decomposition. A user study ($n=15$) reveals that 60\% of participants perceive LLM+RL agents as the most human-like ($p=0.027$), citing behavioral adaptability and tactical variability. These results demonstrate that pretrained LLM reasoning can effectively orchestrate pretrained RL skills, achieving competitive multi-agent coordination and superior perceived believability without manual rule engineering.

21.
arXiv (CS.LG) 2026-06-16

Multi-Agent Framework for Audit Risk Assessment with Explicit Uncertainty and Evidence Conflict Modeling

arXiv:2606.15640v1 Announce Type: new Abstract: Audit risk assessment increasingly benefits from combining heterogeneous evidence sources, yet existing approaches typically produce point predictions without quantifying how well different evidence streams agree. We propose UMAR (Uncertainty-Aware Multi-Agent Risk Assessment), a framework that employs three specialized agents: an MD&A Text Agent, a Financial Ratio Agent, and a CAM Agent, each producing independent risk scores with calibrated uncertainty estimates. An Uncertainty Aggregator based on Dempster-Shafer evidence theory fuses these scores while explicitly measuring inter-agent conflict. We evaluate UMAR on a U.S. dataset of 3,200 firm-year observations from SEC 10-K filings (2019-2023), with financial restatement as the target label. Experimental results show that UMAR achieves an AUROC of 0.782 and a PR-AUC of 0.341, outperforming logistic regression, XGBoost, FinBERT, and single-agent and dual-agent LLM baselines. UMAR attains the lowest expected calibration error (ECE = 0.052) among all methods and identifies evidence-conflict patterns that correlate with actual restatement risk, offering auditors potentially actionable and interpretable risk signals.

22.
arXiv (CS.AI) 2026-06-16

LLM-WikiRace Benchmark: How Far Can LLMs Plan over Real-World Knowledge Graphs?

arXiv:2602.16902v5 Announce Type: replace Abstract: We introduce LLM-Wikirace, a benchmark for evaluating planning, reasoning, and world knowledge in large language models (LLMs). In LLM-Wikirace, models must efficiently navigate Wikipedia hyperlinks step by step to reach a target page from a given source, requiring look-ahead planning and the ability to reason about how concepts are connected in the real world. We evaluate a broad set of open- and closed-source models, including Gemini-3, GPT-5, and Claude Opus 4.5, which achieve the strongest results on the easy level of the task and demonstrate superhuman performance. Despite this, performance drops sharply on hard difficulty: the best-performing model, Gemini-3, succeeds in only 23\% of hard games, highlighting substantial remaining challenges for frontier models. Our analysis shows that world knowledge is a necessary ingredient for success, but only up to a point, beyond this threshold, planning and long-horizon reasoning capabilities become the dominant factors. Trajectory-level analysis further reveals that even the strongest models struggle to replan after failure, frequently entering loops rather than recovering. LLM-Wikirace is a simple benchmark that reveals clear limitations in current reasoning systems, offering an open arena where planning-capable LLMs still have much to prove. Our code and leaderboard available at https:/llmwikirace.github.io.

23.
arXiv (CS.LG) 2026-06-17

Learning to Refine Hidden States for Reliable LLM Reasoning

arXiv:2606.17524v1 Announce Type: new Abstract: Large language models show strong reasoning ability, but their internal reasoning process can remain unstable in complex multi-step settings, where early hidden-state errors may propagate to incorrect predictions. We propose ReLAR, a reinforcement-guided latent refinement framework that iteratively updates hidden representations before decoding. ReLAR maintains a compact latent reasoning state and uses learned depth and action controllers to adaptively determine both the number and direction of refinement steps. The controllers are trained with a policy gradient objective based on step-wise likelihood improvement, enabling efficient input-dependent reasoning without explicit chain-of-thought generation. Experiments on medical, mathematical, multi-hop reasoning, and open-ended generation benchmarks show that ReLAR improves accuracy, generation quality, and reasoning stability with substantially lower inference overhead than explicit reasoning baselines.

24.
medRxiv (Medicine) 2026-06-15

Anti-Platelet Factor 4 Antibody Clonal Heterogeneity and MGUS Status in HIT

Background Monoclonal gammopathy of thrombotic significance (MGTS) is a recently described chronic prothrombotic condition characterized by monoclonal anti-PF4 antibodies that are detected above the polyclonal antibody background in patient sera (i.e. present as monoclonal gammopathy of undetermined significance, MGUS). Due to conflicting data in the published literature on antibody clonality in heparin-induced thrombocytopenia (HIT), we evaluated clonality and abundance of anti-PF4 antibodies in HIT, including investigating whether an MGUS, if present in HIT, represents the causative anti-PF4 antibody. Methods Blood samples from 15 patients with HIT were subject to Platelet Factor 4-dependent antigen-based and functional tests. The unmanipulated serum antibody repertoire and isolated anti-PF4 antibodies were subjected to mass spectrometric evaluation. Results Two of the 15 HIT patients had an IgG MGUS. Notably, anti-PF4 antibodies were not synonymous with the MGUS antibody in either of the two patients. Eight of the 15 patients demonstrated monoclonal anti-PF4 antibodies, however, none of the anti-PF4 antibodies were detectable as an MGUS upon evaluation of the entire serum antibody repertoire, reflecting their low abundance. In the seven patients with multiple anti-PF4 antibodies, non-monoclonality was confirmed by analysis of deglycosylated antibody heavy chains. Conclusions Anti-PF4 HIT antibodies are monoclonal in approximately 50% of HIT patients, however, antibody abundance is low such that they are not detectable over the polyclonal IgG background (i.e. are MGUS-negative), differentiating HIT from MGTS. This observation helps explain the transient nature of HIT relative to the persistent prothrombotic state seen in MGTS.

25.
medRxiv (Medicine) 2026-06-13

Projected population level impact and cost-effectiveness of clinic and community-based tuberculosis screening approaches

The South Africa National Department of Health have set ambitious targets to scale up TB testing, focusing primarily on clinic attendees. In the context of declining funding for TB care and prevention, the most cost-effective approaches for targeting testing should be identified. We developed a mathematical model of TB in South Africa, explicitly incorporating clinic attendance by sex and HIV/ART status. We simulated six screening approaches over 2026-2035 (individually and in combination): three clinic-based (symptom screening, intensified targeted universal TB testing [TUTT, symptom-agnostic sputum testing of clinic attendees in key risk groups], and intensified TUTT allowing saliva samples) and three targeted community-based (community radiographic screening, symptom screening, and universal Xpert Ultra testing), each implemented at a range of coverage levels. Model outputs were combined with a mechanistic cost function to estimate potential impact and cost-effectiveness from a societal perspective. The most cost-effective standalone approach was community radiographic screening at 10% annual population coverage, with an incremental cost-effectiveness ratio (ICER) of $421 per disability-adjusted life year (DALY) averted. 10/11 scenarios along the expansion path included community radiographic screening at progressively higher coverage, combined with a clinic-based approach. Combining complementary approaches to reach both groups at increased risk of TB (e.g. clinic-based screening) and groups with lower screening coverage (e.g. community-based screening) may increase cost-effectiveness of TB screening, compared to standalone approaches. When designing TB screening strategies, both population risk and existing screening coverage should be considered.