探索全球前沿学术脉络

01.

arXiv (CS.CL) 2026-06-19 DOI: arXiv:2603.16941

The Voice Behind the Words: Quantifying Intersectional Bias in SpeechLLMs

作者:

Shree Harsha Bokkahalli Satish↗Christoph Minixhofer↗Maria Teleki↗James Caverlee↗Ond\v{r}ej Klejch↗Peter Bell↗Gustav Eje Henter↗\'Eva Sz\'ekely↗

Speech Large Language Models (SpeechLLMs) process spoken input directly, retaining cues such as accent and perceived gender that were previously removed in cascaded pipelines. This introduces speaker identity dependent variation in responses. We present a large-scale intersectional evaluation of accent and gender bias in three SpeechLLMs using 2,880 controlled interactions across six English accents and two gender presentations, keeping linguistic content constant through voice cloning. Using pointwise LLM-judge ratings, pairwise comparisons, and Best-Worst Scaling with human validation, we detect recurring directional disparities. Eastern European-accented speech receives lower helpfulness scores, particularly for female-presenting voices. Responses remain polite but differ in helpfulness. While LLM judges capture the directional trend of these biases, human evaluators exhibit significantly higher sensitivity, showing stronger accent-level contrasts.

阅读与讨论 → 访问原文 →

02.

arXiv (CS.CL) 2026-06-16 DOI: arXiv:2606.15517

SHARD: Safe and Helpful Alignment via Self-Reframing Distillation

作者:

Viswonathan Manoranjan↗Amogh Gupta↗Anvesh Rao Vijjini↗Thomas Hofweber↗Snigdha Chaturvedi↗

Large language models often struggle with sensitive prompts. They may refuse outright, provide generic safety boilerplate, or fail to address the user's legitimate informational needs that can be answered safely. We introduce SHARD, a self-reframing distillation method to improve safe-helpfulness. It first rewrites sensitive prompts to surface benign intent using philosophical guidelines, then reframes its original responses into safe, more helpful ones, and finally fine-tunes the model on its self-reframed responses. Across DNA and the English subset of LINGUASAFE, SHARD improves helpfulness for most model families while preserving safety. It also remains competitive with distillation from a larger teacher model, suggesting that models can internalize safe and helpful behavior elicited from their own. Warning: This paper contains content that may be offensive or harmful.

阅读与讨论 → 访问原文 →

03.

arXiv (CS.CL) 2026-06-19 DOI: arXiv:2606.20482

Your Mouse and Eyes Secretly Leak Your Preference: LLM Alignment using Implicit Feedback from Users

作者:

Haw-Shiuan Chang↗Jeffrey Gomez↗Mehul Patwari↗Aryan Sajith↗Hamed Zamani↗

To align a Large Language Model (LLM), most existing methods collect explicit human feedback and train a reward model to predict the human preference based on the response text. These existing methods have two key limitations. First, the users rarely provide explicit feedback for LLM responses, which makes the high-quality preference annotation expensive to collect. Second, the methods do not leverage implicit human feedback, which has proven vital to the economic moats of Internet giants. To quantify the value of implicit feedback, we build a new dataset called IFLLM, which collects 1336 multi-turn questions from the 59 Mechanical Turk workers, their mouse trajectories, and eye gazing points to the LLMs' responses from their webcams. IFLLM shows that the users have very diverse types of gazing behavior and mouse trajectories. Our reward model based on the implicit user feedback boosts the accuracy of the text-based reward model from 55% to 64% and nearly triples the relative response quality improvements after applying the DPO to eight LLMs, demonstrating the value of implicit feedback in the wild. Our data collection website, dataset, and codes can be found at https://github.com/themehulpatwari/llm-implicit-feedback/.

阅读与讨论 → 访问原文 →

04.

arXiv (CS.CV) 2026-06-18 DOI: arXiv:2603.13941

Bidirectional Cross-Attention Fusion of High-Resolution RGB and Low-Resolution Hyperspectral Inputs for Multimodal Semantic Segmentation

作者:

Jonas V. Funk↗Lukas Roming↗Andreas Michel↗Paul B\"acker↗Georg Maier↗Thomas L\"angle↗Markus Klute↗

Multimodal semantic segmentation with heterogeneous sensors must reconcile complementary information across modalities that differ in spatial resolution and channel dimensionality. In particular, high-resolution RGB imaging provides detailed spatial structure but often fails to distinguish visually similar materials, whereas hyperspectral imaging (HSI) provides discriminative spectral signatures but at lower spatial resolution. We present Bidirectional Cross-Attention Fusion (BCAF), which aligns high-resolution RGB with low-resolution HSI at their native grids via localized, bidirectional cross-attention, avoiding pre-upsampling or early spectral collapse. BCAF uses two independent backbones: a standard Swin Transformer for RGB and an HSI-adapted Swin backbone that preserves spectral structure through 3D tokenization with spectral self-attention. Although our evaluation targets RGB-HSI fusion, BCAF is modality-agnostic and applies to co-registered RGB with lower-resolution, high-channel auxiliary sensors. On the benchmark SpectralWaste dataset, BCAF delivers strong performance, achieving 75.4% at 55 images/s. We further evaluate a novel industrial dataset: K3I-Cycling (first RGB subset already released on Fordatis). On this dataset, BCAF reaches 62.3% mIoU for material segmentation (paper, metal, plastic, etc.) and 66.2% mIoU for plastic-type segmentation (PET, PP, HDPE, LDPE, PS, etc.). These results show that preserving native-grid spatial detail and spectral structure improves multimodal segmentation under real-time constraints. Code and model checkpoints are publicly available at https://github.com/jonasvilhofunk/BCAF_2026.

阅读与讨论 → 访问原文 →

05.

arXiv (CS.CL) 2026-06-18 DOI: arXiv:2606.18587

Dual Dimensionality for Local and Global Attention

作者:

Zhiyuan Wang↗Xuan Luo↗Sirui Zeng↗Xifeng Yan↗

Decoder-only Transformers compute attention over the KV cache of preceding tokens. Keys (and Values) are typically represented with the same dimensionality, regardless of its distance from the prediction target. In natural language, however, the next word is most strongly influenced by the immediately preceding tokens. We hypothesize that local and distant tokens impose asymmetric demands on representational capacity: local tokens are more critical for predicting immediate outputs and thus require richer representations, whereas distant tokens primarily serve as long-range memory, for which lower-dimensional representations may suffice. We formalize this idea as Distance-Adaptive Representation (DAR), implemented in a controlled setting that preserves full-dimensional representations within a local context window while assigning reduced-dimensional representations (e.g. 1/4 of the original dimensionality) to tokens beyond that window. Across multiple pretraining scales (70M to 410M parameters), as well as continued supervised fine-tuning on a 1B-scale model, this approach closely matches the performance of full-dimensional baselines. In contrast, uniformly reducing dimensionality across all token positions leads to worse performance. These results challenge the common assumption that key and value dimensionality should be uniform across token positions. Our findings suggest a new direction for designing attention architectures that adaptively allocate representational capacity across sequences, enabling further reductions in KV cache during inference.

阅读与讨论 → 访问原文 →

06.

arXiv (CS.AI) 2026-06-17 DOI: arXiv:2606.03089

Constitutional On-Policy Safe Distillation

作者:

Ming Wen↗Yuxuan Liu↗Kun Yang↗Yunhao Feng↗Zhuoer Xu↗Yuhao Sun↗Shiwen Cui↗Xiang Zheng↗Guoyu Wang↗Xingjun Ma↗Yu-Gang Jiang↗

arXiv:2606.03089v2 Announce Type: replace-cross Abstract: On-policy self-distillation (OPSD) has emerged as an efficient post-training paradigm by using a teacher conditioned on privileged information to provide dense token-level supervision. Prior work has shown that OPSD can collapse in verifiable reasoning tasks, but safety alignment differs in that it is guided by high-level constitutions rather than explicit target answers, making it a natural setting to revisit dense distillation. However, our pilot study show that safety OPSD still suffers from severe collapse: constitutional conditioning contracts the teacher distribution toward short and overly conservative responses, and Reverse KL further amplifies this contraction into reduced expressiveness. We formalize this effect as geometric leakage under safety boundaries in a non-orthogonal semantic space, where safety pressure transfers into the expressiveness dimension. Based on this analysis, we propose Constitutional On-Policy Safe Distillation (COPSD), which first calibrates the teacher through a Cross-SFT cold-start and then performs constitution-conditioned on-policy distillation. Experiments on 12 benchmarks show that COPSD achieves a consistently stronger safety–helpfulness trade-off than baselines while substantially reducing the safety tax on general reasoning ability.

阅读与讨论 → 访问原文 →

07.

arXiv (CS.AI) 2026-06-16 DOI: arXiv:2606.16133

InvDesMobility: a reliability-gated first-principles feedback framework for closed-loop materials discovery

作者:

Wen-Kao Li↗Ze-Feng Gao↗Peng-Jie Guo↗Wei Ji↗Zhong-Yi Lu↗

arXiv:2606.16133v1 Announce Type: cross Abstract: Inverse materials design starts from target functionality and searches for structures that can realize it. Its value in closed-loop discovery depends not only on prediction performance, but also on whether expensive first-principles results are independently validated, provenance-recorded, and admitted as feedback only when evidence is sufficient. This is especially important for composite properties such as carrier mobility, where a final scalar value hides intermediate quantities, fit quality, convergence history, and workflow assumptions. Here we present InvDesMobility, a reliability-gated first-principles feedback framework that integrates multi-agent automated DFT, evidence stratification, generative structure proposal, acquisition ranking, and auditable release. Using 516 2DMatPedia-derived candidates, the workflow produced 280 QC-passed materials and 573 retained carrier-direction seed channels after channel-level reliability gating. These records were split into two feedback objects: relaxed structures updated the generative model, while retained mobility channels trained the acquisition model and set validation priority. Over multiple iterations, InvDesMobility screened 2.4 x 10^6 structures, submitted 102 candidates for DFT validation, and retained 86 reliability-gated generated channels across 41 formulas. Overall, the main contribution is not a fixed list of high-mobility materials, but a transferable feedback contract that makes closed-loop inverse design both useful and auditable when learning from expensive calculated properties. All source data, retained feedback records, and workflows are available at https://github.com/DreamLufei/invDesMobility, with an accompanying evidence website at https://dreamlufei.github.io/invDesMobility/.

阅读与讨论 → 访问原文 →

08.

bioRxiv (Bioinfo) 2026-06-19 DOI: HASH:9267e5dd2337c1a7844123940f8ef492

Perturbation Curve models continuous transcriptional response trajectories and improves prediction of genetic modulations

作者:

Zhong↗wang↗Yang↗Yu↗Qi↗Jiang↗

Single-cell CRISPR screens, Perturb-seq, have revolutionized functional genomics by revealing biological causality. However, although perturbation assignments are typically represented as discrete labels, the cell-level effective strength of perturbations is often continuous and diverse. Current analytical frameworks struggle to decouple the variability in perturbation strength from the diversity of downstream responses. Here, we present Perturbation Curve (PertCurve), a nonlinear, curve-based computational framework that models the trajectories of transcriptomic responses by explicitly incorporating diverse perturbation magnitudes and strengths. By ordering cells by perturbation strength, we demonstrate that PertCurve accurately recapitulates the response magnitudes and reveals the distinct modularity and asynchrony patterns of downstream gene behaviors. These patterns are categorized into archetypes, including proportional, sensitive, and threshold responses. By applying this framework across CRISPRi/a modalities, we identify universal response patterns in viral infection, apoptosis, and proliferation genes, and reveal previously overlooked context-specific regulatory features in cell differentiation. Finally, incorporating PertCurve into perturbation prediction models and evaluation metrics enhances predictive performance, delivering actionable insights for refining established models.

阅读与讨论 → 访问原文 →

09.

arXiv (CS.AI) 2026-06-12 DOI: arXiv:2507.07947

Reconstructing Template-Memorized Images from Natural Prompts

作者:

Sol Yarkoni↗Mahmood Sharif↗Roi Livni↗

arXiv:2507.07947v4 Announce Type: replace-cross Abstract: Recent advances in generative models, such as diffusion models, have raised concerns related to privacy, copyright infringement, and data stewardship. To better understand and control these risks, prior work has introduced techniques and attacks that reconstruct images, or parts of images, from training data. While these results demonstrate that training data can be recovered, existing methods often rely on high computational resources, partial access to the training set, or carefully engineered prompts. In this work, we present a new attack that requires low resources, assumes little to no access to the training data, and identifies seemingly benign prompts that can lead to potentially risky image reconstruction. We further show that such reconstructions may occur unintentionally, even for users without specialized knowledge. For example, we observe that for one existing model, the prompt ``blue Unisex T-Shirt'' generates the face of a real individual. Moreover, by combining the identified vulnerabilities with real-world prompt data, we discover prompts that reproduce memorized visual elements. Our approach builds on insights from prior work and leverages domain knowledge to expose a fundamental vulnerability arising from the use of scraped e-commerce data, where templated layouts and images are closely tied to pattern-like textual prompts. The code for our attack is publicly available at https://github.com/TheSolY/lr-tmi.

阅读与讨论 → 访问原文 →

10.

arXiv (CS.LG) 2026-06-16 DOI: arXiv:2510.04902

DP-Hype: Federated Differentially Private Hyperparameter Search

作者:

Johannes Liebenow↗Thorsten Peinemann↗Esfandiar Mohammadi↗

arXiv:2510.04902v3 Announce Type: replace Abstract: Tuning hyperparameters in federated machine learning can substantially impact model performance. When hyperparameters are tuned on sensitive data, privacy becomes an important challenge and to this end, differential privacy has emerged as the de facto standard for provable privacy. A standard setting in federated learning is that clients agree on a shared setup, i.e., find a compromise from a set of hyperparameters, like a model's learning rate. Yet, prior work on privacy-preserving hyperparameter tuning is tailored to specific learning tasks, does not account for the privacy leakage of aggregated results, or offers a sub-optimal privacy-utility trade-off. In this work, we present our algorithm DP-Hype, which performs a federated and privacy-preserving hyperparameter search by conducting a federated voting based on local hyperparameter evaluations of clients. In this way, DP-Hype selects hyperparameters that lead to a compromise supported by a majority of clients, while maintaining scalability and independence from specific learning tasks. We prove that DP-Hype preserves the strong notion of differential privacy called client-level differential privacy and, importantly, show that its privacy guarantees do not depend on the number of hyperparameters. We also provide bounds on its utility guarantees, that is, the probability of finding good hyperparameters, and implement DP-Hype as a submodule in the popular Flower framework for federated machine learning. In addition, we evaluate performance on multiple benchmark data sets in iid as well as multiple non-iid settings and demonstrate high utility of DP-Hype even under small privacy budgets.

阅读与讨论 → 访问原文 →

11.

arXiv (CS.LG) 2026-06-18 DOI: arXiv:2606.18703

Contextualizing Biological Language Models across Modalities via Logit-Space Contrastive Alignment

作者:

Yanjun Shao↗Yundi Chen↗Yashvi Patel↗Aurelien Pelissier↗Mar\'ia Rodr\'iguez Mart\'inez↗

arXiv:2606.18703v1 Announce Type: new Abstract: Pretrained biological language models expose per-token probability distributions through masked-token prediction, providing the likelihood interface central to sequence design, variant scoring, and mechanistic interpretation. Yet these distributions are learned from broad unlabeled corpora and are not naturally conditioned on task-specific biological contexts such as interaction partners, cellular environments, or therapeutic interventions. Existing contextual matching methods often distort this interface through pooled embeddings, contrastive latent spaces, or task-specific prediction heads. We introduce LOGICA (Logit-space Contrastive Alignment), a framework for context-conditioned prediction that performs contrastive learning directly in output-logit space. Using gated cross-modal adapters compatible with each model's native token head, LOGICA preserves the pretrained likelihood interface and converts contextualized token log-likelihoods into matching scores. Alignment is defined through context-sensitive token probabilities rather than proximity in a shared embedding space, enabling learning from sparse paired data across models with distinct vocabularies, without a shared tokenizer or decoder. LOGICA is particularly effective for mutation-local variant ranking, where comparisons reduce to context-conditioned likelihoods of mutant tokens at perturbed sites. Across protein–ligand binding, TCR–peptide activity, and drug-conditioned resistance prediction, LOGICA improves over prior state-of-the-art methods, including matched latent-contrastive and conditional MLM baselines, while retaining a token-level interface for interpretation and generation. On held-out-gene single-mutation drug-resistance prediction, LOGICA improves AUC from near-random latent-space baselines of $\sim$0.55 to $\sim$0.65.

阅读与讨论 → 访问原文 →

12.

arXiv (CS.AI) 2026-06-15 DOI: arXiv:2606.13811

Aligning Quantum Operators with Large Language Models

作者:

Rogerio Feris↗Yunchao Liu↗Pengyuan Li↗Hang Hua↗David Kremer↗

arXiv:2606.13811v1 Announce Type: cross Abstract: Can Large Language Models (LLMs) understand and reason about quantum operators? Despite their remarkable capabilities in mathematics and symbolic reasoning, LLMs remain inherently blind to quantum representations such as unitary matrices. In this work, we take a step toward bridging this gap by introducing an approach that maps unitary operators into the latent space of an LLM, enabling unified modeling over quantum and linguistic inputs. We instantiate this idea on Clifford+T circuit synthesis over a Pauli rotation gate set, where our model achieves results competitive with state-of-the-art methods and scales consistently with training data, with no signs of saturation. Our approach further enables language-conditioned synthesis, allowing gate constraints unseen during training to be specified directly in natural language. This work suggests a path toward quantum–aware foundation models that can natively interpret and reason about quantum operations, which could have broader implications reaching across quantum compilation and algorithm discovery.

阅读与讨论 → 访问原文 →

13.

arXiv (math.PR) 2026-06-18 DOI: arXiv:2606.19115

Finite free perpetuities

作者:

Julia Le Bihan↗Bartosz Ko{\l}odziejek↗

arXiv:2606.19115v1 Announce Type: new Abstract: We introduce and study finite free perpetuities, defined as monic polynomial solutions of degree $n$ to the affine fixed-point equation \[ p(z) = \mathbb{E}\!\left[ A^{n}\,p\!\left(\frac{z-B}{A}\right)\mathbf{1}_{\{A\neq0\}} \right] + \mathbb{E}\!\left[ (z-B)^n\mathbf{1}_{\{A=0\}} \right], \] where $A$ and $B$ are complex-valued random variables with finite moments up to order $n$. Equivalently, if $p(z)=\mathbb{E}[(z-X)^n]$, then $p$ encodes a truncated moment version of the classical perpetuity equation $X\stackrel{d}{=}AX+B$ with $X$ and $(A,B)$ independent. This places finite free perpetuities between classical perpetuities and free-probabilistic fixed-point laws. We prove existence and uniqueness under weak conditions, and we identify a broad class of admissible pairs $(A,B)$ for which the resulting polynomial has only real, nonnegative zeros. Our approach uses finite free additive and multiplicative convolutions together with a probabilistic representation via the $U$-transform. As a motivating example, we exhibit an explicit family of finite free perpetuities expressed in terms of Jacobi polynomials and show that their empirical root distributions converge to a free-beta-prime law. More generally, for admissible sequences of parameters, we prove weak convergence of the empirical root distributions of finite free perpetuities to the law of a free perpetuity characterized by the corresponding free fixed-point equation. This yields a finite-degree polynomial model approximating free perpetuities and clarifies the connection between classical affine recursions, finite free convolutions, and free probability.

阅读与讨论 → 访问原文 →

14.

arXiv (CS.CL) 2026-06-11 DOI: arXiv:2606.11531

Measuring language complexity from hierarchical reuse of recurring patterns

作者:

Junyi Zhou↗Rui Liu↗Pengyu Liu↗Yu Liu↗

We introduce the ladderpath index as a measure of language complexity grounded in algorithmic information theory. It counts the minimum steps needed to reconstruct a sequence through hierarchical reuse of repeated substructures, capturing an exactly computable but constrained form of algorithmic compressibility related to, but distinct from, Kolmogorov complexity. We apply the ladderpath approach to 21 parallel corpora from the Parallel Universal Dependencies dataset. The ladderpath index is approximately invariant across the languages, and varies much less than the corpus length. This is more pronounced when all corpora are mapped to a unified binary representation, providing evidence for the equi-complexity hypothesis from a representation-independent perspective. We also observe trade-offs between character inventory size and corpus length, and between vocabulary-level and corpus-level reconstruction complexity, supporting the trade-off hypothesis that total complexity is conserved and redistributed across linguistic levels. The reusable substructures identified by the ladderpath approach, without any linguistic input, overlap with words and morphological components attested in the natural vocabulary. The hierarchical reuse captured by the ladderpath approach parallels the chunking mechanisms proposed in cognitive science, where the human cognitive system compresses linguistic input into nested, reusable units under shared memory and processing constraints. This connection between cognitive chunking and the ladderpath approach provides a new interpretation for the equi-complexity and trade-off hypotheses, grounding both in the shared cognitive architecture that underlies language processing across human languages.

阅读与讨论 → 访问原文 →

15.

arXiv (CS.CV) 2026-06-11 DOI: arXiv:2606.12142

AerialClaw: An Open-Source Framework for LLM-Driven Autonomous Aerial Agents

作者:

Ke Li↗Jianfei Yang↗Luyao Zhang↗Guo Yu↗Chengwei Yan↗Yuan Ding↗Di Wang↗Nan Luo↗Gang Liu↗Xiao Gao↗Quan Wang↗

Unmanned aerial vehicles (UAVs) are increasingly used in inspection, search and rescue, environmental monitoring, and emergency response. However, most UAV applications still rely on pre-defined command sequences or task-specific pipelines, where developers manually connect perception, planning, flight control, simulation, logging, and safety modules. This limits the flexibility, reproducibility, and extensibility of autonomous aerial systems. This paper presents AerialClaw, an open-source software framework that enables UAVs to operate as decision-making aerial agents rather than merely command-following platforms. Given a natural-language mission, AerialClaw allows an LLM-based agent to understand the task, maintain context, invoke executable aerial skills, observe perception and runtime feedback, and iteratively update its decisions in a closed loop. The framework adopts a modular brain-skill-runtime architecture, combining hard skills for atomic UAV operations, Markdown-based soft skills for reusable task strategies, document-driven agent state and capability boundaries, memory-driven reflection, safety-oriented runtime validation, and platform-agnostic execution adapters. AerialClaw supports lightweight mock execution, PX4 SITL with Gazebo, and AirSim-based simulation, together with a web console, pluggable model backends, example missions, simulation assets, and staged deployment scripts. By combining standardized aerial skills, document-driven agent state, memory, and closed-loop LLM decision-making, AerialClaw provides a reproducible and extensible open-source framework for building UAV systems that can interpret missions, make decisions, execute skills, and adapt their behavior from feedback.

阅读与讨论 → 访问原文 →

16.

arXiv (CS.AI) 2026-06-16 DOI: arXiv:2604.18827

OmniMouse: Scaling properties of multi-modal, multi-task Brain Models on 150B Neural Tokens

作者:

Konstantin F. Willeke↗Polina Turishcheva↗Alex Gilbert↗Goirik Chakrabarty↗Hasan A. Bedel↗Paul G. Fahey↗Yongrong Qiu↗Marissa A. Weis↗Michaela Vystr\v{c}ilov\'a↗Taliah Muhammad↗Lydia Ntanavara↗Rachel E. Froebe↗…

arXiv:2604.18827v2 Announce Type: replace-cross Abstract: Scaling data and artificial neural networks has transformed AI, driving breakthroughs in language and vision. Whether similar principles apply to modeling brain activity remains unclear. Here we leveraged a dataset of 3.1 million neurons from the visual cortex of 73 mice across 323 sessions, totaling more than 150 billion neural tokens recorded during natural movies, images and parametric stimuli, and behavior. We train multi-modal, multi-task models that support three regimes flexibly at test time: neural prediction, behavioral decoding, neural forecasting, or any combination of the three. OmniMouse achieves state-of-the-art performance, outperforming specialized baselines across nearly all evaluation regimes. We find that performance scales reliably with more data, but gains from increasing model size saturate. This inverts the standard AI scaling story: in language and computer vision, massive datasets make parameter scaling the primary driver of progress, whereas in brain modeling – even in the mouse visual cortex, a relatively simple system – models remain data-limited despite vast recordings. The observation of systematic scaling raises the possibility of phase transitions in neural modeling, where larger and richer datasets might unlock qualitatively new capabilities, paralleling the emergent properties seen in large language models. Code available at https://github.com/enigma-brain/omnimouse.

阅读与讨论 → 访问原文 →

17.

arXiv (CS.AI) 2026-06-12 DOI: arXiv:2606.12451

ToolSense: A Diagnostic Framework for Auditing Parametric Tool Knowledge in LLMs

作者:

Ashutosh Hathidara↗Sai Shruthi Sistla↗Sebastian Schreiber↗Sahil Bansal↗

arXiv:2606.12451v1 Announce Type: new Abstract: Large language models deployed as agents over large tool catalogs face a critical tool-retrieval bottleneck. As embedding-based retrieval approaches rely on compact encoders that may under-capture specialized tool semantics, parametric tool retrieval addresses this by encoding each tool as a virtual token appended to the LLM vocabulary, fine-tuned in two stages (memorization then retrieval SFT) to use the LLM as a retriever, achieving strong performance on standard ToolBench retrieval benchmarks. Yet these benchmarks use verbose, fully-specified queries, and their evaluation applies constrained decoding that restricts outputs to valid token paths, neither reveals whether the model actually understands its tools. We introduce ToolSense, an open-source LLM-powered diagnostic framework that takes any tool catalog as input and automatically generates three benchmarks: a Realistic Retrieval Benchmark (RRB) with queries at three ambiguity tiers, an MCQ probing benchmark, and a QA probing benchmark. Applying ToolSense to ToolBench (~47k tools) and evaluating five parametric model training configurations reveals a knowledge-retrieval dissociation: on RRB queries, several configurations collapse by ~50-64 percentage points compared to fully-specified ToolBench benchmarks, falling below the embedding-model baseline. Additionally, despite strong retrieval performance, some models score near-random on factual probes, suggesting a knowledge-retrieval dissociation. We open-source the ToolSense framework and the ToolBench diagnostic benchmarks at https://github.com/SAP/toolsense.

阅读与讨论 → 访问原文 →

18.

arXiv (CS.CL) 2026-06-16 DOI: arXiv:2606.15735

EHRNote-ChatQA: A Benchmark for Evidence-Grounded Multi-Turn Clinical Question Answering over Longitudinal Discharge Summaries

作者:

Jiyoun Kim↗Muhan Yeo↗Eunhye Jang↗Jeewon Yang↗Hangyul Yoon↗Su Ji Lee↗Hee Jo Han↗Hee-Jae Jung↗Doyun Kwon↗Jun young Lee↗Jaehun Lee↗Jung-Oh Lee↗…

Discharge summaries are crucial clinical documents containing the context of a patient's overall hospital stay, and are routinely reviewed by medical experts for patient readmission, ongoing care, and diagnostic decision-making. When reviewing them, medical experts often must iteratively synthesize information across multiple summaries while verifying the evidence supporting each answer. Although large language models (LLMs) are increasingly explored for clinical question answering, existing benchmarks do not sufficiently reflect this setting: they often evaluate exam-style medical knowledge or focus on single-turn question answering with limited evidence-grounding evaluation. We introduce EHRNote-ChatQA, the first benchmark for evidence-grounded multi-turn clinical question answering over patients' multiple discharge summaries. Built from de-identified MIMIC-IV discharge summaries, EHRNote-ChatQA contains 967 patient-level multi-turn samples spanning one to five notes and 16,072 medical-expert-verified QA pairs (8,036 content questions, each paired with an evidence-grounding question) across eight clinical categories. The benchmark is constructed through an expert-informed pipeline combining discharge-summary structuring schema, expert-curated multi-turn QA templates, and LLM-based generation, followed by review and revision of every single QA sample by 11 medical experts. Benchmarking 22 open- and closed-source LLMs reveals several challenges, including that LLMs struggle more with evidence grounding than content answering, multi-turn errors compound across turns, and single-turn clinical QA performance does not reliably transfer to this setting. These findings establish EHRNote-ChatQA as a rigorous and practical benchmark for evaluating clinical QA systems. The dataset will be made publicly available through PhysioNet credentialed access.

阅读与讨论 → 访问原文 →

19.

arXiv (CS.CL) 2026-06-12 DOI: arXiv:2606.13254

Evaluating Pluralism in LLMs through Latent Perspectives

作者:

Laura Majer↗Jan \v{S}najder↗Martin Tutek↗

The growing need to represent diverse perspectives has increased interest in pluralistic LLM generation. Although difficult to operationalize, identifying perspectives expressed in text would provide clear guidance on pluralistic alignment and more clearly articulate the pluralistic gap in LLM generation. While models have been shown to reduce the diversity of training data and generate homogeneously, this has been demonstrated primarily on multiple-choice questionnaires or using high-level characteristics of free-form text. In this paper, we introduce and implement a domain-agnostic multi-layered framework for unsupervised extraction of perspectives suitable for identifying the pluralistic gap in LLM-generated text. We evaluate our framework on book reviews, a highly opinionated dataset representing diverse perspectives, and compare various prompts and models. Our results show that while some models and prompting techniques come close to covering a broad spectrum of perspectives, rarer perspectives remain disproportionately underrepresented, resulting in distributions that diverge from human text.

阅读与讨论 → 访问原文 →

20.

arXiv (CS.AI) 2026-06-16 DOI: arXiv:2606.15376

CoAgent: Concurrency Control for Multi-Agent Systems

作者:

Hongtao Lyu↗Dingyan Zhang↗Mingyu Wu↗Xingda Wei↗Haibo Chen↗

arXiv:2606.15376v1 Announce Type: cross Abstract: Multi-agent LLM systems – coding agents, devops agents, document agents – now routinely run several agents in parallel against the same git tree, Kubernetes cluster, or document. As soon as two of them mutate shared state, they enter the regime classical concurrency control has studied for decades, but classical mechanisms fit LLM agents poorly. A single agent transaction spans minutes of inference, read sets are broad and opaque rather than statically inferable, and the live state agents act on admits neither fork nor buffer, so writes take effect the moment they execute. Locks block long inference intervals; OCC abort-and-retry discards minutes of work on every conflict. This paper builds concurrency control on a capability classical transactions lack: the LLM inside each agent can judge whether a conflicting write invalidates its plan, and can repair exactly the operations that depended on it. Control therefore turns advisory: the runtime informs, the agent repairs. Our protocol, MTPO (Monotonic Trajectory Pre-Order), fixes a serialization order at launch, serves each read the order-filtered value, and applies writes speculatively in place; a one-way notification asks an affected reader to re-judge and patch its plan, while the framework mechanically undoes and reorders misplaced writes through the saga-style inverse each tool registers in advance. At quiescence the run is serializable in the pre-decided order. We realize MTPO as CoAgent, toolcall middleware whose privileged ToolSmith grows footprint-declared, undoable tools online. On ten contended workloads, CoAgent stays within 5\% of serial correctness at a $1.4\times$ speedup and near-serial token cost, where 2PL and OCC surrender nearly all concurrency gains; on a bash-only target system, it grows a 25-tool library online and lifts the task pass rate from 45/71 to 63/71 at $0.80\times$ the time and $0.86\times$ the cost.

阅读与讨论 → 访问原文 →

21.

medRxiv (Medicine) 2026-06-15 DOI: HASH:8fb88c5b5b1a208ef8a72b4887b291b7

GLLaucoMed: A Secure LLM-Powered Agentic Workflow for Automated Medication Extraction from Free-Text Glaucoma Clinical Notes

作者:

Solages↗Scherer↗Samico↗G. A↗Gutkind↗N. E↗Kang↗Medeiros↗F. A↗Swaminathan↗S. S↗

Purpose: To evaluate the efficacy of large language models (LLMs) in extracting medication-related information from glaucoma clinical notes in the electronic health record (EHR). Design: Cross-sectional. Subjects: 1,250 subjects in the Bascom Palmer Ophthalmic Repository. Methods: Extracted clinical notes from glaucoma-related encounters between 2014 and 2024 were labeled by two glaucoma specialists with a third serving as an adjudicator. Graders were asked to label current topical medications (CTM), proposed changes to topical medications ({Delta}TM), current oral medications (COM), and proposed changes to oral medications ({Delta}OM) in a structured fashion. The dataset was split into development (10%), validation (10%), and test (80%) sets stratified by clinician. Development and validation sets were used to engineer and refine prompts, and the held-out test set was used for model assessment. Five LLMs (Claude Opus 4.6, DeepSeek-V3.2, GPT 5.2, Grok 4.1, and Qwen3.6-35B-A3B) were accessed via Microsoft Azure AI Foundry within a HIPAA-compliant environment. Inter-grader agreement was assessed with Gwet AC1. LLM performance was initially assessed in a binary fashion with F1 scores, and the degree of text match among positive cases was evaluated using exact match accuracy and Jaccard Index (JI). Main Outcome Measures: F1 score, exact match accuracy, JI. Results: Gwet AC1 for intergrader agreement was 0.799, 0.888, 0.985, and 0.988 for CTM, {Delta}TM, COM, and {Delta}OM, respectively. F1 scores for CTM were 0.985, 0.971, 0.978, 0.968, and 0.970 for Claude, Deepseek, GPT, Grok, and Qwen, respectively; for {Delta}TM: 0.905, 0.826, 0.897, 0.842, 0.855, respectively; for COM: 0.923, 0.887, 0.899, 0.906, 0.894, respectively; for {Delta}OM: 0.958, 0.815, 0.937, 0.835, 0.940, respectively. Among positive cases, range of exact match accuracies for CTM (N=1354) was 0.730- 0.882 and range of JIs was 0.809-0.918. For {Delta}TM (N=404), exact match accuracy range was 0.619-0.780 and JI range was 0.668-0.827. For COM (N=47), exact match accuracy range was 0.766-0.872 and JI range was 0.765-0.870. For {Delta}OM (N=25), exact match accuracy range was 0.583-0.920 and JI range was 0.583-0.922. Conclusions: The GLLaucoMed pipeline demonstrated high performance in extracting and standardizing medication data from unstructured clinical notes, including both current medications and proposed changes. Claude and GPT exhibited the strongest performance.

阅读与讨论 → 访问原文 →

22.

arXiv (CS.LG) 2026-06-15 DOI: arXiv:2606.14361

SemPiper: Interactive Code Synthesis for Semantic Operators in Machine Learning Pipelines

作者:

Olga Ovcharenko↗Luciano Duarte↗Sebastian Schelter↗

arXiv:2606.14361v1 Announce Type: new Abstract: Machine learning (ML) pipelines require extensive data preparation, feature engineering, and integration across heterogeneous sources, making them tedious and error-prone to develop. While large language models (LLMs) have recently shown promise for assisting programming tasks, chat-based interfaces provide limited control over pipeline behavior and often produce code that is difficult to optimize or integrate into production systems. We demonstrate SemPipes, a novel programming model that extends ML pipelines with declarative, LLM-powered semantic data operators. SemPipes allows developers to specify high-level natural language instructions for data-centric operations, while seamlessly combining these operators with arbitrary Python code from standard data science libraries. For the semantic operators, it synthesizes specialized implementations at pipeline training time, conditioned on dataset characteristics and pipeline context, enabling the flexible yet controlled integration of LLM capabilities. We demonstrate SemPipes through SemPiper, an interactive interface that visualizes computational graphs of the pipelines, synthesized operator implementations, and optimization trajectories produced by an evolutionary search procedure. Attendees can explore three end-to-end scenarios, modify pipelines, inspect generated code, and observe how semantic operators are synthesized and iteratively optimized. The demonstration highlights how declarative semantic operators enable controllable, optimizable, and practical integration of LLMs into ML pipeline development.

阅读与讨论 → 访问原文 →

23.

arXiv (CS.AI) 2026-06-15 DOI: arXiv:2606.14238

When and How Severely: Scenario-Specific Safety Envelopes for Driving VLAs

作者:

Abhinaw Priyadershi↗Jelena Frtunikj↗

arXiv:2606.14238v1 Announce Type: cross Abstract: Safety certification of Vision-Language-Action (VLA) driving planners under ISO 21448 (SOTIF) rests on an Operational Design Domain (ODD) specification that answers two complementary questions: when does the planner start to fail, and how severely does it fail once it does? We evaluate Alpamayo R1, a 10B-parameter open-weight driving VLA, on 15,968 (clip, attack) pairs. We find a conservative-aggregate gap: an aggregate safe threshold of $\sigma \leq 50$ under a 15% average displacement error (ADE) budget masks well-sampled scenarios that tolerate the top of the tested grid ($\sigma = 70$). A Gaussian Mixture Model (GMM) on the changed-explanation subset identifies six discrete severity bands (BIC-optimal $k{=}6$), so two perturbation conditions with the same mean error can differ materially in their share of high-severity (C4/C5) failures. Joining the two analyses on the same corpus surfaces a finding neither yields in isolation: the scenarios with the loosest noise thresholds are not those with the lowest high-severity rate: STOP_SIGNAL concentrates roughly $4\times$ the C4/C5 share of LANE_KEEPING despite tolerating a larger $\sigma$. A deployable SOTIF ODD specification for driving VLAs therefore requires a two-dimensional safety envelope, not a single aggregate value per hazard.

阅读与讨论 → 访问原文 →

24.

medRxiv (Medicine) 2026-06-15 DOI: HASH:43959628b96f9aefe23ac94ed0b82689

Beyond the Apnea-Hypopnea Index: Physiological and Demographic Predictors of Excessive Daytime Sleepiness in Obstructive Sleep Apnea

作者:

Tan↗Parekh↗Wickramaratne↗S. D↗

Excessive daytime sleepiness (EDS) is a common but inconsistently predicted symptom of obstructive sleep apnea (OSA). OSA is typically diagnosed with polysomnography (PSG), and the current standard for severity assessment is the apnea-hypopnea index (AHI). AHI has many limitations, including its inability to explain physiological mechanisms or reflect variability in patient symptoms, such as EDS. This retrospective study aims to find physiological and demographic parameters that better predict EDS in patients with OSA and to evaluate whether these parameters outperform AHI using PSG data from the Mount Sinai Integrative Sleep Center. Clinical variables used to predict EDS included arousal index (AI), average oxygen desaturation during sleep, average heart rate during sleep, and AHI, along with demographic variables including age, sex, and BMI. Hypothesis tests, logistic regression models, and decision tree classifier models were performed on the data to discriminate sleepy from nonsleepy patients as determined by an Epworth Sleepiness Scale (ESS) score [≥] 10. AI and oxygen desaturation were found to be the most predictive physiological variables, and sex and BMI were found to be the most predictive demographic variables. The final decision tree model with these four variables outperformed the AHI in predicting EDS. These findings suggest that daytime sleepiness in OSA can be better explained by measures of apnea burden, oxygenation impairment, and patient demographics than by AHI alone, although these remain only modestly predictive. Future studies should focus on investigating more comprehensive physiological markers, multi-night sleep data, and more objective assessments of sleepiness.

阅读与讨论 → 访问原文 →

25.

arXiv (math.PR) 2026-06-11 DOI: arXiv:2602.20266

Multiple Poisson-Dirichlet diffusions on generalized Kingman simplices

作者:

Cristina Costantini↗Matteo Ruggiero↗

arXiv:2602.20266v2 Announce Type: replace Abstract: We construct a new class of infinite-dimensional diffusions with values in a generalized Kingman simplex with finitely many marks. The model describes the temporal evolution of the relative frequencies of infinitely many types that are labeled by a finite number $H$ of marks, but unlabeled within each mark. We first establish a blockwise skew-product representation for a finite-type Wright-Fisher diffusion, extending the aggregation-renormalization self-similarity property of Dirichlet laws. The decomposition separates an $H$-dimensional Wright-Fisher diffusion governing the evolving random mark masses, from $H$ Wright-Fisher diffusions, each run on its own random clock, which describe the evolution of the relative frequencies within each mark. After ranking the within-mark frequencies in decreasing order, we identify the distributional limit as the number of types per mark tends to infinity and we derive an explicit form of its infinitesimal generator on a suitable domain. The limiting diffusion admits the multiple Poisson-Dirichlet distribution as a stationary distribution; it recovers the infinitely-many-neutral-alleles diffusion when all types share the same mark and yields a diffusion on the Thoma simplex when there are two marks.

阅读与讨论 → 访问原文 →