arXiv:2606.17680v1 Announce Type: new Abstract: Reinforcement learning (RL) has emerged as a powerful paradigm for training Large Language Models (LLMs) as agents. However, conventional RL methods for long-horizon agentic tasks often struggle with sparse outcome rewards. Intuitively, this overlooks the rich environment dynamics information contained in rollout interaction trajectories. We argue that the interaction experience inherently serves as an implicit supervision signal, reveals the underlying transition mechanisms of the environment, and enables the agent to construct a more accurate internal model of the environment.. Therefore, in this work, we investigate how to leverage this additional signal to improve policy learning. Specifically, we propose EnvRL, a framework that incorporates environment dynamics learning into agentic RL via two auxiliary objectives: state prediction and inverse dynamics. By jointly optimizing with the primary RL objective, we encourage the agent to internalize environment dynamics from its own interaction experience. Extensive experiments on two long-horizon agentic benchmarks demonstrate that EnvRL achieves significant improvements on success-rates over RL-only baselines, e.g., when trained with GRPO, lifting Qwen-2.5-1.5B-Instruct from 72.8% to 77.4% on ALFWorld, and from 56.8% to 67.0% on WebShop.
Science Journals
arXiv:2606.17682v1 Announce Type: new Abstract: Reinforcement learning pipelines for Large Language Model (LLM) training often rely on manually redesigned environments between stages, requiring practitioners to heuristically infer which configuration will best improve the current policy. To automate this process, we propose the LLM-as-Environment-Engineer framework in which the current policy model analyzes failure trajectories together with contextual information and proposes modifications to the next-stage training environment configuration. We also introduce MAPF-FrozenLake, a controllable testbed whose generator exposes multi-dimensional environment configurations, making it suitable for studying and benchmarking environment redesign. On this testbed, we condition the environment engineer on structured summaries of policy behavior, failure cases, and environment statistics, from which it produces the configuration for the next training stage. With Qwen3-4B as the backbone, our framework achieves the strongest aggregate performance on our benchmarks, outperforming larger proprietary LLMs (e.g., GPT, Gemini) and fixed-environment training baselines. We further analyze which forms of context are most effective, finding that successful environment updates rely on failure evidence and preserve configurations that already work. Interestingly, the current RL checkpoint serves as a better environment engineer than the original base model, suggesting that policy learning improves the model's ability to diagnose its remaining weaknesses.
arXiv:2606.17819v1 Announce Type: new Abstract: Agent skills -- structured, reusable knowledge artifacts that augment LLM agent capabilities -- have been rapidly adopted in industry, yet their cross-domain impact and use across commercial and open-source models remain under-studied, and no reusable methodology exists for evaluating an individual skill. In this work, we present an evaluation framework that lets a skill author construct realistic tasks to rigorously assess the aspects of a skill that matter most to them, and that estimates skill utility by solving those tasks. Further, we apply our evaluation approach at scale to 500 real-world skills, generating 1,000 tasks derived from the skills' content, along with instruction-following and goal-completion scoring rubrics. Using these metrics, we evaluate how 19 agent-model configurations, both proprietary and open-source, perform on the tasks. Our results show that models vary widely in how closely they adhere to the instructions encoded in skills, leading to substantial differences in their performance gains. Furthermore, we show that access to a skill significantly changes model behavior compared to the no-skill setup, providing an essential mechanism for encoding opinionated workflows into LLM agents. We release our evaluation dataset to support future work on agent skills.
arXiv:2601.21455v2 Announce Type: replace-cross Abstract: Conformal prediction(CP) has become a cornerstone of distribution-free uncertainty quantification, conventionally evaluated by its coverage and interval length. This work critically examines the sufficiency of these standard metrics. We demonstrate that the interval length might be deceptively improved through a counter-intuitive approach termed Prejudicial Trick(PT), while the coverage remains valid. Specifically, for any given test sample, PT probabilistically returns an interval, which is either null or constructed using an adjusted confidence level, thereby preserving marginal coverage. While PT potentially yields a deceptively lower interval length, it introduces practical vulnerabilities: the same input can yield completely different prediction intervals across repeated runs of the algorithm. We formally derive the conditions under which PT achieves these misleading improvements and provide extensive empirical evidence across various regression and classification tasks. Furthermore, we introduce a new metric interval stability which helps detect whether a new CP method implicitly improves the length based on such PT-like techniques. Code is available at https://github.com/benben-cd/PT-Conformal-Prediction.
arXiv:2501.09876v4 Announce Type: replace Abstract: Generative modeling aims to generate new data samples that resemble a given dataset. When using diffusion models for this task, one of the main challenges is solving the problem in the input space, which tends to be very high-dimensional. To address this, recent approaches solve diffusion models in the latent space through an encoder that maps from the data space to a lower-dimensional latent space, improving training efficiency and achieving state-of-the-art results. The variational autoencoder (VAE) is the most commonly used encoder/decoder framework in this domain, known for its ability to learn latent representations and generate data samples. In this paper, we introduce a novel encoder/decoder framework with theoretical properties distinct from those of the VAE, specifically designed to preserve the geometric structure of the data distribution. We demonstrate the significant advantages of this geometry-preserving encoder in the training process of both the encoder and decoder. Additionally, we provide theoretical results proving convergence of the training process, including convergence guarantees for encoder training, and results showing faster convergence of decoder training when using the geometry-preserving encoder.
arXiv:2502.08363v3 Announce Type: replace Abstract: We present Top-Theta (Top-$\theta$) Attention, a training-free method for sparsifying transformer attention during inference. Our key insight is that static, per-head thresholds can be calibrated to retain the desired constant number of significant elements per attention row. This approach enables content-based sparsity without retraining, and it remains robust across data domains. We further introduce compensation techniques to preserve accuracy under aggressive sparsification, establishing attention thresholding as a practical and principled alternative to top-k attention. We provide extensive evaluation on natural language processing tasks, showing that Top-$\theta$ achieves 3-10x reduction in V-cache usage and up to 10x fewer attention elements during inference while degrading no more than 1% in accuracy.
arXiv:2511.01650v3 Announce Type: replace Abstract: Large Language Models (LLMs) are increasingly entering specialized, safety-critical engineering workflows governed by strict quantitative standards and immutable physical laws, making rigorous evaluation of their reasoning capabilities imperative. However, existing benchmarks such as MMLU, MATH, and HumanEval assess isolated cognitive skills, failing to capture the physically grounded reasoning central to engineering, where scientific principles, quantitative modeling, and practical constraints must converge. To enable verifiable process supervision in engineering, we introduce EngTrace, a symbolic benchmark built on 90 parameterized templates, each generating unique, contamination-resistant problem instances, spanning three major engineering branches, nine core domains, and 20 distinct areas, yielding 1,350 test cases that stress-test generalization across diverse physical scenarios. Moving beyond outcome matching, we introduce a verifiable two-stage evaluation framework that uses a tiered protocol to validate intermediate reasoning traces alongside final answers through automated procedural checks and a heterogeneous AI Tribunal. Our evaluation of 27 leading LLMs reveals a distinct trade-off between numeric precision and trace fidelity, identifying a complexity cliff where abstract mathematical pre-training fails to translate into the integrative reasoning required for advanced engineering tasks.
arXiv:2511.03211v4 Announce Type: replace Abstract: This paper examines the role of public interest litigation in promoting accountability for AI and automated decision-making (ADM) in Australia. Since ADM regulation faces political and geopolitical headwinds, effective governance will have to rely on the enforcement of existing laws. Drawing on interviews with Australian public interest litigators, technology policy activists, and technology law scholars, the paper positions public interest litigation as part of a larger ecosystem for transparency, accountability and justice with respect to ADM. The paper explores the tactics and strategies of what one participant described as 'retrofitting' old laws to ADM. These go beyond creative legal argumentation, to encompass practices of community-building, collaboration on theories of change, canny selection of clients and causes of action, and the alignment of the interests of stakeholders in litigation. Naturally, the paper also contends with the limits of these strategies, and of the Australian legal system. Where limits are, however, capable of being overcome, the paper presents findings on urgent needs: the enabling institutional arrangements without which effective litigation and accountability will falter. The paper is relevant to law and technology scholars; individuals and groups harmed by ADM; public interest litigators and technology lawyers; civil society and advocacy organisations; and policymakers.
arXiv:2606.17733v1 Announce Type: new Abstract: Particle-in-cell (PIC) simulations are widely used for kinetic plasma modeling in energy applications, but their efficiency is often limited by repeated field solves on dense meshes. This work proposes a Latent Residual-Closure Fourier Neural Operator (LRC-FNO) for robust surrogate multi-field solving in PIC simulations. Rather than treating field prediction as a purely data-driven regression task, LRC-FNO formulates PIC field solving as a two-level residual-closure problem involving source compression and source-to-field operator mapping. An autoencoder extracts compact representations of particle-deposited source fields, while a Latent Closure Refiner recovers unresolved residual structures lost during compression. A Coarse-FNO Solver captures the dominant field response, and a Residual-Closure FNO restores full-resolution corrections. The method is tested on three benchmarks with increasing complexity: 1D linear Landau damping (LLD), 2D two-stream instability (TSI), and a 2D scrape-off layer (SOL) fusion plasma model. In LLD and TSI, LRC-FNO better preserves charge-to-potential mapping, potential-mode evolution, residual charge structures, and particle-field energy exchange during closed-loop PIC integration. In the SOL case, LRC-FNO achieves relative L2 errors of 0.0447 for the self-consistent potential and 0.0251 for the magnetic vector potential in single-step prediction. More importantly, when used as a neural initial guess with 20 iterative corrections, LRC-FNO maintains strong physical consistency in extrapolated closed-loop simulations, preserving charge and current density structures over a time range close to twice the training horizon. These results demonstrate that LRC-FNO can serve as both a fast surrogate field solver and a high-quality initialization strategy for iterative PIC field solvers.
arXiv:2606.17735v1 Announce Type: new Abstract: Although reinforcement learning (RL) has expanded the cognitive boundaries of large language models (LLMs), it often remains vulnerable to the autoregressive curse in long-horizon logical reasoning: small epistemic perturbations introduced early in generation can propagate irreversibly along the Markov decision process flow, triggering cascading failures that drive the reasoning trajectory toward collapse. To overcome this autoregressive cascade, in which a single early mistake can compromise all subsequent reasoning steps, we propose dynamic epistemic entropy orchestrated erasable reinforcement learning ($\text{E}^3\text{RL}$). $\text{E}^3\text{RL}$ eliminates reliance on external signals by grounding the model's endogenous local autoregressive cross-entropy as an intrinsic coordinate of epistemic uncertainty. By introducing segment-level adaptive dynamic thresholds and advantage allocation, $\text{E}^3\text{RL}$ enables the model to precisely excise localized logical defects while reusing historical key-value (KV) cache streams, thereby endowing the reasoning process with a self-healing capability. We train $\text{E}^3\text{RL}$ on the DeepMath-103k dataset. Experimental results show that $\text{E}^3\text{RL}$ reshapes the exploration efficiency of long-sequence reasoning and improves sample efficiency while maintaining linear memory overhead. On mathematical reasoning benchmarks such as AIME, $\text{E}^3\text{RL}$ achieves substantial performance gains, with the 4B and 8B parameter models surpassing previous state-of-the-art (SOTA) results by 5.349\% and 6.514\%, respectively. These findings suggest that $\text{E}^3\text{RL}$ shatters the autoregressive curse in long-sequence reasoning and establishes a theoretical and systems-level foundation for the next generation of self-healing artificial general intelligence (AGI).
arXiv:2603.03573v2 Announce Type: replace Abstract: Discrete biological sequence optimization often requires goal-directed, parser-valid edits to an existing protein or molecule. Diffusion models support iterative refinement but do not expose a controllable discrete-edit interface, while autoregressive LLMs can be myopic when planning constrained edits over multiple steps. We introduce STRIDE (Sequence Trajectory Refinement via Iterative Discrete Editing), a post-training framework that trains an LLM to emit executable INSERT/DELETE/REPLACE trajectories for variable-length refinement. STRIDE first learns Levenshtein-aligned shortest-edit demonstrations, then uses supervised fine-tuning and group-based policy optimization to align trajectories with task rewards while preserving coherent editing. On an oracle-based full-action protein stress test, STRIDE raises success over Vanilla SFT from 42% to 89% and novelty among unique improvements from 47% to 97%. On instruction-conditioned molecular editing, the GSPO-aligned variant improves strict success, controllability, and SMILES validity over the SFT-only STRIDE model (code: https://github.com/daiheng-zhang/STRIDE).
arXiv:2605.18362v2 Announce Type: replace Abstract: In a previous paper, a process algebra based on ACP (Algebra of Communicating Processes) was proposed in which processes involving data can be handled by means of features originating from imperative programming. In this paper, an extension of that process algebra with probabilistic choice operators is presented that rests on the principle that probabilistic choices are always resolved before choices involved in alternative composition and parallel composition are resolved. This extension is devised, among other things, to be used for modeling and analyzing algorithms that are important in the area of distributed computing. Many canonical problems in that area call for a probabilistic algorithm. In this paper, a probabilistic algorithm for the leader election problem is modeled using the presented process algebra.
arXiv:2605.20708v2 Announce Type: replace Abstract: Diffusion Transformers (DiTs) have become a de facto backbone of modern visual generation, and nearly every major axis of their design -- tokenization, attention, conditioning, objectives, and latent autoencoders -- has been extensively revisited. The residual stream that governs how information accumulates across layers, however, has been directly inherited from the original Transformer. In this paper, we present a systematic empirical analysis of cross-layer information flow in DiTs, jointly along depth and denoising timestep, and identify three concrete symptoms of traditional residual addition, namely monotonic forward magnitude inflation, sharp backward gradient decay, and pronounced block-wise redundancy. Motivated by this diagnosis, we propose Diffusion-Adaptive Routing (\textsc{DAR}), a drop-in residual replacement that performs \emph{learnable, timestep-adaptive, and non-incremental} aggregation over the history of sublayer outputs. Moreover, the proposed \textsc{DAR} is compatible with many modern Transformer enhancement methods, such as REPA. On ImageNet $256\times256$, \textsc{DAR} improves SiT-XL/2 by $2.11$ FID ($7.56$ vs.\ $9.67$) and matches the baseline's converged quality with $8.75\times$ fewer training iterations. Stacked on top of REPA, it yields a $2\times$ training acceleration in the early stage, suggesting cross-layer information routing as an underexplored design axis in diffusion modeling, one that operates orthogonally to existing representation-alignment objectives. Beyond pretraining, \textsc{DAR} can also be applied during the fine-tuning stage of large-scale T2I models and preserves high-frequency details during Distribution Matching Distillation.
arXiv:2606.17798v1 Announce Type: new Abstract: Despite the remarkable progress of Video Large Language Models (Video-LLMs), current online architectures still struggle to simultaneously process continuous video streams, decide autonomously when to respond, and preserve long-horizon contextual memory. These obstacles undermine real-time responsiveness and cause severe forgetting throughout prolonged interactions. In this work, we introduce LiveStarPro, a live streaming assistant that is designed for proactive video understanding over long-horizon streams. The design of LiveStarPro rests on three complementary components. The first component is Streaming Verification Decoding (SVeD), an inference framework that identifies the appropriate response timing through single-pass perplexity verification, thereby eliminating the dependency on explicit silence tokens. The second component is Streaming Causal Attention Masks (SCAM), a training strategy that enforces incremental video-language alignment over variable-length streams. The third component is Tree-Structured Hierarchical Memory (TSHM), a recursive memory architecture that organizes evicted historical information into event chains and consequently enables efficient retrieval from effectively unbounded video streams. To facilitate a comprehensive evaluation under realistic online conditions, we further present OmniStarPro, a large-scale benchmark that spans 15 diverse real-world scenarios and that extends to hour-scale streams for the assessment of long-term recall. Extensive experiments demonstrate that LiveStarPro consistently surpasses existing methods, attaining a 28.9% improvement in semantic correctness and an 18.2% reduction in timing error, while its streaming key-value cache further yields a 1.58x inference speedup over the same model without caching. The model and the code are publicly available at https://github.com/sotayang/LiveStarPro.
arXiv:2606.17800v1 Announce Type: new Abstract: As an increasing majority of global video content is consumed on social platforms for interactive social purposes, video generation models built for social worlds are important but largely overlooked by previous studies. In this work, we define the position of social world models and build a prototype model as the first step towards this goal. While previous world models successfully simulate physical environments or gaming world exploration, they remain fundamentally detached from human-centric social dynamics. To bridge this gap as the first step to social world models, we present MaineCoon, the first real-time audio-visual autoregressive model that has 22B parameters and is capable of real-time streaming generation and sub-second interaction, with a record-breaking frame rate of up to 47.5 FPS, on a single GPU. To the best of our knowledge, MaineCoon is also the first real-time audio-visual generation model specifically optimized for social-interactive applications. To enable efficient and stable training, we introduce several novel techniques into MaineCoon, including self-resampling, cross-modal representation alignment, domain-aware preference optimization, and reinforced online-policy distillation (ROPD). We also design the first agentic streaming inference framework that supports thousand-second-scale or even longer generation while mitigating drift with agentic cache management and prompt planing. These innovations significantly accelerate training while optimizing real-time inference performance. We believe this work not only sets a new state-of-the-art (SOTA) performance benchmark for high-quality, low-latency, and long-horizon audio-visual autoregressive models, but also points out the paradigm shift desired for next-generation AI-native social platforms.
arXiv:2606.17860v1 Announce Type: new Abstract: The coordinated attack problem models the challenge of coordinating a joint action within a bounded time by communicating over unreliable links. It was the first distributed computing problem proven unsolvable. Its analysis also revealed the importance of common knowledge, a central concept in epistemic logic. However, the randomized version of coordinated attack, which is solvable, has not, to the best of our knowledge, been studied through the lens of probabilistic epistemic logic, where processes generate randomness by flipping coins. We present an epistemic logic framework for studying randomized algorithms that execute for a bounded number of rounds. The framework applies to coordinated attack, approximate agreement, and consensus, and supports dynamic graph models: synchronous systems in which reliable processes execute a bounded number of rounds while an adversary determines which messages are lost. Our approach combines techniques from the logical characterization of dynamic networks and task solvability with ideas from probabilistic dynamic epistemic logic. It is inspired by the operational model of Varghese and Lynch for randomized coordinated attack. More broadly, the resulting notion of probabilistic epistemic task solvability provides a foundation for the epistemic study of randomized distributed computation. Using this framework, we analyze the Varghese-Lynch algorithm from a knowledge-theoretic perspective, providing a formal treatment of the algorithm and its lower bound. As a byproduct, we strengthen the lower bound and show it is tight. The proof relies on indistinguishability arguments, demonstrating that reasoning about knowledge remains essential in the probabilistic setting. We also formalize the notion of information level introduced by Varghese and Lynch, showing that it corresponds to a specific epistemic formula.
arXiv:2606.17805v1 Announce Type: new Abstract: Data acquisition is a major bottleneck for learning in real-time streams: analysts must decide on the fly which labels to purchase while respecting a rolling budget. However, existing online active learning rarely unifies pricing, information gain, and rolling budget constraints under concept drift. We introduce QueryMarket, a market-inspired framework that queries each incoming data point based on its estimated utility to the model and its price. Within this framework, we propose OVBAL (online variance-based active learning), which integrates data pricing with information-driven selection by estimating each sample's marginal utility via a D-optimality criterion with exponential forgetting and executing cost-aware purchases under rolling budget constraints. OVBAL yields a simple, fully online decision rule that adapts to nonstationary streams and heterogeneous label costs. Experiments on synthetic data and a real-world solar power generation forecasting task show that OVBAL is particularly effective under seller-centric pricing and yields a more favorable long-run error-cost trade-off in the real-world task under both pricing schemes.
arXiv:2606.17872v1 Announce Type: new Abstract: Large language models (LLMs) outperform earlier architectures on generative inference and long-context tasks, but their large size introduces significant challenges in memory usage, energy cost, and on-device deployment. Since scaling pre-trained language models improves downstream capability \cite{zhao2023survey}, the key-value (KV) cache becomes a dominant inference bottleneck. Recent KV cache compression methods \cite{jo2025fastkv,li2024snapkv,zhou2024dynamickv} reduce this cost by retaining only a subset of attention-relevant tokens. However, while these approaches preserve accuracy on benign workloads, their compression policies either fail to defend against jailbreak attacks \cite{jiang2024robustkv} or degrade safety alignment under aggressive eviction. We propose AnchorKV, a drop-in modification to KV cache compression that biases token retention scores away from directions in key space associated with harmful prompts. AnchorKV constructs an offline safety anchor by adapting a difference-of-means representation engineering approach \cite{arditi2024refusal,zou2023representation} to the layer-specific key projection space used in KV caching. Based on this anchor, a soft penalty token selection rule trades a small amount of utility for substantially improved safety alignment, while reducing to the original compressor when the penalty is zero.
arXiv:2606.17888v1 Announce Type: new Abstract: Chain-of-Thought (CoT) reasoning has extended from purely linguistic domains to multimodal scenarios; however, existing approaches often treat visual inputs as homogeneous or auxiliary signals, failing to capture the intricate and sample-specific dependencies between text and images in mathematical problem-solving. This gives rise to two core issues: first, the supervisory signals for visual content are generalized and coarse-grained, lacking adaptation to the actual necessity of visual information in each sample; second, training feedback becomes inaccurate when visual rewards are uniformly applied without distinguishing the complementary relationships among inputs. These limitations hinder models from achieving precise multimodal reasoning. In this work, we propose a framework for modeling fine-grained visual dependencies in mathematical reasoning. We first construct the MathVis-Fine dataset, augmenting fine-grained visual annotations with visual dependency ratings. Building upon this dataset, we introduce a two-stage progressive visual enhancement training paradigm that balances answer correctness rewards and visual grounding rewards according to the intrinsic visual dependency level of each sample, thereby mitigating reward bias and improving supervision accuracy. Extensive experiments demonstrate that the MathVis-Fine framework effectively enhances visual perception progressively based on visual dependency, offering a more precise training framework for multimodal mathematical reasoning. We will release the dataset upon acceptance.
arXiv:2606.12863v2 Announce Type: replace Abstract: Multimodal attributed graphs (MAGs) integrate graph topology with heterogeneous modality attributes, such as text and images, thereby enabling richer modeling of complex relational systems. However, such expressiveness also makes learning on MAGs depend on multiple semantic sources, including structural topology, textual and visual attributes, each of which can be regarded as a branch for node representation. Node-level branch semantic imbalance arises when these branches differ across nodes in semantic informativeness and reliability: a branch that provides discriminative semantics for one node may mislead another due to bias in modality quality or structural context. Existing methods often mitigate such heterogeneity through cross-branch agreement or alignment, implicitly treating the dominant prediction as reliable supervision. When the dominant branch is biased, forced imitation may propagate its bias to other branches and suppress original semantics that are useful for classification. We propose GraphMNL, a graph-aware multimodal negative learning framework that addresses this issue by using Negative Learning as cross-branch guidance. Instead of forcing inferior branches to imitate a teacher prediction, the model teaches them which classes a node is unlikely to belong to. GraphMNL builds a branch library, identifies dominant and inferior branches via graph-aware reliability arbitration, gates unstable transfer, and applies target-preserving negative learning over non-target classes. This design decouples target supervision from branch guidance so that supervised losses learn the correct class, while Negative Learning suppresses unlikely alternatives when branch agreement is unreliable. Through the comprehensive experimental evaluation, GraphMNL achieves the best performance on Grocery datasets with 72.47% accuracy and 76.60 F1 score on Reddit M datasets.
arXiv:2602.09802v2 Announce Type: replace Abstract: As Large Language Models (LLMs) are increasingly deployed in applications such as travel assistance and purchasing support, they are often required to make subjective choices on behalf of users in settings where no objectively correct answer exists. We study LLM decision-making in a travel-assistant context by presenting models with choice dilemmas and analyzing their responses using multinomial logit models to derive implied willingness to pay (WTP) estimates. These WTP values are subsequently compared to human benchmark values from the economics literature. In addition to a baseline setting, we examine how model behavior changes under more realistic conditions, including the provision of information about users' past choices and persona-based prompting. Our results show that while meaningful WTP values can be derived for larger LLMs, they also display systematic deviations at the attribute level. Additionally, they tend to overestimate human WTP overall, particularly when expensive options or business-oriented personas are introduced. Conditioning models on prior preferences for cheaper options yields valuations that are closer to human benchmarks. Overall, our findings highlight both the potential and the limitations of using LLMs for subjective decision support and underscore the importance of careful model selection, prompt design, and user representation when deploying such systems in practice.
arXiv:2602.10384v4 Announce Type: replace Abstract: Vision-language models (VLMs) perform well on many document understanding tasks, yet their reliability in specialized, non-English domains remains underexplored. This gap is especially critical in finance, where documents mix dense regulatory text, numerical tables, and visual charts, and where extraction errors can have real-world consequences. We introduce Scribe Finance, the first multimodal benchmark for evaluating French financial document understanding. The dataset contains 1,204 expert-validated questions spanning text extraction, table comprehension, chart interpretation, and multi-turn conversational reasoning, drawn from real investment prospectuses, KIDs, and PRIIPs. We evaluate six open-weight VLMs (8B-124B parameters) using an LLM-as-judge protocol. While models achieve strong performance on text and table tasks (85-90% accuracy), they struggle with chart interpretation (34-62%). Most notably, multi-turn dialogue reveals a sharp failure mode: early mistakes propagate across turns, driving accuracy down to roughly 50% regardless of model size. These results show that current VLMs are effective for well-defined extraction tasks but remain brittle in interactive, multi-step financial analysis. Scribe Finance offers a challenging benchmark to measure and drive progress in this high-stakes setting.
arXiv:2606.17096v1 Announce Type: new Abstract: Thin wall treatments modify high-speed boundary-layer instability through the pressure they admit or absorb at the wall. This paper develops a unified admittance formulation for such effects on trapped compressible Rayleigh modes. For a simple rigid-wall eigenpair, we prove the spectral sensitivity law \[ c(A)=c_0+KA+\mathcal O(|A|^2), \qquad \delta\sigma=\alpha\Imag(KA)+\mathcal O(|A|^2), \] where \(A\) is the wall admittance and \(K\) is an explicit functional of the rigid-wall eigenfunction. The formula separates wall physics from outer-mode physics and yields a phase criterion for stabilisation. Matched asymptotics show that viscous and thermal wall layers, blind-pore coatings and shallow non-separating roughness all reduce to this same boundary condition, with additive leading admittances. Mach-4.5 computations validate the sensitivity coefficient and demonstrate porous damping, viscous-wall damping and sign-changing reactive roughness effects.
arXiv:2606.18055v1 Announce Type: new Abstract: MERMAID-v1 is a prototype PET scanner designed to support biomedical research involving adult zebrafish and similar species. The current experimental setup has been characterized, and scans of various phantoms, as well as adult zebrafish have been conducted. A dedicated reconstruction software was implemented, including accurate modeling of the parallax effect. The average energy resolution was 21.6% (FWHM at 511keV), with no significant dead-time effects observed for activities up to 18MBq. The absolute sensitivity at the center of the field of view (FOV) ranged from 0.06% to 0.31%, depending on the energy window (from 450-550 to 300-600keV), reflecting the limitations of the current two-head configuration. In the central 12mm of the transaxial FOV, the averaged spatial resolution is approximately 0.77mm (FWHM) transaxially and 0.66mm axially, as evaluated using a point source. Image quality was assessed using a downscaled NEMA-inspired IQ phantom and a 3D-printed Derenzo phantom. The reconstructed images suggest a spatial resolution around 0.7mm - 0.8mm, despite the lack of depth-of-interaction information. The first ex- and in-vivo PET scans of adult zebrafish were successfully performed, showing detectable tracer uptake in organs such as the brain and eyes despite low initial activity levels. These results confirm MERMAID-v1 capability to obtain useful results from the acquired data from living, anesthetized fish in a water-filled imaging chamber. While no scatter, attenuation, or efficiency corrections have yet been implemented, this work establishes a working proof-of-concept for dedicated PET imaging of small aquatic vertebrates. Future developments will focus on developing correction techniques, expanding the detector array, and integrating complementary modalities such as CT.
arXiv:2606.17905v1 Announce Type: new Abstract: Large language models perform increasingly well on standardized logical reasoning benchmarks, but whether this ability remains robust beyond English is unclear. We introduce ChLogic, an English--Chinese aligned benchmark that tests whether models preserve logical reasoning performance when the same latent logical structure is expressed in English and diverse Chinese surface realizations. Built from formal logical templates, the benchmark contains three data sets: (i) the General aligned set, derived from 60 General Propositions across nine template families; (ii) the Difficult aligned set, derived from 40 Difficult Problems; and (iii) the Chinese-only set, covering 15 language-specific phenomenon types. Each aligned item pairs one English reference expression with five Chinese realizations. Experiments on Qwen3, Ministral, and GLM models reveal a persistent English--Chinese performance gap. Back-translation from standard Chinese into English often improves performance on the General aligned set, but produces mixed effects on the Difficult aligned set, where Qwen3-32B and GLM-5.1 perform worse after translation. These results indicate that Chinese surface realization, translation artifacts, and model-specific behavior jointly affect multilingual logical reasoning. Overall, ChLogic provides a useful stress test for the robustness of multilingual reasoning.