Academic Intelligence · Curated Daily

Explore the Frontier of Global Academia

AcademicHub aggregates real-time literature from top journals and preprint platforms. Build your personal research radar and let large language models compile cross-disciplinary analysis briefings automatically.

01.
arXiv (math.PR) 2026-06-18

Denoising Distances in Metric Measure Spaces

arXiv:2606.18301v1 Announce Type: cross Abstract: Recent work studied the problem of finding clusters and denoising pairwise distances from noisy distances of points sampled on a manifold. We study the same problems in more general metric measure spaces under \lowerphiregularity{}. We give an algorithm that extracts large localized clusters around every sampled point and uses them to denoise distances to any fixed accuracy, with near-linear running time in the dense fixed-accuracy regime. We also show how to achieve much higher accuracy with a non-efficient algorithm. This suggests that unlike the Riemannian case, denoising to higher accuracy in more general metric spaces has a statistical-computational gap.

02.
arXiv (CS.LG) 2026-06-19

Adaptive Distance-Aware Trunk Deep Operator Learning for Long-Span Roadway Bridges

arXiv:2606.20015v1 Announce Type: new Abstract: Long-span roadway bridges exhibit highly localized structural responses under vehicular loading, making repeated FE analysis computationally expensive for applications such as influence surface generation and structural digital twins. Existing SciML approaches struggle to accurately capture these localized responses. To address this challenge, this study proposes an adaptive-trunk DeepONet for localized structural response prediction in large-scale bridge systems. The framework dynamically constructs a load-dependent learning domain using a KNN strategy, allowing the network to focus on structural influence zones. The trunk network is further enhanced using distance-aware features that encode the geometric relationship between the load and structural nodes. A physics-based full-field reconstruction is incorporated through a stiffness-informed Schur complement formulation, enabling predictions at adaptive nodes to be extended to the entire structural domain. To enable scalable training, response data are generated using a reduced-order equivalent shell model that preserves the dominant global behavior while significantly reducing computational cost. The proposed framework is validated on both a benchmark bridge model and the real-world Mussafah Bridge. Results show that the method achieves FEM-level accuracy with relative errors below 5%, while reducing the total response evaluation time (including full-field reconstruction) by approximately 60x; excluding the post-processing reconstruction step, the AD-DeepONet inference is up to four orders of magnitude faster than FEM. In addition, the framework enables rapid generation of full-field responses, influence lines, and influence surfaces under arbitrary vehicular loading configurations, demonstrating strong potential for large-scale bridge analysis and digital twin applications.

03.
arXiv (CS.CV) 2026-06-11

DAM-VLA: Decoupled Asynchronous Multimodal Vision Language Action model

Vision-language-action (VLA) models inherit a shared synchronous clock from vision-language pretraining, processing every input at one rate. This is misaligned with physical interaction, where a high-frequency modality changes at hundreds of hertz, vision evolves more slowly, and language stays constant across an episode. A synchronous VLA oversamples slow modalities, undersamples fast ones, and caps action generation at the lowest effective frequency. We hypothesize that decoupling temporal processing per modality, letting each update and retain information at its own sensor rate, yields stronger representations and more robust control. We present DAM-VLA, which maintains per-modality latent buffers refreshed at sensor rates and read continuously by the action head, integrating new high-frequency modalities through gated cross-attention that leaves the pretrained backbone intact. Across seven contact-rich real-world manipulation tasks, DAM-VLA more than doubles the average success rate of the strongest synchronous baseline (95.2\% vs.\ 40.95\%) while sustaining smooth, reactive 100\,Hz control. Project website: \href{https://intuitive-robots.github.io/DAM-VLA/}{intuitive-robots.github.io/DAM-VLA/}

04.
arXiv (CS.CL) 2026-06-12

Language Model Circuits Are Sparse in the Neuron Basis

The high-level concepts that a neural network uses to perform computation need not be aligned to individual neurons (Smolensky, 1986). Language model interpretability research has thus turned to techniques which decompose the neuron basis into more interpretable units of model computation, such as sparse autoencoders (SAEs). However, not all neuron-based representations are uninterpretable. For the first time, we empirically show that MLP neurons are as sparse a feature basis as SAEs. We use this finding to develop an end-to-end gradient-based attribution pipeline for circuit tracing on the MLP neuron basis, which surfaces causally effective neurons on a variety of tasks. On a standard subject-verb agreement benchmark (Marks et al., 2025), a circuit of $\approx 10^2$ MLP neurons is enough to control model behaviour. On the multi-hop city-state-capital task from (Lindsey et al., 2025), we find a circuit in which small sets of neurons encode specific latent reasoning steps (e.g. mapping a city to its state), and can be steered to change the model's output. This work thus advances automated interpretability of language models without imposing additional training costs.

06.
arXiv (CS.LG) 2026-06-18

Learning from Your Own Mistakes: Constructing Learnable Micro-Reflective Trajectories for Self-Distillation

arXiv:2606.18844v1 Announce Type: new Abstract: Self-distillation improves reasoning in large language models by using the model's own rollouts as training signal, typically through implicit logit-level alignment that minimizes KL divergence toward a privileged target distribution. However, because this supervision is generated via uncontrolled sampling, it provides no diagnostic insight into the model's specific errors or corrective guidance for its individual failure patterns. Consequently, the model learns to imitate a privileged distribution rather than receiving fine-grained corrections that pinpoint where and why its reasoning fails. In this paper, we propose Trajectory-Augmented Policy Optimization (TAPO), which advances self-distillation from implicit distributional alignment to explicit trajectory construction. During RL training, the model produces both correct and incorrect rollouts to the same query, and TAPO leverages this contrastive structure to construct micro-reflective corrections, new training trajectories that retain the model's erroneous reasoning up to the point of failure, then insert a natural-language diagnosis and corrected reasoning guided by a correct reference from the same sampling group. Since each trajectory is anchored in the learner's own prefix and solutions, the corrective signal preserves the model's on-policy distribution to a greater extent than the position-wise alignment imposed by KL-based methods. To integrate these trajectories, TAPO introduces difficulty-aware candidate selection at the model's capability boundary and decoupled advantage estimation to prevent gradient contamination. Experiments on AIME 2024, AIME 2025, and HMMT 2025 show that TAPO achieves consistent improvements over GRPO under the same number of training steps. Further analysis demonstrates that TAPO strengthens both first-pass reasoning and error-correction effectiveness.

07.
arXiv (CS.CV) 2026-06-19

Reliability-Aware Prototype Calibration for Frozen Pose-Flow Video Anomaly Detection

Pose-flow video anomaly detectors are attractive for one-class surveillance because they provide likelihood-based rankings for tracked skeleton windows. However, a single likelihood score may hide multimodal normal behavior and be sensitive to pose-observation noise. We study a frozen-detector setting in which the pose-flow backbone, cached skeleton tracks, and evaluation pipeline are fixed. Reliability-Aware Prototype Calibration (RPC) is a post-hoc score calibration method for this setting. It adds a standardized nearest-prototype deviation in the frozen latent space to the standardized flow score, and uses keypoint confidence only to gate this added geometric evidence. Thus, RPC preserves the original density signal while correcting the ranking with empirical normal-mode structure under pose reliability. Across two frozen pose-flow backbones and four datasets, RPC improves frame-level AUROC in all eight backbone-dataset pairs, with gains ranging from 0.34 to 4.49 percentage points and averaging 2.03 points. Ablation and reliability analyses show that prototype deviation is the main corrective signal, while reliability gating is most useful when pose observations are less trustworthy. These results suggest that lightweight post-hoc calibration can strengthen cached pose-flow systems when retraining or reproducing the full pose pipeline is impractical.

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

Vero: An Open RL Recipe for General Visual Reasoning

What does it take to build a visual reasoner that works across charts, science, spatial understanding, and open-ended tasks? The strongest vision-language models (VLMs) suggest that broad visual reasoning is within reach, yet their closed data and reinforcement learning (RL) pipelines make their gains difficult to study, reproduce, or extend. We introduce Vero, a family of fully open VLMs that match or exceed existing open-weight models across diverse visual reasoning tasks. We scale RL data and rewards across six broad task categories, constructing Vero-600K, a 600K-sample dataset from 59 datasets, and designing task-routed rewards that handle heterogeneous answers. Across VeroEval, our 30-benchmark suite, Vero-600K outperforms existing RL datasets under controlled comparisons. Applied to five starting models, Vero variants gain 2.9-5.4 points on average over their initial models. Notably, Vero-Qwen3I-8B, trained on the Instruct model, surpasses Qwen3-VL-8B-Thinking by 3.8 points on average without additional distillation. Systematic ablations reveal that different task categories elicit distinct reasoning patterns and that broad gains depend on learning them jointly rather than in isolation. All data, code, and models are publicly available.

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

ML Inference Scheduling with Predictable Latency

arXiv:2512.18725v3 Announce Type: replace Abstract: Machine learning (ML) inference serving systems can schedule requests to improve GPU utilization and to meet service level objectives (SLOs) or deadlines. However, improving GPU utilization may compromise latency-sensitive scheduling, as concurrent tasks contend for GPU resources and thereby introduce interference. Given that interference effects introduce unpredictability in scheduling, neglecting them may compromise SLO or deadline satisfaction. Nevertheless, existing interference prediction approaches remain limited in several respects, which may restrict their usefulness for scheduling. First, they are often coarse-grained, which ignores runtime co-location dynamics and thus restricts their accuracy in interference prediction. Second, they tend to use a static prediction model, which may not effectively cope with different workload characteristics. In this paper, we evaluate the potential limitations of existing interference prediction approaches, finding that coarse-grained methods can lead to noticeable deviations in prediction accuracy and that static models degrade considerably under changing workloads.

10.
arXiv (CS.AI) 2026-06-18

SWE-Future: Forecast-Conditioned Data Synthesis for Future-Oriented Software Engineering Agents

arXiv:2606.18733v1 Announce Type: cross Abstract: Realistic coding-agent benchmarks often replay public GitHub issues and pull requests, making them vulnerable to overlap with model pretraining, fine-tuning, synthetic-data generation, or benchmark-driven model selection. Fully synthetic tasks avoid direct historical replay, but can drift away from real repository needs. We propose SWE-Future, a forecast-conditioned data synthesis method for future-oriented coding tasks. Given a forecast snapshot at time $T_0$, the method uses only pre-$T_0$ repository evidence to forecast future feature implementation/enhancement, bugfix, and refactor task families. We first validate this forecasting step retrospectively: after forecasts are fixed, later pull requests are used only to measure whether the predicted task families match future repository work. In an 80-repository study, the forecaster achieves 58.1\% future-work relevance under the main semantic matching metric. We then use validated forecast families as conditioning signals to synthesize a 200-task coding-agent dataset across 61 repositories from a task-generation snapshot, rather than replaying the later pull requests used for validation. SWE-Future shows that repository-evolution forecasts can guide realistic, future-oriented coding-task synthesis while reducing direct dependence on historical pull-request replay.

11.
arXiv (CS.CL) 2026-06-16

Calibrated Triage, Not Autonomy: Confidence Estimation for Medical Vision-Language Models

A vision-language model can answer a question about a medical image fluently and confidently while barely using the image, leaning instead on language priors. In medicine this is the failure that matters most, because the answer looks trustworthy and is not, and the only protection is a confidence score reliable enough to tell the system when to abstain. We ask a deployment question rather than an accuracy one: how much imaging work a model can safely handle alone, and which confidence signal makes that possible. We evaluate seven confidence estimators across five open-weight LVLMs and three medical visual-question-answering datasets spanning broad clinical imaging, radiology, and pathology, with every probe trained only on natural images and applied without adaptation. Recast as bounded selective prediction (automate a case only when confidence clears a threshold, defer the rest), the comparison is cautionary. The standard metrics are poor guides: discrimination barely separates the methods, and the weak calibration of a cheap self-report is cheaply removed by off-domain temperature scaling without changing deployable yield. What distinguishes a usable estimator is the high-confidence region a clinician acts on: the weakest baselines are confidently wrong on 41 to 45 percent of their errors against 1 to 4 percent for the best probe, and no estimator is reliably best across domains or models. Safe handoff is governed at two levels: base-model competence sets a ceiling, so a well-calibrated score recovers roughly a third of radiology cases at a 20 percent error tolerance but almost none of pathology; the confidence layer then decides how much of that ceiling is reachable. The usable role today is calibrated triage, not autonomy: automate the cases a calibrated score marks safe, route the rest to a clinician. We release all outputs, correctness judgments, and confidence scores, with code.

12.
medRxiv (Medicine) 2026-06-22

Three multimodal large language models fail at clinically actionable breast pathology in three different directions

Background. Breast cancer treatment depends on histopathological features, such as grade and receptor-defined subtype; however, specialist pathologist access is constrained when the workforce is limited. Commercial multimodal large language models (MLLMs) accept hematoxylin and eosin (H&E) image tiles through paid interfaces without local hardware or fine-tuning. However, prior pathology evaluations addressed only coarse tasks. Whether they reach treatment-determining accuracy and whether vendors agree remain unclear. Methods. We aimed to evaluate three vendor-designated flagship MLLMs (Claude Sonnet 4.6, Gemini 2.5 Pro, GPT-5.5) in 427 invasive breast cancer cases. Each case went to all three with identical H&E tiles and prompts, and the subtype was inferred in the second call. The reference was an institutional sign-out report of an immunohistochemistry-derived subtype. We calculated the concordance, sensitivity, specificity, Cohen's kappa, and pairwise McNemar and Bowker tests. Findings. Claude ranked highest by raw histologic-type concordance but lowest by kappa, classifying all 23 lobular and seven micropapillary carcinomas as invasive breast carcinoma of no special type. The models anchored the Nottingham grade to three modal grades. None of the models reliably identified human epidermal growth factor receptor 2-positive disease. The failure direction was vendor-specific: Claude and GPT-5.5 were under-detected, whereas Gemini was over-called. Twelve prompt variants (4,056 calls) did not recover sensitivity. Interpretation. No current commercial MLLM reaches deployment-ready accuracy for any treatment-determining feature of breast pathology. As each vendor fails in its own fixed direction, changing vendors alters the type of error rather than removing it; therefore, the value of these models is assistive rather than autonomous. At USD 0.20-0.50 per case, they may serve as supervised draft generators that leave the diagnosis with the pathologist.

13.
arXiv (CS.CL) 2026-06-16

Extending Item Response Theory for Efficient and Meaningful Multilingual Evaluation

Multilingual benchmarks are central to evaluating large language models (LLMs) across languages, but they suffer from three issues: exhaustive evaluation scales linearly with the number of languages, automatic translation introduces errors that are easily missed at scale, and some items conflate general and culture-specific knowledge. We address all three with a unified statistical framework, Multilingual-IRT, which extends Item Response Theory with per-language difficulty deviations, split discriminability separating content from language effects, and per-language ability residuals. Fitting Multilingual-IRT on 25 LLMs across 29 languages of MMLU-Pro-X, we show that its fitted parameters support three practical applications: predicting unobserved (item, LLM, language) instances with 11-16% lower binary cross-entropy than the strongest accuracy-based baseline, surfacing candidate translation errors distributed across all 28 non-English languages, whereas accuracy-based baselines concentrate detections in a few languages, and recovering culture-specific items that accuracy-based baselines miss.

14.
arXiv (CS.CL) 2026-06-11

Neuron-based Personality Trait Induction in Large Language Models

Large language models (LLMs) have become increasingly proficient at simulating various personality traits, an important capability for supporting related applications (e.g., role-playing). To further improve this capacity, in this paper, we present a neuron-based approach for personality trait induction in LLMs, with three major technical contributions. First, we construct PersonalityBench, a large-scale dataset for identifying and evaluating personality traits in LLMs. This dataset is grounded in the Big Five personality traits from psychology and is designed to assess the generative capabilities of LLMs towards specific personality traits. Second, by leveraging PersonalityBench, we propose an efficient method for identifying personality-related neurons within LLMs by examining the opposite aspects of a given trait. Third, we develop a simple yet effective induction method that manipulates the values of these identified personality-related neurons. This method enables fine-grained control over the traits exhibited by LLMs without training and modifying model parameters. Extensive experiments validate the efficacy of our neuron identification and trait induction methods. Notably, our approach achieves comparable performance as fine-tuned models, offering a more efficient and flexible solution for personality trait induction in LLMs. We provide access to all the mentioned resources at https://github.com/RUCAIBox/NPTI.

15.
medRxiv (Medicine) 2026-06-22

Repeat expansions in Parkinson's disease and parkinsonism across ancestries: insights from a global genetic cohort

Expanded short tandem repeats contribute to a broad spectrum of neurodegenerative diseases, yet their roles in Parkinson's disease (PD) and parkinsonism remain incompletely characterized, especially across diverse ancestries. We analyzed short-read whole-genome (WGS) and clinical exome sequencing (CES) data from 38,365 individuals (28,861 WGS; 9,504 CES), encompassing 23,242 patients with PD, 4,729 patients with atypical parkinsonism and 10,394 healthy controls from 11 genetic ancestries. To determine carrier frequencies and characterize repeat structures across diverse ancestries, we genotyped 12 established pathogenic loci where normal, intermediate, and pathogenic alleles can be reliably differentiated using short-read sequencing data. Additionally, we conducted threshold-based associations to determine the minimum threshold associated with increased PD risk in 15,995 individuals (8,591 PD, 7,404 controls) of European ancestry. Pathogenic repeat expansions were detected in 62 patients (56 PD and 6 atypical parkinsonism) and 5 controls across seven loci (AR, ATXN1, ATXN2, ATXN3, CACNA1A, HTT and THAP11), spanning seven ancestries. Among these, ATXN2 expansions were the most frequently observed in PD and were present in African, East Asian, European and Middle Eastern ancestries. Additionally, intermediate ATXN2 repeat expansions exhibited a strong, length-dependent association with PD risk in the European population, with individuals with [≥]32 repeats having a more than four-fold increased risk (odds ratio 4.25, 95% confidence interval 1.80-12.05). Overall, >92% of expanded alleles harbor CAA interruptions within the CAG tract. Pathogenic expansions at other loci, such as ATXN3 and THAP11, showed more ancestry-specific distributions. Clinically, individuals with pathogenic ATXN2 and ATXN3 expansions most often presented with typical PD features but frequently showed earlier disease onset and a strong family history of PD. This large-scale, multi-ancestry study comprehensively maps the genetic landscape of pathogenic and intermediate repeat expansions in PD. Our findings confirm a length- and structure-dependent risk association for ATXN2 with PD in the European population, and highlight the pleiotropic effects of repeat expansions across the parkinsonian spectrum.

16.
arXiv (CS.AI) 2026-06-18

DN-Hypo-Pipeline: An AI-Driven Workflow for Hypothesis Generation via Large Language Models and Scientific Explanations

arXiv:2606.08532v2 Announce Type: replace Abstract: A scientific hypothesis is the first step in research and undergoes experimental validation, yet it also reflects a deep understanding of and reasoning about scientific phenomena. We introduce DN-Hypo-Pipeline, an AI-powered workflow based on large language models, designed to support structured scientific thinking and hypothesis generation by leveraging scientific explanations as prior knowledge. This pipeline assists researchers in deriving novel hypotheses from existing literature. Given the explanandum (i.e., the conclusion) of a research paper, it identifies underlying laws, theories, and principles, and reconstructs a new, yet-to-be-verified explanation for the observed phenomenon. We evaluated DN-Hypo-Pipeline in the field of data science modeling using three highly cited papers. Statistical inference, supported by both LLM-as-judge assessment and human expert evaluation, demonstrates that our pipeline is more effective than direct generation methods. Additionally, we validated the two highest-scoring generated hypotheses by developing corresponding novel algorithms, which outperformed the baseline models presented in the original papers. Beyond application in data science, DN-Hypo-Pipeline provides a theoretical framework that not only encompasses theory-guided data science modeling methods but also reveals a more fundamental structure of the modeling process. Moreover, this approach is essentially a generalization of theory-guided modeling, offering potential for extension to other domains and across a broader range of scientific disciplines.

17.
arXiv (CS.CL) 2026-06-11

ICA Lens: Interpreting Language Models Without Training Another Dictionary

Finding interpretable directions in language-model representations is critical for understanding and controlling model behavior. Sparse autoencoders (SAEs) have become the standard tool for this purpose, but using them as the default first lens often requires training, storing, and evaluating large overcomplete dictionaries. This bottleneck limits rapid exploration and raises a fundamental question: how much interpretable structure is already visible from activation geometry before training another neural dictionary? Our intuition is simple: many interpretable directions are selective on tokens, and these directions should look less Gaussian than random directions. We therefore revisit independent component analysis (ICA), a classical method for finding non-Gaussian directions, as a compact lens for language-model interpretability. We find that ICA has been underestimated for LLM interpretability, because prior uses often relied on off-the-shelf ICA implementations that are brittle on LLM activations and lacked systematic tools for inspecting and evaluating the recovered directions. To bridge these gaps, we introduce ICALens, the first practical workflow for stable, efficient, and auditable ICA analysis of LLM representations. It combines an optimized GPU-parallel FastICA pipeline with LLM-specific stability recipes and better fitting diagnostics, enabling efficient and reliable layer-wise analysis. Across GPT-2 Small, Gemma 2 2B, and Qwen 3.5 2B Base, ICALens efficiently recovers compact, human-interpretable directions without per-layer gradient-based dictionary training. On SAEBench, ICA is competitive with public SAEs in sparse probing and outperforms them in targeted probe perturbation under small-to-medium budgets. These results suggest that ICA should not be viewed as a weak baseline, but as an efficient and complementary first lens for exploring language-model representations.

18.
arXiv (CS.LG) 2026-06-19

Approximate Next Policy Sampling: Replacing Conservative Target Policy Updates in Deep RL

arXiv:2605.05481v2 Announce Type: replace Abstract: We revisit a classic "chicken-and-egg" problem in reinforcement learning: to safely improve a policy, the value function must be accurate on the state-visitation distribution of the updated policy. That distribution over states is unknown and cannot be sampled for the purposes of training the value function. Conservative updates solve this problem, but at the cost of shrinking the policy update. This paper explores an alternative solution, Approximate Next Policy Sampling (ANPS), which addresses the problem by modifying the training distribution rather than constraining the policy update. ANPS is satisfied if the distribution of the training data approximates that of the next policy. To demonstrate the feasibility and efficacy of ANPS, we introduce Stable Value Approximate Policy Iteration (SV-API). SV-API modifies the standard approximate policy iteration loop to hold the target policy fixed while an iteratively updated behavioral policy gathers relevant experience. It only commits to a new policy once a convergence criterion has been met. If certain stability criteria are met, the update is guaranteed to be safe; otherwise, it remains no less safe than standard approximate policy iteration. Applying SV-API to PPO yields Stable Value PPO (SV-PPO), which matches or improves performance on high-dimensional discrete (Atari) and continuous control benchmarks while executing substantially larger target policy updates. These results demonstrate the viability of ANPS as a new solution to this classic challenge in RL.

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

FraudSMSWalker: Benchmarking Agentic Large Language Models for SMS-to-Webpage Fraud Detection

SMS fraud is increasingly cross-channel: a message directs the user to a webpage, and the final risk depends on how the SMS claim aligns with the page content and requested user action. However, existing evaluations either focus on message-only smishing classification or expose URL and domain cues that allow models to rely on reputation shortcuts. To address this gap, we introduce FraudSMSWalker, a controlled benchmark for URL-masked SMS-to-webpage fraud judgment. FraudSMSWalker contains 699 bilingual chains, including 332 fraudulent and 367 benign cases, across ten service scenarios. The model-visible input consists of the SMS context and sanitized webpage evidence, while raw URLs, hosts, domains, IPs, redirects, and reputation metadata are withheld. The benchmark further includes hard benign cases whose pages contain login, payment, verification, or account-management elements that are plausible under the service context but also appear in scam flows. We evaluate nine web agents under masked browser-agent protocols and conduct URL-visibility ablations. The results show that current agents can detect suspicious cues, but struggle to preserve benign recall and often produce positive predictions that are weakly supported by the observed evidence. These findings position FraudSMSWalker as a benchmark for measuring whether web agents can make fraud judgments that remain both accurate and evidence-grounded when direct reputation shortcuts are suppressed. The associated code and dataset are accessible at the \href{https://anonymous.4open.science/w/FraudMessageWalker-Bench}{anonymous link}.

20.
arXiv (CS.CV) 2026-06-15

How do Self-Supervised Remote Sensing Vision Models Transfer to Downstream Tasks?

Self-supervised geospatial foundation models (GeoFMs) learn transferable representations from remote sensing data, but their downstream behavior is difficult to characterize. We study six representative GeoFMs spanning joint-embedding, reconstruction, and multimodal pretraining families, and evaluate transfer across classification, regression, and segmentation benchmarks under different label availability and downstream pipelines. We find that model rankings change across tasks and adaptation settings. Layerwise probing shows that, in most cases, task-relevant information is more accessible in intermediate transformer blocks compared to final-layer embeddings, and that GeoFMs exhibit distinct depthwise profiles. In segmentation case studies on PASTIS and Sen1Floods11, downstream adaptation settings such as decoder design and fine-tuning can be as impactful as the choice of GeoFM, and standard dense-prediction heads may be poorly aligned with how GeoFMs organize information over depth. Finally, CKA analysis on case studies shows that fine-tuning does not rewrite GeoFMs uniformly across depth, and the strongest changes are localized to the first linear layer of the MLP in ViT blocks. These results help explain why GeoFM rankings shift across benchmarks and motivate more representation-aware evaluation and adaptation strategies.

21.
arXiv (CS.AI) 2026-06-17

Constitutional On-Policy Safe Distillation

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.

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

TextResNet: Decoupling and Routing Optimization Signals in Compound AI Systems via Deep Residual Tuning

arXiv:2602.08306v2 Announce Type: replace Abstract: Textual Gradient-style optimizers (TextGrad) enable gradient-like feedback propagation through compound AI systems. However, they do not work well for deep chains. The root cause of this limitation stems from the Semantic Entanglement problem in these extended workflows. In standard textual backpropagation, feedback signals mix local critiques with upstream contexts, leading to Attribution Ambiguity. To address this challenge, we propose TextResNet, a framework that reformulates the optimization process to achieve precise signal routing via four key innovations. Firstly, in the forward pass, it enforces Additive Semantic Deltas to preserve an Identity Highway for gradient flow. Secondly, in the backward pass, it introduces Semantic Gradient Decomposition via a Semantic Projector to disentangle feedback into causally independent subspaces. Thirdly, it implements Causal Routing, which routes projected signals to their specific components. Finally, it performs Density-Aware Optimization Scheduling to leverage the disentangled signals to dynamically allocate resources to key system bottlenecks. Our results show that TextResNet not only achieves superior performance compared to TextGrad, but also exhibits remarkable stability for agentic tasks in compound AI systems where baselines collapse. Code is available at https://github.com/JeanDiable/TextResNet.

23.
arXiv (CS.CV) 2026-06-12

VISTA: An End-to-End Benchmark for Visual Spec-to-Web-App Coding Agents

We present VISTA (VIsual Spec-To-App Benchmark), a benchmark for evaluating the end-to-end web-app generation capabilities of LLM-based agents. Unlike prior code generation benchmarks that focus on algorithmic tasks, VISTA targets realistic UI-centric development, where agents must produce functional, visually coherent applications from underspecified inputs. We define five prompt-information conditions that vary along two axes, visual/structural fidelity and stack constraint: (1) text only with free stack choice, (2) text with reference screenshots under three specified stacks, (3) text with reference screenshots under free stack choice, (4) text with screenshots and pruned Figma structure under a single specified stack, and (5) text with screenshots and pruned Figma structure under free stack choice. To enable robust evaluation, each page in the benchmark is manually annotated with interactive UI components and around three visual anchor points, addressing the well-known limitations of script-based testing tools such as Playwright in open-ended code generation settings. Evaluation combines DOM-grounded reference matching, behavior-specific browser tests, and CLIP-based visual similarity, jointly measuring structural alignment, behavioral completeness, and overall visual fidelity. We use VISTA to assess four agent systems drawn from two model families and two harnesses, finding that visual fidelity and functional correctness are partially decoupled across both input conditions and agents, and that agent editing style varies sharply but is largely orthogonal to task quality. VISTA establishes a rigorous and reproducible foundation for advancing agent-based software engineering research.

24.
arXiv (CS.CL) 2026-06-15

EiCAP: Beyond Fluency, Probing and Improving Emotional Intelligence in LLMs via Psychologically Grounded Multi-Turn Dialogue

Large Language Models increasingly serve in emotionally sensitive roles, including mental health support, education, and crisis response, yet they lack a principled framework for assessing or improving Emotional Intelligence (EI). We introduce EiCAP, a unified, psychologically grounded six-layer EI taxonomy operationalized into two complementary resources. EiCAP-Bench is a multi-turn, one-vs-three forced-choice evaluation suite with 3,174 probes across 24 subcategories and cross-turn dependencies that reflect real conversational EI demands. EiCAP-SFT is a 152,820-dialogue supervision corpus aligned to the same taxonomy, enabling controlled, interpretable fine-tuning. Two key findings emerge. First, generic conversational supervised fine-tuning does not confer EI: fine-tuning on UltraChat yields no significant gain in any of the 24 subcategories, with a macro score of 24.6%, near the chance level of 25%. Second, applying EI-grounded LoRA, using approximately 0.8% of parameters, directly to Qwen-2.5-7B-Base achieves significant gains in all 24 subcategories, reaching a macro score of 75.33%, a gain of 51.7 percentage points over Base and 37.1 percentage points over Instruct. Crucially, an ablation shows that the UltraChat pre-stage is counterproductive, reducing performance by 21.4 percentage points: direct EI-grounded training is both necessary and sufficient.

25.
arXiv (math.PR) 2026-06-12

Characterizing metric-space-valued processes: separating classes and weak invariance principles for measure-theoretic inference

arXiv:2606.13084v1 Announce Type: cross Abstract: This article investigates stochastic processes taking values in metric spaces that lack a topological vector space structure, a regime characterized by intricate interplay between topological, geometric, and temporal dependence structures. It is formally established that spaces admitting an isometric Hilbertian embedding constitute a strict subclass within the much broader class of metric spaces possessing the ball property. While traditional kernel methods are susceptible to geometric distortion when the underlying space cannot be isometrically embedded into a Hilbert space, we bypass such limitations by exploiting a fundamental structural property inherent to this broader class; namely, that Borel probability measures are uniquely determined by their values on balls. These separating classes provide the foundation for the subsequently introduced measure-theoretic inference methodology. We derive uniform convergence of a family of time-dependent random measures, alongside weak invariance principles for the corresponding nonstationary random fields. This framework explicitly exposes how dependence and geometric complexity influence sample path regularity. Furthermore, because the rapid decay of small-ball probabilities can prohibit the existence of limiting distributions for supremum-based discrepancy measures, we develop $L^p$-based alternatives. By directly leveraging the introduced convergence results, this approach circumvents the need for higher-order $U$-process formulations. Finally, for spaces that do admit an isometric Hilbertian embedding, and where $U$-processes naturally arise, we establish limit theory for both degenerate and nondegenerate multi-parameter $U$-processes, and demonstrate that local discrepancy tests maintain asymptotic stability under dynamic parameter regimes.