Academic Intelligence · Curated Daily

探索全球前沿学术脉络

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

01.
arXiv (CS.CV) 2026-06-16

Are Neuro-Inspired Multi-Modal Vision-Language Models Resilient to Membership Inference Privacy Leakage?

In the age of agentic AI, the growing deployment of multi-modal models (MMs) has introduced new attack vectors that can leak sensitive training data in MMs, causing privacy leakage. This paper investigates a black-box privacy attack, i.e., membership inference attack (MIA) on multi-modal vision-language models (VLMs). State-of-the-art research analyzes privacy attacks primarily to unimodal AI-ML systems, while recent studies indicate MMs can also be vulnerable to privacy attacks. While researchers have demonstrated that biologically inspired neural network representations can improve unimodal model resilience against adversarial attacks, it remains unexplored whether neuro-inspired MMs are resilient against privacy attacks. In this work, we introduce a systematic neuroscience-inspired topological regularization (tau) framework to analyze MM VLMs resilience against image-text-based inference privacy attacks. We examine this phenomenon using three VLMs: BLIP, PaliGemma 2, and ViT-GPT2, across three benchmark datasets: COCO, CC3M, and NoCaps. Our experiments compare the resilience of baseline and neuro VLMs (with topological regularization), where the tau > 0 configuration defines the NEURO variant of VLM. Our results on the BLIP model using the COCO dataset illustrate that MIA attack success in NEURO VLMs drops by 24% mean ROC-AUC, while achieving similar model utility (similarities between generated and reference captions) in terms of MPNet and ROUGE-2 metrics. This shows neuro VLMs are comparatively more resilient against privacy attacks, while not significantly compromising model utility. Our extensive evaluation with PaliGemma 2 and ViT-GPT2 models, on two additional datasets: CC3M and NoCaps, further validates the consistency of the findings. This work contributes to the growing understanding of privacy risks in MMs and provides evidence on neuro VLMs privacy threat resilience.

02.
arXiv (CS.AI) 2026-06-11

Geometric Erasure by Contrastive Velocity Matching in Rectified Flows

arXiv:2606.00140v2 Announce Type: replace-cross Abstract: While the rapid adoption of multimodal generative models offers immense potential, it has also increased the risks of harmful content synthesis, deepfakes, and copyright infringements. To address these challenges, concept erasure has emerged as a prospective safeguard. However, as the field gradually transitions from U-Net-based diffusion models to Rectified Flow Transformers, erasure research has struggled to keep pace. In this work, we introduce GEM, a simple but highly effective erasure framework for Rectified Flow models. As part of our contribution, we establish a principled bridge between trajectory-based unlearning grounded in Generative Flow Networks and classic teacher-guided erasure: we translate trajectory-based signals into a teacher-guided flow-matching setup that unifies the strengths of both paradigms. Concretely, a teacher provides complementary attraction and repulsion signals that we combine into a single geometric guidance objective, yielding targeted suppression of unwanted concepts while preserving benign generation.

03.
arXiv (CS.LG) 2026-06-19

A deep learning framework for jointly solving transient Fokker-Planck equations with arbitrary parameters and initial distributions

arXiv:2604.06001v2 Announce Type: replace-cross Abstract: Efficiently solving the Fokker-Planck equation (FPE) is central to analyzing complex parameterized stochastic systems. However, current numerical methods lack parallel computation capabilities across varying conditions, severely limiting comprehensive parameter exploration and transient analysis. This paper introduces a deep learning-based pseudo-analytical probability solution (PAPS) that, via a single training process, simultaneously resolves transient FPE solutions for arbitrary multi-modal initial distributions, system parameters, and time points. The core idea is to unify initial, transient, and stationary distributions via Gaussian mixture distributions (GMDs) and develop a constraint-preserving autoencoder that bijectively maps constrained GMD parameters to unconstrained, low-dimensional latent representations. In this representation space, the panoramic transient dynamics across varying initial conditions and system parameters can be modeled by a single evolution network. Extensive experiments on paradigmatic systems demonstrate that the proposed PAPS maintains high accuracy while achieving inference speeds four orders of magnitude faster than GPU-accelerated Monte Carlo simulations. This efficiency leap enables previously intractable real-time parameter sweeps and systematic investigations of stochastic bifurcations. By decoupling representation learning from physics-informed transient dynamics, our work establishes a scalable paradigm for probabilistic modeling of multi-dimensional, parameterized stochastic systems.

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

Cloze: An Open Research Platform for Studying Human-AI Conversations in Mental Health Contexts

Cloze is an open-source web platform for conducting controlled, monitored studies of human-AI conversation in mental health research contexts. Consumer large language model (LLM) products such as ChatGPT, Claude, and Gemini are built for individual productivity, and offer researchers little experimental control, inconsistent data export, and no shared safety scaffolding that holds across providers. Cloze gives research teams a single environment in which they configure which models participants converse with, how the AI is instructed, how conversations are scheduled over time, and which safety constraints apply unconditionally, while every message is captured with full provenance (model version, prompt configuration, timing). The platform currently supports OpenAI, Anthropic, Google, and locally hosted open-weight models served through Ollama behind a unified interface, and runs in the cloud or fully on premises so that participant data need never leave an institution. Cloze is research infrastructure for building an evidence base on human-AI interaction in mental health contexts. It is not a therapeutic product.

05.
arXiv (CS.LG) 2026-06-17

Performance-Driven Environment Abstraction with Multi-Timescale Learning

arXiv:2606.17377v1 Announce Type: new Abstract: We study performance-driven environment abstraction for decision-making in large Markov decision processes. Rather than preserving geometric or topological structure, we seek abstractions that directly optimize decision quality. We model abstraction as a controlled approximation obtained by aggregating the state space and enforcing a shared action distribution within each aggregated state. For a fixed partition, we establish a performance guarantee that separates value-function approximation error from the loss introduced by action sharing. Guided by this analysis, we develop a multi-timescale reinforcement learning framework that jointly adapts the policy and a tree-structured environment abstraction. The resulting algorithm refines and coarsens regions of the state space based on Q-value discrepancies, balancing performance against abstraction size and complexity. Empirical results demonstrate substantial state compression, improved sample efficiency, and faster replanning compared to actor-critic baselines.

06.
Nature (Science) 2026-06-17

Molecular basis of polyadenylated RNA fate determination in the nucleus

作者:

Eukaryotic genomes generate a plethora of polyadenylated (pA+) RNAs1,2, which are packaged into ribonucleoprotein particles (RNPs). To ensure faithful gene expression, functional pA+ RNPs, including protein-coding RNPs, are exported to the cytoplasm, whereas transcripts within non-functional pA+ RNPs are degraded in the nucleus1–4. How cells distinguish these opposing fates remains unknown. The DExD-box ATPase UAP56 (also known as DDX39B) is a central component of functional pA+ RNPs, and promotes their docking to the nuclear pore complex-anchored TREX-25,6, which triggers transcript release from UAP56 to facilitate export7. Here we reveal that the poly(A) tail exosome targeting (PAXT) connection8 binds a TREX-2-like module, which releases pA+ RNAs from UAP56 for decay by the nuclear exosome. The core of this module consists of a LENG8–PCID2–SEM1 trimer, which we show is structurally and biochemically equivalent to the central GANP–PCID2–SEM1 trimer of TREX-2. Mutagenesis and transcriptomic data demonstrate that the nuclear fate of pA+ RNPs is governed by the contending actions of nucleoplasmic PAXT and nuclear pore complex-associated TREX-2, which interpret RNA-bound UAP56 as a signal for RNA decay or export, respectively. As RNA targets of PAXT are generally short and intron-poor, we propose an overall model for pA+ RNP fate determination whereby the distinct sub-nuclear localizations of PAXT and TREX-2 govern the degradation of short non-functional pA+ RNAs while allowing export of their longer and functional counterparts. Biochemical, structural and cell biological analyses reveal that UAP56 (DDX39B) assembles with a TREX-2–like module that redirects non-functional polyadenylated RNAs from export to degradation.

07.
arXiv (CS.CV) 2026-06-24

A Geometry-Informed Computer Vision Method for Detecting and Examining Overtaking Vehicles From A Bicycle

Instrumented bicycle studies have produced direct field evidence on vehicle passing behavior, but extracting overtaking events from continuous rear-facing video has remained dependent on manual, frame-by-frame annotation. This bottleneck constrains sample sizes and limits naturalistic cycling safety research. We present a geometry-informed computer vision pipeline that automates overtaking event detection from a single bicycle-mounted camera without multi-sensor configurations or explicit camera calibration. The system combines RT-DETR object detection with ByteTrack multi-object tracking through a three-stage geometric validation module enforcing bearing angle trend, apparent size growth, and spatial confirmation criteria derived from perspective projection principles. Validated on 315 manually annotated real-world overtaking events from urban roads in Ann Arbor, Michigan, the pipeline achieved 97.8% recall with zero false positives. The system identified overtaking intentions a mean of 2.44 seconds before vehicle passage, with 84.1% of events exceeding the 1.5-second human reaction time threshold, demonstrating feasibility for active cyclist warning. Lateral passing distance measurements from 96 events revealed 33.3% of passes below the 5-foot (152.4 cm) threshold, consistent with non-compliance rates in prior field and self-reported studies. A preliminary calibration-free lateral distance estimation approach using bounding box geometric features achieved mean absolute errors of 13-14 cm under leave-one-out cross-validation, sufficient to distinguish close passes from standard passes for safety categorization. By automating event isolation from consumer-grade footage, the system removes the primary annotation bottleneck of instrumented bicycle research and provides a scalable foundation for vehicle-bicycle interaction analysis across larger datasets and diverse urban environments.

08.
arXiv (CS.CL) 2026-06-17

Atlas: Orchestrating Heterogeneous Models and Tools for Multi-Domain Complex Reasoning

The integration of large language models (LLMs) with external tools has significantly expanded the capabilities of AI agents. However, as the diversity of both LLMs and tools increases, selecting the optimal model-tool combination becomes a high-dimensional optimization challenge. Existing approaches often rely on a single model or fixed tool-calling logic, failing to exploit the performance variations across heterogeneous model-tool pairs. In this paper, we present ATLAS (Adaptive Tool-LLM Alignment and Synergistic Invocation), a dual-path framework for dynamic tool usage in cross-domain complex reasoning. ATLAS operates via a dual-path approach: (1) training-free cluster-based routing that exploits empirical priors for domain-specific alignment, and (2) RL-based multi-step routing that explores autonomous trajectories for out-of-distribution generalization. Extensive experiments across 15 benchmarks demonstrate that our method outperforms closed-source models like GPT-4o, surpassing existing routing methods on both in-distribution (+10.1%) and out-of-distribution (+13.1%) tasks. Furthermore, our framework shows significant gains in visual reasoning by orchestrating specialized multi-modal tools.

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

Prediction of Runtime Parameters of Parallel Chemistry Applications via Active and Generative Learning

arXiv:2606.16226v1 Announce Type: new Abstract: In this work, we develop two main Machine Learning based approaches to predict the runtime parameters of highly scalable parallel chemistry computations.These approaches employ active and generative learning together with the empirically determined gradient boosted regression tree models chosen among a rich suite of machine learning models. When evaluated on Coupled-Cluster with Singles and Doubles computations, our models achieve a mean absolute error percentage (MAPE) as low as 0.023 and a coefficient of determination as high as 99.9%. Furthermore, when combined with active learning to mitigate the lack of large amounts of training data, our models score a MAPE about 0.2 with 20-25% of the original dataset.

10.
arXiv (quant-ph) 2026-06-19

Robust Generation of Topological Biphoton Mode via Adiabatic Passage

arXiv:2606.19786v1 Announce Type: new Abstract: Topological waveguide arrays support robust mode propagation in the presence of fabrication imperfections, providing a significant advantage for on-chip quantum information processing. However, this robustness does not fully extend to nonlinear biphoton generation. Structural disorder can enhance the excitation of non-topological biphoton modes during nonlinear interactions, which degrades the quantum properties of the generated state. To overcome this limitation, we propose an adiabatic passage that connects an isolated site to a topological defect array. By initiating the nonlinear process in a strongly isolated regime, nonlinear coupling to unwanted modes is effectively suppressed, thereby preserving the Schmidt number of the generated state. The subsequent adiabatic connection facilitates the high fidelity transfer of the generated biphoton into the topological biphoton mode. Our numerical simulations demonstrate that, unlike conventional topological structures, the adiabatic scheme maintains both high biphoton fidelity and a unit Schmidt number in the presence of waveguide gap disorder. Furthermore, we show that this robustness extends to path entangled NOON states, achieving a near-unity quantum interference visibility. Our approach provides a practical design strategy for disorder-tolerant integrated quantum photonic devices.

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

FLaRA: Predicting Future Latent Representations for Accident Anticipation

Anticipating traffic accidents from dashcam videos is a critical challenge in intelligent transportation systems. Existing methods typically map visual context directly to a collision probability without explicitly modeling the future evolution of the driving scene. In this paper we propose FLaRA (Predicting Future Latent Representations for Accident Anticipation), a novel predictive architecture that shifts this paradigm by forecasting future latent representations for accident anticipation. Building upon the Video Joint-Embedding Predictive Architecture (V-JEPA2), our model conditions a predictor network on observed context frames to predict the forthcoming latent features of the scene. A classifier then operates on these predicted future representations rather than only on past observations. To ensure these forecasts remain grounded in realistic future dynamics, we introduce a joint training objective that simultaneously optimizes an auxiliary feature-level reconstruction loss and a cross-entropy classification loss. Extensive evaluations on the Nexar dataset, alongside cross-domain validations on the DAD, DADA-2000, and DoTA benchmarks, demonstrate that our approach achieves state-of-the-art performance while maintaining realistic early warning capabilities.

12.
medRxiv (Medicine) 2026-06-11

Computer Vision for Real-Time Anatomical Navigation in Neurosurgery: First-in-Human Clinical Evaluation and Iterative Development (IDEAL Stage 1)

Introduction: Precise anatomical navigation is fundamental to safe endoscopic pituitary surgery, a high-stakes procedure characterised by a challenging learning curve. While traditional navigation systems often rely on workflow-disrupting probes or static preoperative imaging, advancements in computer vision AI (CVAI) now enable dynamic, real-time anatomical segmentation directly from live surgical video1-3. Our group has previously conducted a series of preclinical human-computer interaction studies to refine the system's design, alongside digital and high-fidelity physical simulations demonstrating the benefit of AI assistance in improving overall performance, training, and safety4-8. Building on this foundation, the current study represents a first-in-human application of real-time CVAI assistance in the neurosurgical operating room, serving to assess feasibility and safety, and to iteratively improve the system. Method: Guided by DECIDE-AI and IDEAL frameworks, this single-centre evaluation comprises an initial proof-of-concept phase (n=6) for endoscopic transsphenoidal pituitary surgeries. The AI model utilised a DINOv3-derived vision transformer architecture, deployed via a high-performance edge computing unit to achieve low-latency, real-time inference without reliance on cloud infrastructure2. Given the high-risk nature of the procedure and the early stage of clinical AI integration, the system was initially deployed as an educational adjunct on a secondary monitor, ensuring the primary surgical feed remains uncompromised. Functionality and safety were assessed via structured questionnaire, prospective observation, and blinded retrospective review of the recordings of the endoscopic surgical video feed and wider operating room environment. Continuous multi-stakeholder feedback through validated human factors surveys drove iterative technical refinements between cases. Results: Six patients with pituitary adenomas were enrolled. The CVAI system was successfully deployed in four cases, demonstrating acceptable real-time sella segmentation accuracy. Deployment failed pre-operatively in two cases owing to a single recurring system reboot bug. Iterative refinement between cases were driven by our experience and surgical team feedback. This resulted in the integration of additional anatomical structure segmentations (e.g., carotid arteries), enhanced model accuracy via training dataset expansion, and hardware firmware upgrades. Multi-stakeholder surveys demonstrated satisfactory system feasibility, usability, and acceptability among the surgical team. Both prospective observation and retrospective video review confirmed the absence of adverse events, including no significant distraction to the primary surgeon, and there were no AI-related clinical complications. Conclusion: This first-in-human early clinical evaluation demonstrates the feasibility, safety and iterative development of real-time, CVAI-based anatomical navigation during high-stakes neurosurgery. Future work will include a larger single-centre case series (IDEAL Stage 2a) with more surgical teams to further iterate the system and explore its impact on training and workflow. As the underpinning technology improves, deployment will transition to direct intra-operative decision support and integration with other intra-operative navigational technologies.

13.
arXiv (CS.CV) 2026-06-11

IB-HFN: Information Bottleneck-Driven SAR-Optical Fusion Network for High-Fidelity Cloud Removal

Synthetic aperture radar (SAR)-assisted optical cloud removal aims to recover surface information obscured by clouds in optical remote sensing images by exploiting complementary SAR observations. Existing multimodal fusion methods typically rely on direct spatial concatenation and pixel-wise supervision, which can propagate SAR speckle noise into optical reconstruction and lead to over-smoothed results. To address these limitations, we propose an Information Bottleneck-driven High-Fidelity Network (IB-HFN) for SAR-assisted optical cloud removal. IB-HFN employs a dual-stream backbone to preserve modality-specific representations before deep semantic fusion, thereby mitigating premature cross-modal contamination. At the fusion stage, we introduce a Spatial Information Bottleneck Fusion module that compresses SAR features through a channel-wise variational information bottleneck to suppress unstructured speckle noise. In parallel, a local-global gating mechanism predicts clear-sky regions and routes reliable optical details through a Dirac-initialized skip connection, decoupling noise suppression from texture preservation. We further develop a joint optimization strategy that integrates feature-level bottleneck regularization with image-level constraints on reconstruction accuracy, structural consistency, spectral fidelity, and contrastive sharpness. A dynamic weighting schedule balances these objectives to stabilize training and reduce hazy artifacts. Experiments on the SEN12MS-CR dataset under challenging spatio-temporal splits demonstrate that IB-HFN achieves superior structural preservation and spectral fidelity over existing methods.

14.
arXiv (CS.CL) 2026-06-17

RubricsTree: Scalable and Evolving Open-Ended Evaluation of Personal Health Agents across Health Memory and Medical Skills

The LLM-empowered personal health agents with user health (sensor) metrics have offered a promising pathway to alleviate global disparities in healthcare access. However, large-scale clinical deployment remains constrained by an open-ended evaluation bottleneck: physician annotation is reliable but costly and unscalable, while LLM-as-a-judge evaluators are scalable but subjective, inconsistent, and sometimes clinically misaligned. We introduce RubricsTree, a scalable evaluation framework with an expert-aligned hierarchical taxonomy of over 100 atomic, clinically-verifiable Boolean rubrics, evolving from the insights of 4,000 real user queries through an iterative human-in-the-loop curation protocol with an expertise panel led by an experienced physician. A context-aware adaptive router activates only the relevant auto-weighted rubric subset per query, providing the throughput needed for scalable evaluation with expert-aligned quality. Through a systematic meta-evaluation, we show that RubricsTree (i) substantially exceeds a strong large-scale evaluation baseline in expert alignment on challenging open-ended queries; (ii) reliably penalizes contextually degraded responses; and (iii) when used as structured instructions, text feedback, or training rewards for performance optimization, yields up to ~66% relative gains on HealthBench for Gemini, GPT, and Qwen model families. RubricsTree thus provides a scalable, auditable, and evolving evaluation infrastructure required for the continuous optimization of product-level personal healthcare AI.

15.
arXiv (CS.CV) 2026-06-18

BindEdit: Taming Attention Leakage for Precise Multi-Object Image Editing

Real image editing enables precise manipulation of visual content, yet existing methods often fail in complex multi-object scenarios, causing semantic blending, object duplication, or incomplete edits. We attribute these failures to attention leakage, where signals across spatial regions and text tokens become entangled during the denoising process. Specifically, we identify two distinct forms of leakage: Edit-Token Leakage, where ambiguous token-region alignment leads to object blending, and Source Dominance Leakage, where tokens of unchanged source objects overwhelm the attention intended for target entities. To resolve these leakages, we propose BindEdit, which enforces attention-level constraints within a single diffusion trajectory. To suppress Edit-Token Leakage, BindEdit jointly regularizes cross- and self-attention so that each target token group is bound to its corresponding spatial region while maintaining instance-level separation. To suppress Source Dominance Leakage, a cross-attention re-balancing mechanism amplifies target token influence and attenuates residual source semantics within editable regions. Moreover, a region fidelity term ensures that each target concept is expressed coherently across the entire editing mask. Additionally, we propose a comprehensive multi-object benchmark encompassing diverse object counts and categories. Extensive experiments demonstrate that BindEdit consistently outperforms existing methods within a single diffusion trajectory, maintaining robust performance across both single- and multi-object editing scenarios.

16.
arXiv (CS.AI) 2026-06-12

Learning What to Remember: A Cognitively Grounded Multi-Factor Value Model for Agentic Memory

arXiv:2606.12945v1 Announce Type: new Abstract: Long-running LLM agents accumulate interaction histories far larger than any context window, forcing a standing decision: what to encode deeply, what to forget, and what to retrieve under a fixed memory budget. Production systems answer with semantic similarity or recency – both mis-specified for the forgetting decision, which is made at consolidation time before the future query is known. We propose a multi-factor memory value function V(m)=\sum_i w_i f_i(m) over seven interpretable factors (emotional intensity, goal relevance, value alignment, self/user relevance, task utility, reliability, and usage history) drawn from cognitive psychology, whose weights are learned from a downstream objective by a gradient-free optimiser, and whose single scalar uniformly controls encoding depth, forget risk, and retrieval rank. We make a methodological point: on LongMemEval, scoring goal relevance against the held-out evaluation question saturates gold-evidence retention at \approx 0.98 – this measures retrieval, not forgetting. In the realistic blind regime, a learned multi-factor value retains 0.770 \pm 0.011 of gold evidence across 479 usable cases, versus 0.657 for uniform weights, 0.518 for the best single factor, and 0.368 for recency; every paired gap's 95% bootstrap CI is above zero, and a neural network over the same factors ties the linear model. The learned weights are interpretable – reliability, emotional intensity, and self/user relevance dominate, while query-time goal similarity is correctly down-weighted for the forgetting decision. A controlled synthetic task with planted confounds confirms the learner recovers a separating weighting (1.00 retention) where uniform weighting fails (0.62). The substrate is open-source; all experiments run on a single CPU with no API calls.

17.
bioRxiv (Bioinfo) 2026-06-22

When Less Is Not More: DICEPro Mitigates the Impact of Incomplete Reference Matrices on Cellular Frequency Deconvolution.

Cellular deconvolution aims to estimate the frequencies of different cell populations from gene expression measurements in a biological sample. Supervised approaches, such as CIBERSORTx and DISSECT, critically depend on the reference signature matrix, which encodes the gene expression profiles of cell-types based on prior knowledge. Despite numerous deconvolution methods, the impact of missing cell populations in the reference matrix remains understudied. Here, we evaluate the robustness of state-of-the-art deconvolution approaches using simulations based on real dataset examples combined with statistical modeling, validated against published data, and multiple real benchmark datasets. Results show that deconvolution performance remains stable when the reference matrix includes most cell-types, but declines sharply as the matrix becomes incomplete, especially for abundant cell populations. To address the limitations of incomplete reference matrices, we introduce DICEPro, an optimization-based framework designed to enhance existing deconvolution methods. By systematically adjusting the reference signatures, DICEPro better accounts for missing or underrepresented cell populations, leading to improved precision and robustness. We show that DICEPro consistently boosts deconvolution performance across both simulated datasets, derived from real data examples, and multiple real biological datasets, offering a practical solution when standard methods are hindered by incomplete references.

18.
arXiv (quant-ph) 2026-06-16

Scalable generation of heralded single photons via active feed-forward switching of a fiber delay line

arXiv:2606.16741v1 Announce Type: new Abstract: Quasi-deterministic single-photon generation is a key requirement for many photonic quantum technologies. Photon sources based on spontaneous parametric down-conversion (SPDC) are widely used for producing high-quality photons; however, the probabilistic nature of the process limits the generation of synchronized multi-photon states. Here, we demonstrate temporal synchronization of multiple photon-generation events using a free-space-fiber hybrid delay line with feed-forward control, enabling fast and efficient switching and scalable operation. Narrow-band, telecom-wavelength photons compatible for fiber transmission are heralded from a monolithic cavity SPDC source and synchronized across 20 time bins. This yields a sixfold enhancement in synchronized rates and enables multi-photon synchronization, with only a marginal increase of higher-order photon-number contributions.

19.
arXiv (CS.CL) 2026-06-24

Are LLM Evaluators Really Narcissists? Sanity Checking Self-Preference Evaluations

Recent research has shown that large language models (LLMs) favor their own outputs when acting as judges, undermining the integrity of automated post-training and evaluation workflows. However, it is difficult to disentangle which behaviors are explained by narcissism versus experimental confounds. Specifically, LLM evaluators may deliver self-preferring verdicts when comparing responses to questions they fail on; these verdicts may not depend on the identity of the author, but on evaluator quality. We correct this by directly comparing the judge's voting distribution in cases where it evaluates itself versus another model. This evaluator quality baseline reveals that only 51% of examples in previous findings retain statistical significance against this null hypothesis, covering 89.6% of total self-preference probability mass. Finally, we compare the entropy of voting distributions, suggesting uncertainty-driven overlap, and show that our procedure enables more careful documentation against the backdrop of judge-bias research.

20.
medRxiv (Medicine) 2026-06-15

Diabetes and the Life-Course: Evidence from Panel Data and Electronic Health Records

Incidence of type 2 diabetes is increasing at ages when education, work, family, and financial transitions are taking place, yet we lack robust evidence of whether earlier treatment changes life-course outcomes and over which time span this takes place. This paper uses the medical cutoff for diabetes diagnosis (HbA1c of 6.5 percent) as a natural experiment to study the effects of diabetes treatment using electronic health records (EHR) and panel data. This paper has three main findings. First, using EHR data, we find that there is a sharp increase in the probability of both diagnosis of diabetes and prescription when the HbA1c equals 6.5 percent. Second, we find that treating diabetes reduces HbA1c levels, weight, BMI, and blood pressure and increases the amount of care received, proxied by the number of HbA1c tests. Both the diagnosis and a prescription are independently able to produce positive changes in metabolic health, although a prescription is more effective in this regard. Third, we conclude that treating diabetes does not have a significant effect on life-course outcomes for a cohort of young Americans aged 24-32, although it does result in a reduction in HbA1c levels that are seen even eight years after the intervention. Taken together, these findings suggest that receiving a diagnosis and prescription are both effective treatments for diabetes, but they do not translate to significant alterations in the lives of young adults in the medium-term.

21.
arXiv (CS.AI) 2026-06-12

Creating and Evaluating K-12 GenAI Assessment Graders Through Context Engineering

arXiv:2606.12422v1 Announce Type: cross Abstract: The integration of large language models (LLMs) into educational assessment represents a transformative shift in classroom grading practices. While automated scoring systems and machine learning techniques have existed for decades, generative AI (GenAI) now enables educators to implement standards-based grading (SBG) with unprecedented efficiency and scale. This paper examines the theoretical foundations and evaluates an LLM grader that uses commercially available foundation models with context and prompt engineering to score student work against a rubric. Drawing on an empirical interrater agreement study using Massachusetts Comprehensive Assessment System (MCAS) data, we observed the Quadratic Weighted Kappa (QWK) and Proportional Reduction in Mean-Squared Error (PRMSE) across mathematics, science, and ELA, using Claude Sonnet 4, Haiku 4.5, GPT-5, and GPT-5 Mini. The results demonstrate that LLM graders, especially when based on foundational models with more parameters, achieve substantial agreement with human raters in mathematics and science assessments, while the performances vary in ELA, suggesting generic foundation models can be effective at scoring in given contexts. Additional analysis of teacher and student feedback reveals strong acceptance of AI-generated narrative feedback but skepticism toward numerical scores, suggesting that LLMs function most effectively as formative tools rather than summative evaluators. Our findings indicate that thoughtfully designed hybrid models that combine AI efficiency with teacher judgment can reduce workload, enhance feedback quality, and support equitable assessment practices without displacing professional expertise.

22.
arXiv (CS.CL) 2026-06-18

Learning User Simulators with Turing Rewards

Learning to simulate human users in interactive settings could advance the training of agent assistants, evaluation of personalization systems, research in the social sciences, and more. Existing approaches generally do so by training a large language model (LLM) to match a single ground truth response, either by maximizing the log probability or by using a similarity reward. We instead propose {Turing-RL}: a Turing-Test-based reinforcement learning approach for training user simulator models. {Turing-RL} uses a discriminative Turing reward with an LLM judge to score how indistinguishable a generated response is from the real user's given the user's history, and the user simulator LLM learns to produce responses indistinguishable from what the user could have said with such rewards. Across two different domains–conversational chat and Reddit forum discussion–we find that {Turing-RL} consistently outperforms baseline methods on both LLM and human evaluation metrics. Our study suggests that optimizing for indistinguishability, rather than response matching, is effective for learning user simulators.

23.
arXiv (CS.CL) 2026-06-16

Emergent retokenization symmetry in large language models: phenomenology and applications

Tokenization introduces representational redundancy: under a fixed token vocabulary, every byte string admits many valid token encodings, or segmentations, that decode to the same surface string. However, given a prompt, most language model tokenizers break this representational symmetry by returning a canonical segmentation. Training only on canonical segmentations should influence inference behavior, and there is little reason to expect models to respect segmentation symmetry on downstream tasks. We find that this symmetry partially emerges during training. Here, we probe this emergent symmetry through experiments testing token compositional understanding, representation diversity, and task focused benchmark performance. We primarily use retokenization – replacing a prompt's canonical tokenization with an alternative segmentation while preserving its bytes exactly. Relative to other prompt perturbations, retokenization is unusually clean because it isolates segmentation effects without changing syntax, semantics or surface form. We use retokenization to study sensitivity and robustness to semantically identical input representations across pretraining and post-training. Moreover, this partial retokenization symmetry suggests a distinct inference-time sampling axis. While temperature sampling generates diverse outputs from the model using its next-token probability distribution, retokenization generates diversity from the model's internal computations through semantically equivalent input representations. We find that while this retokenization sampling strategy can hurt performance on easy problems, it can also recover solutions that conventional sampling does not find. Overall, our work presents retokenization as a simple yet powerful probe of large language models, shedding light on compositional understanding and prompt sensitivity, and offering a novel sampling strategy.

24.
arXiv (CS.AI) 2026-06-24

E-MRL: Cross-view Aligned Evidence-driven Multimodal Reinforcement Learning for Reliable 3D Tumor Analysis

arXiv:2606.23888v1 Announce Type: cross Abstract: While Vision-Language Models (VLMs) show great promise in volumetric medical report generation, they frequently suffer from visual hallucinations and a lack of grounding in 3D CT data. Current Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) strategies typically optimize text fidelity alone, essentially rewarding correct diagnoses derived from language priors rather than genuine visual perception. To address this, we propose cross-view aligned Evidence-driven Multimodal Reinforcement Learning (Evidence-MRL, noted as E-MRL), a reliable RL reasoning framework that formulates the generation process as a Markov Decision Process of "diagnosis-localization-verification". Unlike standard approaches, our model is explicitly trained to identify a "key evidence slice" alongside the global diagnostic report, grounding its findings in verifiable visual evidence. Crucially, we introduce a novel cross-view consistency reward, which validates the semantic alignment between the golden-standard report and a local visual re-query of the selected key slice, providing additional rewards for correctly-localized reasoning. Experiments on large-scale 3D CT tumor datasets demonstrate that E-MRL significantly reduces hallucinations and improves diagnostic accuracy compared to SFT and RL baselines, offering a clinically interpretable solution for visually-grounded and tumor analysis.

25.
arXiv (CS.AI) 2026-06-24

DTT-BSR+: A Generative-Regression Cascade for Music Source Restoration

arXiv:2606.24127v1 Announce Type: cross Abstract: Music source restoration (MSR) requires jointly addressing source unmixing and the inversion of non-linear production effects. Current methods struggle to achieve accurate target signal reconstruction while maintaining semantic consistency. To address this limitation, we propose DTT-BSR+, a two-stage cascade MSR system that decouples distribution fitting from signal reconstruction into separate stages. A generative DTT-BSR separator in the first stage produces stems matching the prior of clean sources, and a modified Demucs network in the second stage enhances the first stage output using time-domain and multi-resolution spectral losses. DTT-BSR+ improves multi-mel signal-to-noise ratio (MMSNR) over the single-stage DTT-BSR across all stems, and surpasses the state-of-the-art X-LANCE MSR system on five stems. We also reveal through Fréchet Audio Distance (FAD) decomposition an implicit trade-off between signal reconstruction accuracy and semantic distribution fitting across stems.