×

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。

作者: Gen Li ×
换一批
01.
arXiv (CS.CV) 2026-06-16

FireRed-Image-Edit-1.0 Technical Report

We present FireRed-Image-Edit, a diffusion transformer for instruction-based image editing that achieves state-of-the-art performance through systematic optimization of data curation, training methodology, and evaluation design. We construct a 1.6B-sample training corpus, comprising 900M text-to-image and 700M image editing pairs from diverse sources. After rigorous cleaning, stratification, auto-labeling, and two-stage filtering, we retain over 100M high-quality samples balanced between generation and editing, ensuring strong semantic coverage and instruction alignment. Our multi-stage training pipeline progressively builds editing capability via pre-training, supervised fine-tuning, and reinforcement learning. To improve data efficiency, we introduce a Multi-Condition Aware Bucket Sampler for variable-resolution batching and Stochastic Instruction Alignment with dynamic prompt re-indexing. To stabilize optimization and enhance controllability, we propose Asymmetric Gradient Optimization for DPO, DiffusionNFT with layout-aware OCR rewards for text editing, and a differentiable Consistency Loss for identity preservation. We further establish REDEdit-Bench, a comprehensive benchmark spanning 15 editing categories, including newly introduced beautification and low-level enhancement tasks. Extensive experiments on REDEdit-Bench and public benchmarks (ImgEdit and GEdit) demonstrate competitive or superior performance against both open-source and proprietary systems. To support future research, our code, models, and benchmark suite are publicly available at https://github.com/FireRedTeam/FireRed-Image-Edit/ .

02.
arXiv (CS.CL) 2026-06-15

SANA: What Matters for QA Agents over Massive Data Lakes?

Exploratory question answering (EQA) over data lakes requires an LLM agent to discover relevant sources, analyze retrieved data, and adapt its actions based on intermediate results. End-to-end accuracy alone cannot distinguish failures in search, planning, data analysis, or the agent's Action Policy: its decisions about what to do next and when to submit an answer. We present SANA (Search Agent Navigation Ablation framework), a diagnostic ablation framework that transforms EQA tasks into runtime profiles containing gold source sequence, sanitized subquestions, and execution records. SANA uses these profiles to construct idealized search, planning, and data-analysis tools, allowing each component to be ablated; the residual gap is diagnostic evidence for policy failures. To illustrate SANA as a reusable evaluation framework, we adapted two recent EQA benchmarks, LakeQA and KramaBench, and evaluated lightweight and mid-sized agents under fixed prompts, budgets, data lakes, and runtimes. Across both benchmarks, data analysis is a consistent bottleneck while planning is less so. Search is a major limitation in LakeQA's large data-lake setting, but less so for the smaller-scale KramaBench. SANA thus deconstructs end-to-end task accuracies into a diagnosis of where data-lake agents fail, and allows for systematic comparisons of progress in search, planning, data analysis, and agent design.

03.
medRxiv (Medicine) 2026-06-22

Agentic Artificial Intelligence for Hospital Readmission Review: A Single-Center Blinded Evaluation and Exploratory Qualitative Analysis

Background: Manual review of 30-day hospital readmissions can identify actionable quality and safety problems, but it is labor-intensive. We developed and evaluated an agentic AI workflow for evidence-grounded readmission review. Materials and methods: We studied adult patients with unplanned 30-day readmission after discharge from a medicine hospitalist service at a single academic health system. An AI agent using a large language model queried a database containing notes, encounters, procedures, laboratory results, and other clinical data, and completed the same structured readmission-review rubric used by physicians. In the primary comparative evaluation, 20 randomly selected readmissions from 2025 were each reviewed by two physicians and the AI system. Blinded physician evaluators rated review quality. After rubric refinement, the AI workflow was applied to 100 recent readmissions in an exploratory expanded-cohort analysis of recurring improvement opportunities. Results: In the primary comparative evaluation, the AI classified 9/20 readmissions (45%) as preventable, compared with 19/40 physician reviews (47.5%). Blinded overall quality ratings were similar for AI and physician reviews (4.35 vs. 4.20 on a 1-5 scale; mean difference 0.15, 95% CI -0.20 to 0.48; p=0.49), as were factuality/support and usefulness/actionability ratings. No AI hallucinations were identified during factuality review. Agreement on preventability and primary readmission category was low for both AI-human and human-human comparisons. The AI system cost $0.23 per chart; physician reviewers took a median of 15 minutes, corresponding to an estimated $42.43 per chart. In the exploratory expanded-cohort analysis, AI-assisted review identified recurring vulnerabilities in post-discharge follow-up plans, incomplete inpatient workups, medication-safety transitions, and indwelling-device transitions. Conclusions: Agentic AI produced readmission reviews with similar blinded quality ratings to physician reviews in this small single-center primary comparative evaluation and supported identification of recurring quality-improvement themes in the exploratory expanded-cohort analysis. Preventability judgments remained variable among both AI and physicians, underscoring the need for human oversight and prospective evaluation before operational use.

04.
medRxiv (Medicine) 2026-06-24

Development and External Validation of a Machine Learning Model for 10-Year Ischemic Stroke Risk Prediction in Diverse Populations

Importance: Machine-learning models for ischemic stroke risk prediction are rarely validated across ancestrally distinct cohorts, and the contributions of polygenic risk scores (PRS) and self-reported race in such models remain unclear. Objective: To develop and externally validate a 10-year ischemic stroke risk model and quantify the incremental contributions of laboratory trajectories, PRS, and self-reported race and ethnicity across populations. Design, Setting, and Participants: Retrospective cohort study with model development in the All of Us (AoU) Research Program (n = 34,987; 1,920 incident strokes) and external validation in the BioMe Biobank at Mount Sinai (n = 10,693; 107 incident strokes). Adults aged 45 years or older with at least 1 year of pre-baseline electronic health record data were anchored to a January 2010 baseline with 10-year follow-up. Exposures: Three XGBoost model tiers added laboratory feature trajectories (M2) and 20 PRS (M3) to clinical baseline features (M1); evaluated under race-blind and race-aware specifications. Main Outcomes and Measures: First inpatient ischemic stroke within 10 years; discrimination (area under the receiver operating characteristic curve [AUROC]) and calibration (observed-to-expected [O/E] ratio). Results: In the AoU test partition (n = 6,998; 384 cases), M3 achieved an AUROC of 0.813 (95% CI, 0.788-0.837), outperforming the Revised Framingham Stroke Risk Profile (AUROC difference, 0.164) and Pooled Cohort Equations (AUROC difference, 0.181; both P < 0.001). Discrimination transferred to BioMe (AUROC, 0.745), but predictions were systematically high (aggregate O/E ratio, 0.12 vs 1.00 in AoU), consistent with intercept-shift miscalibration; BioMe-fitted intercept recalibration restored calibration in African American and Hispanic participants but not European American participants. The PRS contribution was significant only among Hispanic participants in BioMe (AUROC difference, 0.042; P = 0.003), with no significant within-stratum gain in the other 5 cohort-by-race combinations. Adding self-reported race produced small gains when combined with PRS (BioMe AUROC difference, 0.022; P = 0.034; AoU AUROC difference, 0.006; P = 0.052) but not when added without PRS. Conclusions and Relevance: A machine-learning ensemble combining clinical, laboratory, and polygenic features outperformed traditional risk scores by 0.16 to 0.18 AUROC and retained discriminative validity in an ancestrally distinct external cohort but required site-specific recalibration of absolute risk. The marginal contribution of self-reported race overlapped with polygenic signal, supporting per-ancestry calibration over universal race-aware model deployment.

05.
medRxiv (Medicine) 2026-06-24

Structural variant discovery and diagnostic impact in rare diseases from short-read and long-read sequencing

Rare diseases collectively affect 1 in 10 individuals, yet current genetic testing fails to identify a causal variant for most cases. At present, cytogenetic methods and/or sequencing approaches such as exome (ES) or short-read genome sequencing (srGS) represent the state-of-the-art for comprehensive clinical discovery of sequence and structural variants (SVs), including copy number variants, balanced SVs, complex SVs, and tandem repeats (TRs). Recently, long-read genome sequencing (lrGS), coupled with multiomics data, has presented great promise to resolve variation in genomic regions recalcitrant to characterization by srGS such as highly repetitive simple repeat sequences and segmental duplications. However, there are few guidelines to enable clinical interpretation of genetic variation in these highly repetitive genomic regions, and the enthusiasm of the field in adopting lrGS has made it difficult to assess the true added diagnostic yield of this technology due to widely variable and inconsistently applied analytic pipelines and variable degrees of pre-screening by ES or srGS. Here, we investigated the contribution of SVs to rare diseases using srGS as a front-line strategy when paired with highly sensitive SV discovery and evaluate the added diagnostic yield of incorporating lrGS for a subset of cases. Our srGS analysis encompassed 1,462 families (3,450 individuals) recruited through the Broad Institute Center for Mendelian Genetics and the Genomics Research to Elucidate the Genetics of Rare Diseases (GREGoR) programs. Diagnostic SVs were identified in 5.4% of cases (79/1,462), of which 80% were uniquely detectable by srGS compared to standard cytogenetic techniques. For 96 families (including 10 families with a heterozygous variant observed in a known recessive gene of clinical relevance), we performed lrGS with methylation profiling, as well as long-read transcriptomic analyses in a subset of 20 trios. Analyses with lrGS yielded over 25,000 SVs per genome, 63% of which were not captured by srGS, along with an additional ~200 rare SNV/indels per genome not previously captured and 12 differentially methylated regions per genome. Among these, we identified only one diagnostic variant not interpreted by srGS, an apparently mosaic de novo SNV in CASK that was absent in the srGS callset due to allelic imbalance. No new diagnoses were supported by long-read transcriptomics or episignatures. In this well characterized rare disease cohort, the added diagnostic yield was thus 1.04% (1/96 families). Following a systematic literature review of prior lrGS studies, we find that most reported diagnoses were detectable by srGS and that our added diagnostic yield is consistent with those prior studies. These studies emphasize the significant impact of comprehensive SV discovery in rare disease cases and further demonstrate the power for increased discovery of novel genomic variation and episignatures from lrGS. Nonetheless, they also serve to temper expectations of dramatic diagnostic advances in rare disease patients until there is more extensive annotation of the functional and clinical impact of all coding and noncoding variation uniquely accessible to lrGS with extensive reference databases spanning highly repetitive genomic sequencing that could be enabled by this transformative technology.

06.
medRxiv (Medicine) 2026-06-22

A Controlled Human Malaria Infection model for relapsing Plasmodium vivax

Background Plasmodium vivax malaria relapses are a major source of morbidity and onward transmission of infection. The underlying mechanisms are poorly understood and current therapies sub-optimal. We examined the safety and feasibility of a controlled human malaria infection (CHMI) model for relapsing P. vivax. Methods We conducted an open-label, proof-of-concept, CHMI study of relapsing P. vivax. Healthy, malaria-naive, Duffy-positive adults aged 18-45 years with extensive CYP2D6 metaboliser phenotype and normal blood glucose-6-phosphate dehydrogenase (G6PD) levels were recruited in Oxford, UK. Mosquito-bite CHMI was performed in Nijmegen, The Netherlands, using Anopheles stephensi mosquitoes infected with PvW1, a clonal isolate of P. vivax from Thailand. All follow-up visits were conducted in Oxford, UK. Primary P. vivax infections (qPCR > 500 genome copies/mL) were treated with artemether-lumefantrine (80mg/480mg at 8, 24, 36, 48 and 60 hours). From Day 28 following CHMI, participants attended a fortnightly clinic for clinical review and qPCR blood sampling, with additional assessments performed for any reported symptoms. P. vivax relapse infections (qPCR > 500 genome copies/mL) were treated with artemether-lumefantrine as per primary infection. Definitive anti-malarial treatment with atovaquone-proguanil (1000mg/400mg once daily for three days) and primaquine (0{middle dot}5 mg/kg/day for 14 days) was administered six months following CHMI, regardless of parasitaemia or symptoms. The primary objective was to assess the safety, feasibility and frequency of relapsing P. vivax after CHMI. Remote follow-up (5 years) is ongoing. The study is registered with ISRCTN registry (ISRCTN48625883). Findings 20 participants were screened for eligibility from 21 January 2025. Five participants (median age 22 years) underwent CHMI (five infected mosquitoes per participant) on 15 April 2025. All participants developed primary P. vivax infection and experienced at least one relapse infection. Two participants experienced a second relapse. Overall incidence rate was 3{middle dot}6 relapse infections per person-year. Solicited adverse events were mild or moderate and there were no serious adverse events. Definitive anti-malarial treatment was administered to all participants. One participant experienced primaquine-induced methaemoglobinaemia, resolving with early discontinuation of treatment (total dose 5{middle dot}3 mg/kg). To date, more than six months after primaquine treatment, no further relapses have been recorded. Interpretation CHMI of relapsing P. vivax is safe and feasible, allowing exploration of the mechanisms underlying relapse infections and providing a platform for future anti-relapse efficacy studies. Funding European Union Horizon Europe programme and UK Research and Innovation (UKRI) via OptiVivax consortium; UK National Institute for Health and Care Research Biomedical Research Centre: Oxford; and UK Medical Research Council.

07.
arXiv (quant-ph) 2026-06-24

Higher-Order Adiabatic Elimination in Atom-Cavity Systems and Its Impact on Spin-Squeezing Generation

arXiv:2506.22383v4 Announce Type: replace Abstract: Spin-squeezed states are metrologically useful quantum states where entanglement allows for enhanced sensing with respect to the standard quantum limit. Key challenges include the efficient preparation of spin-squeezed states and the scalability of estimation precision with the number $N$ of probes. Recently, in the context of the generation of spin-squeezed states via coupling of three-level atoms to an optical cavity, it was shown that increasing the atom-cavity coupling can be detrimental to spin squeezing generation, an effect that is not captured by the standard second-order adiabatic cavity removal approximation. We describe adiabatic elimination techniques to derive an effective Lindblad master equation up to third order for the atomic degrees of freedom. Numerical simulations show that the spin squeezing scalability loss is correctly reproduced by the reduced open system dynamics, highlighting the role of higher-order contributions. Furthermore, we conjecture an extension beyond leading order of the adiabatic elimination technique to the case of conditional dynamics under quantum non-demolition continuous measurement and fast cavity loss, whose reliability is again confirmed by numerical simulation of the dynamics and the corresponding behavior of spin squeezing as a function of $N$.

08.
arXiv (CS.AI) 2026-06-11

Towards Data-free and Training-free Compression for Speech Foundation Models Using Parameter Clustering

arXiv:2606.11836v1 Announce Type: cross Abstract: This paper presents a novel data-free and training-free compression approach for speech foundation models using channelwise clustering via k-means. More fine-grained, mixed sparsity pruning by layer-level varying number of parameter clusters is also explored. Experiments conducted on the LibriSpeech dataset suggest that when operating with pruning sparsity of 50% on HuBERT-large, consistent WER reductions of 27.73%/18.61% absolute (34.37%/21.91% relative) over the magnitude-based pruning were obtained on the test-clean and test-other subsets before fine-tuning and 0.19%/0.79% absolute (3.36%/4.62% relative) after fine-tuning with only 3 epochs. Similar WER reductions of 2.86%/5.02% absolute (59.21%/55.29% relative) were observed against magnitudebased pruning on Whisper-large-v3 at 10% sparsity, all with no significant WER increase relative to the uncompressed baseline.

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

Not What, But How: A Framework for Auditing LLM Responses across Positioning, Generalization, Anthropomorphism, and Maxims

Large language models (LLMs) are being increasingly used to answer subjective, information-seeking questions, where users are sensitive to how responses are communicated, not just whether the answers are correct. Existing LLM evaluations for subjective cultural queries largely focus on factual correctness, ignoring how the response is framed. To this end, we introduce FRANZ, an automated FRAmework for respoNse characteriZation to conduct communicative audit of LLM responses along four dimensions: cultural positioning, use of generalizing language, anthropomorphic cues, and adherence to conversational maxims. To enable this evaluation, we contribute SQUARE - a corpus of 376k subjective questions sourced from 57 subreddits, and mapped to 7 countries and 19 question categories. We demonstrate FRANZ's applicability by scoring responses from three open-weight LLMs. We observe that LLMs show statistically significant differences in the frequency with which they employ each response characteristic. Unlike single-dimensional audits, FRANZ reveals that insider positioning and anthropomorphism are positively coupled, with the degree of coupling varying by country, providing a diagnostic lens for identifying framing divergences.

10.
arXiv (CS.AI) 2026-06-16

Mind-Studio: Executable World Models with Lookahead Evaluation for Partially Observable Games

arXiv:2606.16070v1 Announce Type: new Abstract: World-model synthesis aims to turn interaction experience into an internal model of environment dynamics. Existing symbolic approaches often fit observed transitions or mixtures of local rules, but they do not produce a complete executable program that can run independently of the real environment. We present Mind-Studio, a framework that synthesizes executable pygame-style world models from state-action-next-state trajectories using large language models. Mind-Studio combines entropy-selected traces with a lightweight game skill file containing object, action, and static scene information extracted from screenshots. We evaluate synthesis quality with a K-step lookahead fidelity protocol that compares generated world-model rollouts against Real-ALE rollouts from the same state. On Montezuma's Revenge, Mind-Studio improves chosen-action next-state prediction from 0.3% for PoE-World to 48.7% while verifying 5 of 8 subgoals; across Alien, Assault, and Skiing, it achieves stronger branch-level fidelity than prior learned lookahead sources.

11.
arXiv (CS.LG) 2026-06-19

Humanoid Everyday: A Comprehensive Robotic Dataset for Open-World Humanoid Manipulation

arXiv:2510.08807v2 Announce Type: replace-cross Abstract: From loco-motion to dextrous manipulation, humanoid robots have made remarkable strides in demonstrating complex full-body capabilities. However, the majority of current robot learning datasets and benchmarks mainly focus on stationary robot arms, and the few existing humanoid datasets are either confined to fixed environments or limited in task diversity, often lacking human-humanoid interaction and lower-body locomotion. Moreover, there are a few standardized evaluation platforms for benchmarking learning-based policies on humanoid data. In this work, we present Humanoid Everyday, a large-scale and diverse humanoid manipulation dataset characterized by extensive task variety involving dextrous object manipulation, human-humanoid interaction, locomotion-integrated actions, and more. Leveraging a highly efficient human-supervised teleoperation pipeline, Humanoid Everyday aggregates high-quality multimodal sensory data, including RGB, depth, LiDAR, and tactile inputs, together with natural language annotations, comprising 10.3k trajectories and over 3 million frames of data across 260 tasks across 7 broad categories. In addition, we conduct an analysis of representative policy learning methods on our dataset, providing insights into their strengths and limitations across different task categories. For standardized evaluation, we introduce a cloud-based evaluation platform that allows researchers to seamlessly deploy their policies in our controlled setting and receive performance feedback. By releasing Humanoid Everyday along with our policy learning analysis and a standardized cloud-based evaluation platform, we intend to advance research in general-purpose humanoid manipulation and lay the groundwork for more capable and embodied robotic agents in real-world scenarios. Our dataset, data collection code, and cloud evaluation website are made publicly available on our project website.

12.
arXiv (CS.CV) 2026-06-25

Cross-Attention Multimodal Learning for Predicting Response to Neoadjuvant Imatinib in Gastrointestinal Stromal Tumors: A Multicenter Retrospective Study

Background: Response to neoadjuvant imatinib in gastrointestinal stromal tumors (GISTs) is highly variable and cannot be reliably predicted using current clinical or molecular markers. This study developed and evaluated an explainable multimodal deep learning framework integrating computed tomography (CT) imaging and clinical variables to predict treatment response. Methods: Patients from four tertiary centers were retrospectively included between 2000-2023 in independent pretraining (n=935) and prediction (n=213) cohorts. A cross-attention framework integrating clinical variables and tumor-centered CT imaging was developed to predict response to neoadjuvant imatinib. Two training strategies were evaluated: (1) self-supervised pretraining with low-rank adaptation and (2) training from scratch. Hyperparameters were optimized using SMAC3. Performance was assessed through internal cross-validation and external testing. Ablation analyses and attention-based explanations were used to quantify modality contributions. Results: Among 213 patients (54.5% responders), responders had larger tumors (112 vs. 89 mm, P=0.026), higher mitotic index (3 vs. 0, P

13.
arXiv (CS.AI) 2026-06-19

Beyond Static Endpoints: Tool Programs as an Interface for Flexible Agentic Web Services

arXiv:2606.19992v1 Announce Type: cross Abstract: In the agentic web era, LLM-based agents increasingly invoke web services as tools, yet most interfaces remain static endpoints that poorly express long-horizon workflows with loops, conditionals, joins, and retries. We present ToolPro, which represents an agent's tool intent as an executable tool program that compactly encodes multi-step service interactions with explicit effect types. ToolPro combines constraint-guided program construction, effect-aware replay for exactly-once state-modifying calls, and a profile-driven policy that decides when program execution outperforms stepwise calling. We instantiate ToolPro over MCP-style services with WebAssembly sandboxing and evaluate it on diverse workflows of real-world applications. ToolPro reduces end-to-end latency by up to 53.4\% and client-side traffic by up to 96.1\%, with larger gains under higher network latency and workflow complexity.

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

Parthenon Law: A Self-Evolving Legal-Agent Framework

arXiv:2606.04602v3 Announce Type: replace Abstract: As agents grow more capable, legal-domain LLM agents promise to turn document-heavy matters into reviewable work products – yet reliable deployment faces three obstacles: no large-scale evidence on how today's strongest model-and-harness combinations behave on end-to-end legal matters; no agent architecture adapted to the legal vertical, only general-purpose harnesses; and, in a setting that keeps shifting with new facts, authorities, and deadlines, no mechanism for systems to learn from their own outcomes. We address each. A large-scale empirical study on Harvey LAB – $12{,}510$ agent trajectories – shows that even frontier agents remain far from completing matters in a single pass: per-criterion accuracy climbs with stronger models while strict matter completion stalls. We then introduce \textsc{Parthenon}, a self-evolving legal-agent framework that factors Model, Harness, Agent roles, legal Knowledge, deterministic Tools, and procedural Skills into auditable surfaces for source traceability, date and number grounding, deliverable compliance, and issue closure. Finally, an anti-leakage learning loop converts scored failures into task-agnostic edits to skills, tools, and knowledge, letting the system improve with experience – as a firm refines its checklists and playbooks after each matter – without touching model weights. Across our large-scale empirical analysis, \textsc{Parthenon} substantially improves the performance of state-of-the-art models and harnesses on legal-matter tasks.

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

Clin-JEPA: A Multi-Phase Co-Training Framework for Joint-Embedding Predictive Pretraining on EHR Patient Trajectories

arXiv:2605.10840v3 Announce Type: replace-cross Abstract: We present Clin-JEPA, a multi-phase co-training framework for joint-embedding predictive (JEPA) pretraining on EHR patient trajectories. JEPA architectures have enabled latent-space planning in robotics and high-quality representation learning in vision, but extending the paradigm to EHR data – to obtain a single backbone that simultaneously forecasts patient trajectories and serves diverse downstream risk-prediction tasks without per-task fine-tuning – remains an open challenge. Existing JEPA frameworks either discard the predictor after pretraining (I-JEPA, V-JEPA) or train it on a frozen pretrained encoder (V-JEPA 2-AC), leaving the encoder unaware of the rollout signal that the retained predictor must use at inference; co-training the encoder and predictor under a shared JEPA prediction objective would supply this grounding, but naïve co-training is unstable, with representation collapse and online/target drift causing autoregressive rollout to diverge. Clin-JEPA's five-phase pretraining curriculum – predictor warmup, joint refinement, EMA target alignment, hard sync, and predictor finalization – addresses each failure mode by phase, stably co-training a Qwen3-8B-based encoder and a 92M-parameter latent trajectory predictor. On MIMIC-IV ICU data, three independent evaluations support the framework: (1) latent $\ell_1$ rollout drift uniquely converges ($-$15.7%) over 48-hour horizons while baselines and ablations diverge (+3% to +4951%); (2) the encoder learns a clinically discriminative latent geometry (deteriorating-patient cohorts displace 4.83$\times$ further than stable patients in latent space, vs $\leq$2.62$\times$ for baseline encoders); (3) a single backbone outperforms strong tabular and sequence baselines on multi-task downstream evaluation. Clin-JEPA achieves mean AUROC 0.851 on ICareFM EEP and 0.883 on 8 binary risk tasks (+0.038 and +0.041 vs baseline average).

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

From Memorization to Parameter Interference: How Overtraining Experts Harms Model Merging

arXiv:2506.14126v2 Announce Type: replace-cross Abstract: Modern deep learning is increasingly characterized by the use of open-weight foundation models that can be fine-tuned on specialized datasets. This has led to a proliferation of expert models and adapters, often shared via platforms like HuggingFace and AdapterHub. Model merging has recently emerged as an effective way to leverage these existing resources, enabling the composition of capabilities from different model checkpoints. A natural pipeline has thus formed to harness the benefits of transfer learning and amortize sunk training costs: models are pre-trained on general data, fine-tuned on specific tasks, and then multiple checkpoints are merged to obtain a more capable model. A prevailing assumption is that improvements at one stage of this pipeline propagate downstream, leading to gains at subsequent steps. In this work, we challenge that assumption by examining how expert fine-tuning affects model merging. We show that long fine-tuning of experts that optimizes for their individual performance leads to degraded merging performance across vision and language modalities, multiple model scales, and both fully fine-tuned and LoRA-adapted models. We trace this degradation to the memorization of a small set of difficult examples that dominate late fine-tuning steps. This causes negative parameter interference and encodes knowledge that is forgotten during merging. Finally, we demonstrate that task-dependent aggressive early stopping strategies can significantly improve model merging performance.

17.
medRxiv (Medicine) 2026-06-22

Virtual Responsive Neurostimulation Implantation: From Intracranial Connectivity to Optimized Lead Placement

Responsive neurostimulation (RNS) is an implanted device that delivers direct brain stimulation for drug-resistant focal epilepsy. Individual responses are highly variable, and no validated framework exists to predict outcome or guide lead placement before implantation. We hypothesized that this variability is partly explained by lead placement in relation to patterns of functional connectivity in brain networks. Fourty-nine patients with drug-resistant focal epilepsy who underwent pre-implantation intracranial EEG (iEEG) and RNS implantation across three independent epilepsy centers were retrospectively studied. We developed a composite functional connectivity score, based on simple Spearman correlation, combining the standard deviation and kurtosis of interictal iEEG connectivity distributions to predict the response outcome in a training cohort (HUP, n=18) and validated in two independent cohorts (NYU, n=17; UCSF, n=14). We accounted for a spatial mismatch between iEEG and RNS electrodes with a distance-based correction. The score was extended to generate patient-specific 3D maps of predicted RNS efficacy across 200 simulated, or virtual RNS, lead configurations. Accuracy of the score in predicting clinical outcome was 72% at the group level, 61% at the individual patient level, and, after distance-based optimization, 100% in patients with RNS electrodes placed close to location of iEEG electrodes. Applied to the validation cohort, the same score reached 68% accuracy (71% balanced accuracy, 55% sensitivity, 88% specificity). The spatial combination of the scores at different SEEG contacts localization gives a spatial score for each patient. Responders showed significantly higher spatial scores than non-responders, supporting that actual RNS lead placement in responders was located in map-identified favorable regions. Interictal iEEG functional connectivity predicts individual RNS response across independent epilepsy centers, and patient-specific 3D maps derived from this biomarker could prospectively guide lead implantation toward favorable network regions, opening a promising avenue toward network-informed RNS surgical planning.

18.
arXiv (CS.CV) 2026-06-12

PP-OCRv6: From 1.5M to 34.5M Parameters, Surpassing Billion-Scale VLMs on OCR Tasks

Vision-Language Models (VLMs) have achieved impressive results on general vision-language tasks, yet they suffer from hallucination, imprecise localization, and prohibitive computational cost when applied to dedicated OCR scenarios. This paper presents PP-OCRv6, a lightweight OCR system that combines architectural innovation with data-centric optimization. PP-OCRv6 redesigns the backbone, detection neck, and recognition neck around a unified MetaFormer-style building block with structural reparameterization, decoupling spatial token mixing from channel mixing and supporting both tasks through task-specific stride configurations. Three model tiers (medium, small, tiny) share the same block primitives, covering deployment scenarios from server to edge. On our in-house benchmarks, PP-OCRv6_medium achieves 83.2% recognition accuracy and 86.2% detection Hmean, outperforming PP-OCRv5_server by +5.1% and +4.6% respectively while surpassing Qwen3-VL-235B, GPT-5.5, and Gemini-3.1-Pro with orders of magnitude fewer parameters. The tiny tier achieves 3.9$\times$ faster inference than PP-OCRv5_mobile on Intel Xeon CPU while maintaining comparable accuracy.

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

Scaling Self-Play for End-to-End Driving

End-to-end autonomous driving models are typically trained on offline human-demonstration datasets that provide limited state coverage and often no closed-loop feedback, making them prone to compounding errors when deployed in closed-loop and brittle to long-tail agent interactions. To overcome these limitations, we propose an alternative strategy for training end-to-end driving models: large-scale self-play directly from pixels in simulation. While prior self-play approaches have shown promising transfer to real-world driving, they typically assume vectorized Bird's-Eye-View (BEV) observations that are incompatible with end-to-end policies operating directly on sensor observations. To this end, we introduce Gigapixel, a high-throughput batched driving simulator with perspective rendering, enabling scalable self-play directly from pixel observations. Rather than targeting compute-costly photorealistic sensor simulation, Gigapixel renders a simplified bounding-box world that preserves essential scene structure while achieving throughput at 50k agent steps per second. Since direct pixel-space self-play RL is prohibitively sample-inefficient at end-to-end model scale, we propose self-play DAgger training: we train pixel-based policies in self-play via on-policy distillation from a privileged RL teacher. To bridge the sim-to-real gap, we subsequently transfer the self-play trained policies to real-world sensor data through lightweight perception adaptation. Policies trained in Gigapixel and adapted to real-world sensor data achieve competitive performance on the HUGSIM and NAVSIM-v2 benchmarks without human trajectory supervision. Moreover, scaling self-play training yields proportional gains in policy performance, establishing self-play as a practical and scalable strategy for training end-to-end models.

20.
arXiv (CS.AI) 2026-06-17

CyberEvolver: Structured Self-Evolution for Cybersecurity Agents On the Fly

arXiv:2605.26195v2 Announce Type: replace-cross Abstract: LLM-based agents are increasingly used for cybersecurity tasks, but most existing systems rely on fixed, human-designed scaffolds that struggle to adapt across diverse targets and failure modes. We introduce \textsc{CyberEvolver}, a self-evolving cybersecurity agent framework that iteratively revises its own scaffold based on experience from failed execution attempts. Self-evolution in cybersecurity is challenging because the space of possible scaffold changes is largely unstructured, execution feedback is sparse and often obscured by the environment, and low-diversity updates can cause errors to compound over repeated iterations. \textsc{CyberEvolver} addresses these challenges with a four-layer evolvable agent architecture that decomposes scaffold optimization into structured components, a trace-to-diagnosis mechanism that converts noisy execution logs into actionable revision signals, and a population-based beam search strategy that preserves diverse agent variants during evolution. We evaluate \textsc{CyberEvolver} on CTF challenges, vulnerability exploitation, and penetration-testing tasks using four open-source LLMs. Across these settings, \textsc{CyberEvolver} improves the seed agent's success rate by $13.6$\,\% on average, and outperforms six human-designed cybersecurity agents as well as two self-improvement methods adapted from other domains. These results suggest that scaffold self-evolution is a promising direction for building adaptive LLM agents for security testing.

21.
arXiv (CS.LG) 2026-06-18

Neural Network Implementation of the Renormalization Group for Fault Diagnosis with Class Imbalance

arXiv:2606.18326v1 Announce Type: new Abstract: The application of machine learning models in practical tasks faces challenges such as class imbalance and multidimensional noise. This paper proposes RGNet, a neural network architecture based on the concept of the renormalization group (RG), for hierarchical coarse-graining of the feature space. The model sequentially compresses the input dimensionality and concatenates all scales before classification, allowing it to capture both local details and global patterns. The notion of RG-flows is introduced - interpretable low-dimensional representations whose visualization via t-SNE reveals a discrete curvilinear structure confirming the effectiveness of coarse-graining. Experimental results are presented on the imbalanced AI4I dataset. The obtained results demonstrate that RGNet is a universal, interpretable, and competitive solution for fault prediction in applications with imbalanced classes.

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

CacheMuon: Using Temporal Preconditioning To Approximate Polar Factor

arXiv:2606.16371v1 Announce Type: new Abstract: Muon is an optimizer that computes updates using the polar factor of the momentum matrix and has shown strong empirical performance across a range of training settings. A key component of Muon is the Newton-Schulz iteration used to compute this polar factor. Although this avoids the cost of an exact singular value decomposition, it remains expensive in practice because it is applied at every optimization step. At the same time, the momentum matrix changes smoothly over training, suggesting strong temporal correlation in the corresponding polar factors. In this paper, we exploit this structure and propose CacheMuon, a temporal preconditioning method that reuses information from previous optimization steps to approximate the polar factor at the current step. This reduces redundant orthogonalization computation across iterations. We analyze CacheMuon as an inexact Muon update, with error controlled by fresh-solver error and cache staleness. Empirically, CacheMuon provides a controllable quality-efficiency frontier: conservative thresholds closely match fresh Muon on language-model and vision training while reducing orthogonalization FLOPs, whereas more aggressive thresholds yield larger arithmetic savings at the cost of modest validation-quality degradation.

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

FlexPooling with Simple Auxiliary Classifiers in Deep Networks

In computer vision, the basic pipeline of most convolutional neural networks consists of multiple feature extraction layers, where the input signal is downsampled to a lower resolution in each subsequent layer. This downsampling process is commonly referred to as pooling, which is an essential operation in CNNs. Pooling improves robustness against transformations, reduces the number of trainable parameters, increases the receptive field, and lowers computation time. Since pooling is a lossy process but remains important for extracting high-level information from low-level representations, it is important to preserve the most prominent information from previous activations to improve network discriminability. Standard pooling is usually performed using dense pooling methods, such as max pooling or average pooling, or through strided convolutional kernels. In this paper, we propose a simple yet effective adaptive pooling method, called FlexPooling, which generalizes average pooling by learning a weighted average over activations jointly with the rest of the network. We further show that attaching Simple Auxiliary Classifiers (SAC) to the CNN improves performance and demonstrates the effectiveness of the proposed method compared with standard pooling methods. Experiments on multiple popular image classification datasets show that FlexPooling consistently outperforms baseline networks, achieving approximately 1 to 3 percent improvement in accuracy.

24.
arXiv (CS.AI) 2026-06-16

STRIDE: Strategic Trajectory Reasoning via Discriminative Estimation for Verifiable Reinforcement Learning

arXiv:2606.15866v1 Announce Type: new Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) has become an effective post-training paradigm for improving the reasoning abilities of large language models. However, existing RLVR methods typically rely on final-answer correctness to assign trajectory-level rewards, providing sparse supervision and treating all tokens uniformly regardless of their actual contribution to reasoning. Although recent studies introduce intermediate signals such as process rewards, high-entropy tokens, and semantic uncertainty, these signals are often not inherently verifiable and may fail to distinguish beneficial strategic patterns from harmful ones. To address this limitation, we propose STRIDE (Strategic Trajectory Reasoning with Discriminative Estimation), a fine-grained RLVR framework that derives strategic reasoning supervision from verifiable outcomes. STRIDE contrasts successful and failed trajectories within each response group to estimate the outcome-discriminative preference of each $n$-gram strategic pattern, and further combines this signal with reasoning saliency entropy to identify decision-relevant strategic patterns. These patterns are assigned differentiated advantage values during RL optimization, enabling more precise credit assignment while preserving the verifiability of RLVR. Extensive experiments demonstrate that STRIDE consistently improves reasoning performance across diverse models, tasks, and extended settings, including VLMs and agent-based systems.

25.
arXiv (CS.LG) 2026-06-15

Arbitrary control over multimode wave propagation for machine learning

arXiv:2402.17750v2 Announce Type: replace-cross Abstract: Controlled multimode wave propagation can enable more space-efficient photonic processors than architectures based on discrete components connected by single-mode waveguides. Instead of defining discrete elements, one can sculpt the continuous substrate of a photonic processor to perform computations through multimode interference in two dimensions. Here we designed and demonstrated a device with a refractive index that can be rapidly reprogrammed across space, allowing arbitrary control of wave propagation. The device, a two-dimensional programmable waveguide, uses parallel electro-optic modulation of the refractive index of a slab waveguide with about $10^4$ programmable spatial degrees of freedom. We implemented neural network inference on benchmark tasks with up to $49$-dimensional vectors in a single pass, without digital pre-processing or post-processing. Theoretical and numerical analyses further indicated that two-dimensional programmable waveguides may offer not only a constant-factor reduction in device area but also a scaling benefit, with the area required growing as $N^{1.5}$ rather than $N^2$.