This study fine-tunes two pretrained latent diffusion models, Protogen v3.4 and Stable Diffusion v1.4, on a curated dataset of high-resolution Ulos motifs to generate culturally consistent yet novel designs. Protogen v3.4 significantly outperforms Stable Diffusion v1.4 in terms of FID and IS, highlighting a fidelity-diversity tradeoff. A guidance scale of 5–9 is recommended for optimal balance.

CoFINN is a physics-informed deep learning framework for predicting compressible flow fields, embedding finite-volume conservation physics directly into the training process. It outperforms traditional data-driven CNNs and classical physics-informed methods, reducing drag prediction error by up to 34% on transonic airfoil flows and 15% on average, especially beneficial in limited-data regimes.

The paper proposes a unified deep learning pipeline integrating semantic segmentation, regression-based severity estimation, and disease classification for plant disease severity quantification. On the Apple Tree Leaf Disease Segmentation dataset, U-Net with MobileNetV2 achieves 98.20% pixel accuracy, 0.70 mIoU, and 99.41% detection accuracy at 14.7 ms per image, suitable for real-time use. The computed severity index strongly correlates with expert annotations (r=0.968), demonstrating reliability for automated crop monitoring.

A new evaluation framework predicts item psychometric parameters from text embeddings, revealing that difficulty is highly predictable while discrimination and pseudo-guessing are limited by reliability ceilings. The study highlights the need for repeated cross-validation and scale-free metrics in benchmark construction.

This paper analyzes over-calling in multi-teacher on-policy distillation for tool-using language models and proposes Soft Clamp, a per-token divergence calibration method that reduces over-calling from 13.7% to 9.0% while matching decision accuracy.

This work proposes Riemannian Mean Pooling (RMP), which extracts per-token pullback metrics from a learned encoder's analytical Jacobian and aggregates them via Fréchet mean on the SPD manifold. RMP outperforms Euclidean mean pooling on CoLA, CREAK, and RTE, while staying at chance on FEVER-Symmetric. Ablations show that even a randomly initialized encoder with Fréchet aggregation beats Euclidean pooling on most datasets, locating the gain in geometric aggregation.

Large language models increasingly improve reasoning via test-time computation, but most methods treat problems in isolation. MILES introduces modular memory units with learnable selection heads to accumulate reusable experience across sequential problems, achieving superior accuracy-efficiency tradeoffs.

A new multi-factor scoring framework integrates five dimensions to evaluate LLM response quality, revealing strengths in reasoning but significant weaknesses in factual consistency and ambiguity handling.

Large language models (LLMs) systematically prefer Standard American English (SAE) continuations even when the preceding context is in African American English (AAE), effectively rewriting AAE into SAE. The authors propose an end-to-end framework to audit and mitigate this bias, including conditional Dialect Group Invariance (cDGI) and activation steering. They also release REAL-AAE, the largest real-AAE parallel corpus to date.

A gradient-based speech-to-text alignment method applicable to any differentiable ASR model, including CTC, transducer, attention-based encoder-decoders, and speech LLMs. No training or model modification required. It aligns on the input grid for finer precision. Evaluated on 16 models, it provides usable alignment for all, outperforming native alignment on streaming models, at the cost of one backward pass per token.

This paper proposes a programmatic solution to generate advertising headlines for e-commerce using a self-critical masked language model optimized with reinforcement learning policy gradients. The method conditions on multiple products jointly and outperforms existing Transformer and LSTM+RL methods in overlap metrics and quality audits, even surpassing human-written headlines in grammar and creativity.

Large language models (LLMs) achieve promising results on medical question answering benchmarks, yet their use in public health is constrained by hallucinations and the rapid evolution of official guidance. Retrieval-Augmented Generation (RAG) mitigates these risks by grounding responses in an explicitly maintained corpus, but end-to-end performance depends critically on retrieval configuration and on evaluation beyond multiple-choice formats. We extend PubHealthBench into a retrieval-augmented setting and systematically evaluate retrieval and generation choices. Hybrid retrieval consistently improves recall and ranking quality. Providing retrieved context substantially increases multiple-choice accuracy across a diverse set of LLMs, enabling smaller open-weight models to match or outperform larger models used without retrieval. We introduce a rubric-based LLM-as-a-judge covering faithfulness, completeness, clarity, and factual consistency, and validate it against dual human annotations. Judge-human agreement is strongest for faithfulness and completeness, while factual consistency and clarity are less reliably reproduced.

This paper proposes a multimodal solution that integrates audio and automatically generated multilingual text transcripts via cross-modal transformers, achieving significant performance gains in sentiment polarity classification. Knowledge distillation further enhances an audio-only model without extra inference cost.

TSF uses LLMs offline to build task-semantic fields, enhancing online forecasting without LLM overhead. Achieves 6.4% average MAE reduction with only 1.8-3.0k additional parameters.

This paper investigates how the spectral properties of the attention score matrix are influenced by positional encoding. Analyzing seven pretrained models, they find that previous-token heads under RoPE exhibit rotational spectra, while those under learned-absolute and ALiBi do not. Dynamic analysis shows spectral signatures emerge after behavior, and causal experiments demonstrate that no spectral channel is necessary but bans delay formation.

Inertia-1 is a fully open exploration of wearable motion foundation models using over 18.2 million hours of accelerometer data. It systematically studies data, model, and training choices, evaluating on 15 datasets for tasks like human activity recognition, freezing-of-gait detection, and disease prediction. The work provides state-of-the-art recipes and an open cookbook for wearable motion representation learning.

A new method, FedEAS, addresses label skew in federated learning by adaptively allocating per-class generation budgets to clients based on their local label distributions. It recovers most accuracy gains of full class balancing while reducing the generation budget by 94.1%.

STAGformer is an efficient spatio-temporal agent graph transformer achieving global modeling with linear complexity via a two-step agent attention mechanism, outperforming state-of-the-art on NYC Citi-Bike and Chicago Divvy-Bike datasets.

This paper proposes a multi-objective reliability-based portfolio optimization framework using deep reinforcement learning (MORP-DRL) that jointly optimizes expected return and downside risk. It incorporates three risk measures (variance, CVaR, EVaR), models market uncertainty with GARCH and extreme value theory, and uses PPO under practical constraints, outperforming NSGA-II on ten global equity indices across different market regimes.

D2PO (Dynamic Direct Preference Optimization) is a framework for optimizing diffusion sampling policies, including timestep schedules and classifier-free guidance weights. It addresses the limitation of student-teacher regression where low-NFE student samplers sacrifice high-frequency texture fidelity. By reformulating optimization as preference alignment using energy-based models and dynamic preferences, it achieves better perceptual quality. Experiments show superiority over regression schedulers under low-NFE constraints.

Researchers propose NEST, a framework that addresses dataset-level distribution shifts by identifying distinct operational regimes via unsupervised clustering and using a regime-oriented mixture-of-experts architecture. It achieves state-of-the-art performance on long-term forecasting tasks across network traffic and physical phenomena benchmarks.

Marginal conformal prediction, used in drug discovery, can severely undercover minority classes in imbalanced datasets despite achieving global coverage targets. A class-conditional (Mondrian) approach restores per-class reliability.

TriRoute is a lightweight unified controller that jointly coordinates attention mode, expert selection, and KV-cache bit-width for every token at every layer, outperforming independent optimizations under matched compute and memory budgets.

A study on arXiv shows AI agents often fail in dynamic coordination due to neglecting implicit social norms. From pedestrian-vehicle interaction experiments, researchers identified three principles: outcome predictability, value alignment, and advantage awareness. Incorporating these into LLMs boosted closed-loop task scores nearly fourfold, outperforming human-human interactions by 43%, paving the way for more natural AI integration.

We present the Large Behavioral Model (LBM) that learns customer decision making directly from large-scale retail transactions through a unified Person-Environment formulation. Using behavioral profiles from historical purchases and retrieval-augmented generation, LBM is trained with continued pre-training, supervised fine-tuning, and reinforcement learning with verifiable rewards. It outperforms frontier general-purpose language models on purchase prediction, basket completion, and other retail tasks, with strong zero-shot and fine-tuned transfer across retailers. Ablation studies highlight continued pre-training as the main driver of generalization, and reinforcement learning improves reliance on behavioral evidence.

A new study reveals that optimizing the orchestration layer (Harness) can significantly reduce costs and improve efficiency in enterprise agentic AI, more than model choice alone. Experiments show the Writer Agent Harness cuts cost per task by 41%, time by 44%, and tokens per task by 38%, while maintaining quality.

Recent advances in AI for Mathematics have focused largely on autoformalization and theorem proving. This paper proposes a ReAct-style agentic setup combining LLM reasoning with verifiable feedback from SageMath and Context7 documentation. Evaluated on research-level problems from RealMath, the setup shows substantial performance gains averaging 9.7 pp, with GPT-5.5 achieving 75.2% solve rate. Accepted to ICML 2026 3rd AI for Math Workshop.

Researchers propose two cost-effective agent architectures—Explorer-Definer Pipeline and Reflective Orchestrator—achieving 57.50% and 67.25% pass@2 on ARC-AGI-1 at $0.25 and $0.62 per task respectively, without benchmark-specific training or heavy test-time compute.

QANTIS treats a quantum processor as a calibrated belief-update service for autonomous systems under partial observability. Using IBM Heron hardware on the Tiger POMDP, the study shows that all-step fixed-point amplification preserves the posterior across sequential steps, with hardware posteriors matching exact Bayes posteriors in all decision checks. Boundary-aware BIQAE stabilizes amplitude estimation, and a rare-event sweep maps sample complexity for one-in-a-million evidence.

Researchers introduced a scalable Hybrid Agent-based and Language-driven Epidemic (HALE) modeling framework that leverages large language models (LLMs) to predict human decision-making in agent-based modeling (ABM), with a proof-of-concept simulation of COVID-19 in Salt Lake County, UT.