Ax Quang Hung Pham, Ryad Zemouri, Martin Gagnon, Luc Vouligny 7/7/2026

Modular Foundation Models for Time-Series Perception in Digital Twins

Modular foundation models for time-series perception in digital twins and prognostics health management systems.

Ax Xiaoyue Liu, Zheng Dong 7/7/2026

LLM-Guided Transportation Hub Capacity Planning with Textual Business Inputs

LLM agent framework for transportation hub capacity planning that iteratively proposes decisions guided by natural-language business context.

Ax Zhuoer Shen, Mingyi Wang, Shaofeng Zou, Yuheng Bu 7/7/2026

Rethinking AI-Generated Text Detection: A Strong Baseline and the Distribution-Shift Problem That Remains

Shows fine-tuned RoBERTa matches specialized detectors for AI-generated text detection; challenges recent architectural complexity in detection methods.

Ax Kaixuan Liu, Guojun Xiong, Weinan Zhang, Shengpu Tang 7/7/2026

Social Networks of LLM Agents

Studies how populations of LLM agents form collective beliefs and whether they aggregate genuine knowledge or collapse into false consensus.

Ax Byoungkwon Kim, Minhyuk Sung 7/7/2026

Tensor-Train Joint Modeling for Few-Step Discrete Diffusion

Proposes tensor-train joint modeling to improve discrete diffusion models for faster sequential generation compared to autoregressive approaches.

Ax Yoshihiro Maruyama 7/7/2026

Foundations of Equivariant Deep Learning: Unifying Graph and Sheaf Neural Networks

Extends geometric deep learning with order-equivariant neural networks that generalize graph message passing and sheaf neural networks using equivariant bundle theory.

Ax Bhavesh Sood, Jaromir Savelka 7/7/2026

Punching Above Their Weight: Classification-Head Fine-Tuning of Tiny Language Models (TLMs) for Verifiable Multiple-Choice Tasks

Study of classification-head fine-tuning for tiny language models (under 3B parameters) on multiple-choice reasoning tasks, comparing LoRA paradigms.

Ax Benjamin Wiriyapong, Oktay Karakus, Can Eyupoglu, Kirill Sidorov 7/7/2026

Stable Global Weighting of Flow Mixtures using Simplex Exponential Moving Average

Two-stage framework for normalizing flow mixtures using simplex exponential moving average for stable weighting across heterogeneous posterior geometries.

Ax Zhen Huang, Peicheng Xu, Junbiao Pang, Yulong Zheng 7/7/2026

Adversarial LassoNet: Robust Feature Selection via Stability-Driven Sparse Learning

Adversarial training approach for robust feature selection in high-dimensional learning, improving stability of sparse feature supports under noise.

Ax Aakash Kumar, Emanuele Natale 7/7/2026

A Unified Framework for Quantized and Continuous Strong Lottery Tickets

Unified theoretical framework for strong lottery ticket hypothesis in both quantized and continuous settings using random subset sum problem.

Ax Sadra Saremi 7/7/2026

AdaptiveSD A Stability-Aware, Runtime-Adaptive Speculative Decoding Framework with Multi-Policy Orchestration for CPU-Constrained LLM Inference

Runtime-adaptive speculative decoding framework for CPU-constrained LLM inference, using multi-policy orchestration to optimize small quantized model performance.

Ax Jiayi Guan, Tianle Zhang, Li Shen, Ruiqi Zhang, Ao Zhou, Lusong Li, Guai Chen, Mengjie Li, Alois Knoll, Xiaodong He, Changjun Jiang 7/7/2026

CDCP: Conditional Diffusion Model with Contextual Prompts for Multi-task Offline Safe Reinforcement Learning

Conditional diffusion model framework for multi-task offline safe reinforcement learning, handling safety constraints and out-of-distribution actions.

Ax Subhajit Dandapat, Alvin J. K. Chua 7/7/2026

Transformers with Physics-Informed Encodings and Simulation-Based Inference for Robust Detection of Eccentric Binary Black Holes in Pulsar Timing Array Data

Transformer with physics-informed encodings for gravitational wave detection in pulsar timing array data using simulation-based inference.

Ax Weibin Li, Wendu Li, Yushan You, Chen Wei, Quanying Liu 7/7/2026

NeuroOnline: Bridging Pretraining and Online Adaptation for EEG Foundation Models

Framework combining pretraining with online adaptation for EEG foundation models to handle distribution shifts and task-specific requirements.

Ax Shubhadip Nag, Srinjoy Das, Agniva Saha, Anushree Ghosh, Soumi Das, Tarun Kumar, Suparna Bhattacharya, Sourangshu Bhattacharya 7/7/2026

MPSelectTune: Prompt-type Selection for Fine-tuning improves Concept Unlearning in LLMs

Prompt-type selection framework for fine-tuning to improve concept unlearning in LLMs, removing biased/harmful concepts across diverse prompt variations.

Ax Matan Avitan, Yoav Goldberg, Yanai Elazar 7/7/2026

MANCE: Manifold Aware Concept Erasure

Manifold-aware approach to concept erasure from neural representations, addressing preservation of correlated information during target concept removal.

Ax Yaron Anavi, Mor Aisenberg, Nadav Nesher, Elena Khabibullina, Isabella Cattinelli 7/7/2026

Knowing When to Stop: Predicting Execution-Consistency Convergence in Text-to-SQL

Convergence prediction method for Text-to-SQL pipelines using lightweight models to determine when repeated LLM calls reach sufficient consistency.

Ax Ashmitha R, J\"org Frochte 7/7/2026

Directional Curvature from Armijo Backtracking: A Low-Cost Sharpness Probe and a Calibration-Free Learning-Rate Safeguard for Adam

Low-cost method to measure loss sharpness via Armijo backtracking for calibrating Adam learning rates without Hessian computations.

Ax AlJawharh S. AlOtaibi, Mohamed Eltahir, Jude AlSubaie 7/7/2026

When Does Small Data Work? Accuracy and Efficiency Trade-offs Between Tabular Foundation Models and Conventional Methods for Crowd-State Classification at Hajj and Umrah

Empirical study comparing tabular foundation models to conventional methods for few-shot learning on crowd-state classification at religious gatherings.

Ax Ronaldo C. Prati 7/7/2026

A Unified Algebraic Framework for Classification Performance Evaluation

Unified algebraic framework for classification metrics covering binary, multiclass, multilabel, and other evaluation settings in single formalism.

Ax Siyuan Li, Jiabao Pan, Yumou Liu, Zhuoli Ouyang, Xin Jin, Xinglong Xu, Jingxuan Wei, Shengye Pang, Jintao Che, Xuanhe Zhou, Conghui He, Cheng Tan 7/7/2026

OmniOpt: Taxonomy, Geometry, and Benchmarking of Modern Optimizers

Unified survey and benchmark of 100+ optimizers for large-scale model training, providing taxonomy and selection guidance for compute-constrained scenarios.

Ax Mohammad Sadegh Akhondzadeh, Vijay Lingam, Atula Tejaswi, Chanakya Ekbote, Sujay Sanghavi, Aleksandar Bojchevski 7/7/2026

Reward-Gated On-Policy Distillation

On-policy distillation method using reward gating to improve teacher supervision reliability when transferring reasoning from strong to smaller student models.

Ax Chenrui Liu, Chuanlong Xie, Falong Tan, Yicheng Zeng, Lixing Zhu 7/7/2026

A Unified Framework for In-Context Learning with Causal and Masked Language Models

Statistical framework analyzing in-context learning in both causal and masked language models, extending theoretical understanding beyond autoregressive models.

Ax Zijian Wang, Pengfei Li, Guangyu Yang, Qiong Zhang 7/7/2026

FedSPM: Routing-Enabled Federated Learning under Dual Heterogeneity via Semiparametric Mixture

Federated learning framework with routing mechanism addressing dual heterogeneity using semiparametric mixtures. Handles both inter-client and intra-client latent subpopulation variations.

Ax Yi Lan, Ye Yuan 7/7/2026

Target-Aware Interaction-Guided Reinforcement Learning for Black-Box Node Injection Attacks on Graph Neural Networks

Reinforcement learning approach for black-box node injection attacks on GNNs. Jointly optimizes malicious node features and edge connections for adversarial attacks.

Ax Pan Li 7/7/2026

Dictionaries, Not Darwin: Set-Level Selection Beats LLM Evolution in Scientific Equation Discovery

Comparison of evolutionary LLM-based scientific discovery versus dictionary selection for equation discovery. Shows independent sampling outperforms parent-conditioned evolution under matched budgets.

Ax Silin Gao, Hao Zhao, Zeming Chen, Sepideh Mamooler, Antara Raaghavi Bhattacharya, Qiyu Wu, Hiromi Wakaki, Yuki Mitsufuji, Li Mi, Syrielle Montariol, Antoine Bosselut 7/7/2026

DynaVieW: Schema-Guided World Modeling for Understanding Hierarchical Visual Dynamics

Schema-guided world model for multimodal LLMs to predict visual dynamics. DynaVieW models temporal evolution of videos across multiple hierarchical levels of visual change.

Ax Shiheng Zhang 7/7/2026

Asymptotic-Preserving A Posteriori Analysis of Diffusion and Flow-Matching Samplers

Theoretical analysis of diffusion and flow-matching samplers treating terminal noise scale as singular perturbation. Determines asymptotic-preserving properties of fixed-step samplers.

Ax Arunkumar Ramachandran 7/7/2026

GlacierCastAI: Predicting Glacier Retreat from Multi-Modal Satellite Imagery and Climate Signals

Multi-modal spatiotemporal forecasting system predicting glacier retreat. Combines Landsat satellite imagery with ERA5 climate variables for boundary prediction.

Ax Fengxian Ji, Zhuohan Xie, Jingpu Yang, Fan Zhang, Zirui Song, Xiuying Chen 7/7/2026

Parametric Memory Decoding for Zero-Shot Routing in LoRA-Based External Parametric Memory

Zero-shot routing method for LoRA-based external parametric memory. Eliminates need for additional routing component in modular LLM solutions.

Ax Huqin Weng, Jiayang Huang, Yimin Wen, Jie Du, Chi-Man Vong, Chuangquan Chen 7/7/2026

Masked Generative-Contrastive Representation Learning for Cross-Dataset EEG-Based Emotion Recognition

Self-supervised learning approach combining masked and contrastive learning for EEG emotion recognition. Improves cross-dataset transfer with spatiotemporal dependency modeling.

Ax Kai Zhao 7/7/2026

Mask-based Predictive Representations for Reinforcement Learning

Self-supervised learning method using mask prediction for vision-based reinforcement learning. Addresses sample efficiency in high-dimensional image inputs.

Ax Saksham Bassi, Sharvi Tomar 7/7/2026

Geometry of Ordinal Representations in Language Models

Analysis of how language models represent ordinal information geometrically. Studies attention heads performing geometric transformations across bracket depth, indentation, and numeric tasks in Gemma and Qwen models.

Ax Liyang Yuan, Yibo Yang, Dandan Guo 7/7/2026

FedFFT: Taming Client Drift in Federated SAM via Spectral Perturbation Filtering

Federated learning approach combining Sharpness-Aware Minimization with spectral perturbation filtering. Addresses client drift and convergence problems in decentralized training.

Ax Jinfeng Zhu, Shiyu Long, Ye Yuan 7/7/2026

Physics-Informed Graph Learning with Uncertainty Awareness for Open-Set Domain Generalization in Fault Diagnosis

Graph learning method for fault diagnosis in rotating machinery. Combines physics-informed approaches with uncertainty awareness for open-set domain generalization.

Ax Liyang Yuan, Yibo Yang, Dandan Guo, Peter Richtarik, Zhouchen Lin 7/7/2026

SpecGradFilter: A Spectral Gradient Filtering Framework for Taming Federated Heterogeneity

Federated learning framework addressing statistical heterogeneity via spectral gradient filtering. Uses frequency-domain analysis to mitigate client drift in non-IID data scenarios.

Ax Francisco Passos 7/7/2026

Exploring Convolutional Neural Processes for Weather Downscaling

Applies Convolutional Conditional Neural Processes to weather downscaling. Uses neural processes to increase ERA5-Land resolution from 11km to 1km for temperature prediction in mountainous regions.

Ax Zubaida Fatima, Zubair Shaban, Yusuf Jamal, Nazreen Shah, Ranjitha Prasad, B. N. Bharath 7/7/2026

Channel-Adaptive Robust Aggregation for Over-the-Air Federated Learning in Heterogeneous Networks

Federated learning approach for over-the-air wireless aggregation in heterogeneous networks. Addresses noise and fading in privacy-preserving IoT and autonomous systems.

Ax Jaeyeon Kim, Jewon Lee, Bo-Kyeong Kim 7/7/2026

Quantize the Target, Quantize the Drafter: Efficient Inference with Qwen3.5-4B

Efficient inference method for Qwen3.5-4B using quantization and speculative decoding. Combines quantized target model with block-diffusion drafter for low-latency GPU serving.

Ax Zitao Shuai, Zongzhe Xu, Yuntian Wu, Sirui Li, Tianhong Li, Yuzhe Yang 7/7/2026

Signal or Noise? Understanding Generative Models for Real-World Sensor Time Series

Research on generative models for sensor time series data. Studies how generative models handle continuous, high-dimensional, noisy sensor data across different modalities and tasks.

HN cybermango 7/7/2026

Claude Fable 5 Backlash Grows

Claude Fable 5 performance degradation after July release. Users report coding and agentic capabilities decline; Anthropic attributes to safety updates.

HN matmult 7/7/2026

The AI-Native Founder

Opinion piece critiquing 'AI-native founder' concept and leadership approach. Commentary rather than technical content.

HN gokuljs 7/7/2026

The Hard Parts of Real-Time Voice Agent Audio

Technical deep-dive on real-time voice agent audio engineering challenges including latency, buffering, and quality. Blog post with code.

HN matt_d 7/7/2026

PuzzleMoE: Efficient Compression of Large Mixture-of-Experts Models

PuzzleMoE: Compression technique for large MoE models addressing memory overhead through expert strategies. Research paper.

HN verdverm 7/7/2026

Jackrong LLM Fine-Tuning Guide

Jackrong: Open-source LLM fine-tuning educational resource. Covers SFT, RL, GRPO, dataset distillation, GGUF deployment, agent workflows.

HN isfttr 7/7/2026

Claude Sonnet 5: Anthropic's Most Agentic AI Model Arrives at a Reduced Price (2026)

Claude Sonnet 5 release announcement with agentic improvements and price reduction in 2026.

BL 7/7/2026

Australian Payments Plus moves faster with ChatGPT and Codex

Australian Payments Plus uses ChatGPT Enterprise and Codex to improve payments processing speed and quality while maintaining human oversight.

HN sscaryterry 7/6/2026

Small AI Models Gain Traction In places with unreliable networks

TinyML models gaining adoption in regions with poor infrastructure. Focus on resource-constrained ML deployment for healthcare applications.

HN pw 7/6/2026

Show HN: Instantvideos.org – short documentaries in ~30 seconds

Pipeline for autonomous video generation using Claude and GLM-5.2 fast models, achieving 30-second generation times.

HN proy24 7/6/2026

Samsar: Model-Composable 1-Shot Generative Video Harness and Software Factory

Samsar: Full-stack generative video platform with text-to-video, image editing, recommendations, audio generation. Deployable Docker factory.