STAR-VAE is a Transformer-based variational autoencoder that uses SELFIES encoding and a bidirectional encoder with an autoregressive decoder pretrained on 79 million PubChem molecules. It incorporates a property signal to jointly condition the prior, posterior, and decoder, and employs LoRA adapters for fine‑tuning on small datasets without altering the backbone. The model achieves 100 % validity and near‑perfect novelty in MOSES sampling, low KL divergence on several GuacaMol descriptors, strong synthetic‑accessibility conditioning, and effective docking‑score control across multiple protein targets, while also enabling scaffold recovery and diverse label‑conditioned generation on ChEMBL targets.
By Bum Chul Kwon, Ben Shapira, Moshiko Raboh, Shreyans Sethi, Shruti Murarka, Joseph A Morrone, Leili Zhang, Wendy Cornell, Jianying Hu, Parthasarathy Suryanarayanan
The paper proposes a Quantum-Based Parallel Model (QBPM) that uses two quantum circuits running in parallel to classify Alzheimer's disease stages from MRI data. It demonstrates high accuracy on two datasets, remains robust under Gaussian noise, and outperforms five classical transfer learning methods while using fewer circuit parameters. The study highlights QBPM’s potential as a faster, more efficient alternative to classical AI for high-dimensional, noisy medical data.
By Emine Akpinar, Murat Oduncuoglu
MONA is a new optimizer that extends the Muon optimizer by adding a Nesterov‑style acceleration term derived from an exponential moving average of gradient differences. The paper provides a convergence analysis showing that this term offers curvature‑aware corrections while maintaining Muon’s spectral‑norm regularization. Empirical results demonstrate that MONA outperforms both Muon and AdamW on Mixture‑of‑Experts pretraining across models ranging from 1 B to 68 B parameters, and achieves state‑of‑the‑art performance on downstream benchmarks after fine‑tuning the largest model.
By Jiacheng Li, Jianchao Tan, Hongtao Xu, Jiaqi Zhang, Yifan Lu, Yerui Sun, Yuchen Xie, Xunliang Cai
The paper presents a framework that adapts general-purpose Vision Foundation Models (VFMs) to seismic data denoising using Parameter‑Efficient Fine‑Tuning with Low‑Rank Adaptation (LoRA). It introduces a kurtosis‑guided unsupervised test‑time adaptation module that updates only LoRA parameters to self‑calibrate for site‑specific noise without ground truth. Experiments on exploration seismic images and DAS data demonstrate that the approach matches or surpasses domain‑specific models and generalizes well to unseen cross‑site data.
By Jiahua Zhao, Umair bin Waheed, Jing Sun, Yang Cui, Nikos Savva, Eric Verschuur
The paper introduces a retrieved‑span training approach for query‑focused meeting summarization on the QMSum benchmark. By fine‑tuning a 406 M Fusion‑in‑Decoder model on 2,000‑word retrieved spans, the authors recover a 6.30 ROUGE‑1 loss incurred when moving from capped long input and achieve a test score of 36.33 ROUGE‑1, comparable to a larger 1.2 B system. The smaller model uses roughly one‑third the parameters and less than half the peak inference memory, while span‑regime fine‑tuning adds significant gains over the baseline.
"whyItMatters":"The study demonstrates that efficient, smaller models can match or exceed larger systems on QMSum using span‑based fine‑tuning, offering a practical path for scalable query‑focused meeting summarization."
By Edward Xi Yang (Ertas AI)
ChainDoRA is a new parameter‑efficient fine‑tuning framework for large language models that replaces the dense low‑rank factorization of LoRA with a connected Tensor‑Train (TT) chain. By separating weight magnitude and direction and using a TT rank to control representation capacity, ChainDoRA achieves higher average accuracy on seven commonsense reasoning benchmarks while dramatically reducing trainable parameters—down to 5.35 M versus 56 M for LoRA and DoRA. Ablation studies show that the TT parameterization offers controllable trade‑offs between parameter cost and accuracy.
By Ashfak Yeafi, Mehedi Hasan, Md Khairul Islam
TelecomGPT‑R1 is an open‑source family of unified telecom reasoning models that address the limitations of existing telecom‑specific and general‑purpose LLMs. It is built on a four‑axis framework—protocol, knowledge, modeling, and fault—and trained on a curated corpus of 104,880 verified question‑answer pairs with chain‑of‑thought reasoning. After supervised fine‑tuning, dynamic sampling policy optimization with task‑routed rubric rewards is used to stabilize reinforcement learning across heterogeneous telecom tasks, achieving an 89.64% mean score on seven GSMA Open Telco Leaderboard benchmarks, surpassing leading proprietary models.
By Bohao Wang, Chenwei Wu, Hang Zou, Yu Tian, Lina Bariah, Li Wei, Chongwen Huang, Yongliang Shen, Zhaoyang Zhang, Merouane Debbah
The paper introduces an expert‑annotated dataset of 42 U.S. federal tax opinions on corporate reorganizations under I.R.C. §368, marking the first tree‑structured argument corpus in this domain. Each legal passage is labeled with one of five functional categories—Rule, Analysis, Conclusion, Background Facts, and Procedural History—and can be linked into directed support trees. Experiments demonstrate that functional labels are learnable and that supervised fine‑tuning improves within‑case retrieval, though cross‑case generalization remains weak.
By Luis Brena, William Jurayj, Gregory Deyesu, Zaid Al-Huneidi, Andrew Blair-Stanek, Benjamin Van Durme
The paper introduces a four‑stage framework—SFT, PG‑CoT, Dynamic, and K‑RL—to improve large language models for Traditional Chinese Medicine prescription generation. It addresses three key gaps: lack of auditable reasoning (SR Gap), failure to adjust prescriptions over time (LA Gap), and non‑enforcement of absolute contraindication rules (SC Gap). Experiments on 12 fine‑tuned models and 6 zero‑shot baselines show that the framework, particularly a 7B Mistral model, outperforms zero‑shot GPT‑5 on all three TCM evaluation metrics.
By Zheng Chen, ZhiCheng Du, Haoxuan Li, Peiwu Qin
A new large-scale Arabic fact‑checking dataset called Arafa has been created using an automated pipeline that generates claims from Arabic Wikipedia, mutates them into counterfactuals, and validates them against supporting or refuting evidence. The dataset contains 181,976 claim‑evidence pairs labeled as supported, refuted, or not enough information, and human evaluation shows high inter‑annotator agreement and strong validation accuracy. Fine‑tuned transformer models on Arafa achieve a Macro F1‑score of 77%, demonstrating its usefulness for Arabic fact‑checking tasks.
By Christophe Khalil, Shady Elbassuoni, Rida Assaf
The paper introduces DA-Cramming, a cost‑effective pretraining method that incorporates dependency agreement information into BERT‑style language models. It builds on the Cramming technique to enable training with a single GPU in a day, using a dual‑stage workflow and four submodels to embed chunk‑level dependency agreements. Experiments show that this approach outperforms prior methods on a range of tasks.
By Martin Kuo, Jianyi Zhang, Dongting Li, Yiran Chen
The paper introduces ConRad, a reinforcement learning framework that fine‑tunes large vision‑language models to generate calibrated verbalized confidence estimates for radiology reports. ConRad offers both a single report‑level confidence score and a sentence‑level variant, trained with the GRPO algorithm and logarithmic scoring rewards to encourage truthful self‑assessment. Experiments show significant calibration improvements over existing methods, and clinical evaluation indicates that report‑level scores align well with clinicians’ judgments, enabling targeted review of low‑confidence statements.
By David Bani-Harouni, Chantal Pellegrini, Julian L\"uers, Su Hwan Kim, Markus Baalmann, Benedikt Wiestler, Rickmer Braren, Nassir Navab, Matthias Keicher
The paper introduces ImIR, a method that tunes a large pretrained image‑editing model for all‑in‑one image restoration by replacing text prompts with continuous image‑derived instructions. The approach uses a lightweight token mapper to shift the degraded image’s vision‑language embedding toward that of a clean image, enabling a single adapter to handle six restoration tasks in about three hours on one GPU. ImIR outperforms text conditioning in matched comparisons and supports task‑agnostic restoration without requiring a degradation label.
By S\"uleyman Aslan, G\"orkay Aydemir, M{\i}sra Yavuz, Yunus Bilge Kurt, Nasrin Rahimi, Ahmet Rasim Emirda\u{g}{\i}, Burak Can Biner, M. Ak{\i}n Y{\i}lmaz
The paper presents a foundation-guided auto‑annotation pipeline that improves standard autonomous driving object detectors in adverse weather. By benchmarking YOLOv8, Co‑DETR, and SAM3 on a custom dataset of 25 operational scenarios, the authors find SAM3 to be the most robust and use it offline to generate pseudo‑labels. Fine‑tuning YOLOv8 on these labels boosts overall mAP by 16.04% and yields significant gains in specific conditions such as Residential Direct Sunlight (32.73%) and Highway Fog (28.65%).
By Sepideh Gohari, Goodarz Mehr, Azim Eskandarian
The study presents an interpretable AI framework for glaucoma screening using low-cost handheld retinal cameras in a West African population. Trained on 681 participants, the system achieved high performance across vessel segmentation, cup/disc segmentation, and optic nerve head feature detection, with classification AUCs of 0.85 for the handheld device and 0.93 for a tabletop camera. The model provides confidence scores and visual explanations to aid clinical interpretation.
By Charis Y. N. Chiang, Tarela Sarimiye, Adeyinka Ashaye, Martin Buist, Michael A. Hauser, Olusola Olawoye, Micha\"el J. A. Girard
The paper presents a new monocular spacecraft pose estimation model that achieves the lowest reported mean rotation errors on the SPEED+ lightbox and sunlamp test sets. By replacing smaller encoders with a large self‑supervised ViT foundation model (DINOv3) and scaling up to 840 M parameters, the authors improve accuracy from 300 M to 840 M parameters without saturation. The 840 M model also runs on a Jetson Orin NX 16 GB with 133.8 ms per crop and 32.0 W power draw, demonstrating embedded inference feasibility while training solely on synthetic data.
By John Church, Vazghen Nikolian
arXiv:2609.25040v1 Announce Type: cross
Abstract: Banana crop diseases threaten food security across the world, yet field diagnosis remains difficult because of limited expert access and visual simil...
By Sangam Kumar Jena, Pandarasamy Arjunan
arXiv:2609.25884v1 Announce Type: cross
Abstract: Users commonly combine multiple Low-Rank Adaptation (LoRA) adapters to personalize images with different subjects, styles, and visual attributes. Yet...
By Jin Wei, Rundong Li, Ruihao Yang, Yikai Wang, Xiaoyuan Duan, Jianxiong Wu, Yanbo Wang, Chang Xu, Lingyun Zhang, Zhuyang Yu, Ping Chen, Jun Dai, Xiaoyan Sun
arXiv:2603.17043v2 Announce Type: replace
Abstract: Physics-aware multimodal large language models (MLLMs) can localize and characterize two-dimensional (2D) material flakes from optical microscopy i...
By Sankalp Pandey, Thanh-Dat Truong, Xuan-Bac Nguyen, Hoang-Quan Nguyen, Tim Faltermeier, Nicholas Borys, Hugh Churchill, Khoa Luu
arXiv:2609.24510v2 Announce Type: replace
Abstract: Recent advances in vision foundation models (VFMs) have shown remarkable capabilities across diverse unimodal visual tasks. However, adapting VFMs...
By Xiaoqiang Lu, Licheng Jiao, Lingling Li, Yuting Yang, Long Sun, Wenping Ma, Xu Liu, Fang Liu