The paper introduces an AI-assisted pipeline for retrieving goldsmith marks from silverware images, combining mark localization with metric‑learning fine‑tuning on three backbone models (ResNet‑50, ViT‑S/16, and DINOv2 ViT‑S/14). Systematic tests of cropping strategies show that manual cropping and metric‑learning fine‑tuning yield the best performance, with DINOv2 ViT‑S/14 achieving 62.63% mAP and 73.74% Top‑1 accuracy. The authors release a manually annotated dataset, code, and a public web interface to support reproducibility and adoption in digital humanities.
By Atmik Tiwari, Vincent Christlein, Mark Fichtner, Freya Gohlke, Birgit Sch\"ubel, Theresa Witting, Heike Zech, Mathias Zinnen
GAPrompt++ is a multi-granular geometry-aware prompting method designed to adapt pre-trained 3D vision models to downstream tasks efficiently. It introduces a Point Shift Prompter for multi-scale geometric feature extraction, a Keypoint Prompter for local geometric saliency, and a Prompt Propagation mechanism to embed these cues throughout the model hierarchy. Experiments demonstrate that GAPrompt++ outperforms other prompting-based PEFT methods and even surpasses full fine-tuning while using less than 2% trainable parameters, and the authors provide two new challenging benchmarks for future research.
By Zixiang Ai, Zhenyu Cui, Yufei Guo, Wenwen Qiang, Lei Chen, Jiwen Lu, Jiahuan Zhou
The paper introduces ActObs, a supervised fine‑tuning method that, unlike standard approaches, also predicts environment observations in agent trajectories. While both ActObs and action‑only training perform similarly after initial fine‑tuning, ActObs diverges during subsequent reinforcement learning, yielding higher pass@k scores on several benchmarks and better cross‑domain task performance. The authors attribute this advantage to ActObs’s joint supervision, which preserves observation gradients and prevents the policy from over‑specializing on actions alone.
By Juzheng Zhang, Disha Makhija, Manoj Ghuhan Arivazhagan, Vinayshekhar Bannihatti Kumar, Rashmi Gangadharaiah
The paper investigates an output‑selection bottleneck in frozen DETR‑family segmentation models, showing that many useful mask proposals are computed but never exposed. A lightweight selector called HYDRA, trained only on cached outputs, can recover up to +7.41 mIoU on ADE20k and COCO and +9.4 class‑macro prompt‑IoU on SAM 3 across eight domains by selectively choosing better candidates. The study demonstrates that evaluating segmenters should consider both exposed masks and the hidden candidates they suppress.
By Ignacio M. De la Jara, Cristian Rodriguez-Opazo, Damith Ranasinghe
AURA: Adaptive Uncertainty-Routed Analysis for Email Threat Detection is a multimodal system that evaluates both email content and embedded URLs to detect spam and phishing. It uses a two-layer approach: first, a URL classifier estimates prediction uncertainty, and only messages with high uncertainty are passed to a fine-tuned transformer encoder for deeper semantic analysis. Evaluated on eight diverse training corpora and two real-world datasets covering a decade of attacks, AURA achieves a macro F1-score of 0.9858 in-distribution and maintains scores above 0.94 on the NazPhish-Eval and GuenterTrap-Eval datasets, demonstrating strong generalization to new attack scenarios.
By Omran Berjawi, Walid fahs, Rida Khatoun
The paper addresses the lack of statistical validity in language‑model‑based network intrusion detection. It introduces traffic‑aware conformal prediction, which calibrates on attacker‑expected traffic to restore coverage guarantees, and further mitigates adaptive attacks by removing attacker‑controllable features, achieving exact pathwise coverage. Experiments on three benchmarks show that this approach maintains coverage while incurring a modest accuracy cost.
By Zhenpeng Li
The paper introduces Riemannian–Lorentz Parameter Fusion (RLPF), a method for merging a Vision Transformer and a state‑space model without gradient descent. RLPF aligns parameter groups by semantic role, projects them onto a common coordinate system, lifts selected coordinates to the Lorentz hyperboloid, computes a regularized geodesic barycenter, and decodes the result back into the two branches, with a learned gate combining their logits. The resulting fine‑tuned system achieves 82.37 % on CIFAR‑10, 75.04 % on Oxford‑IIIT Pet, and 78.58 % top‑1 accuracy on ImageNet‑1K, surpassing the best‑parent accuracies of 76.54 %, 71.42 %, and 76.42 % respectively.
By Badri N. Patro, Vijay S. Agneeswaran
STR-Agent is an LLM-driven framework designed for QoS-aware routing in Low Earth Orbit satellite networks. It integrates intent perception, tool-based execution, experience accumulation, and reflection-based policy adaptation to translate natural-language service requests into adaptive routing decisions. In simulations on a Walker-Delta constellation, STR-Agent reduces end-to-end delay by up to 60% compared with DQ-Dijkstra and improves intent-understanding accuracy from 45.4% to 92.45% after fine-tuning, with the Reflection Module providing additional delay reductions.
By Bowen Lu, Mugen Peng, Yaohua Sun, Hongyu Wang, Kerui Guo, Wenjia Xu
HerHealthEval is a controlled evaluation framework that tests multilingual understanding of women's-health communication across English, French, and Modern Standard Arabic. It provides six communicative forms—canonical, clinical, layperson, indirect or hedged, emotionally concerned, and deliberately under-specified—each conveying the same clinical concern except the under-specified form omits details to assess clarification needs. The study evaluates multilingual instruction models and QLoRA-adapted variants on tasks such as concern classification, risk calibration, clarification behavior, parse compliance, and cross-form consistency, revealing that high aggregate accuracy can mask safety-relevant failures and that language-invariant risk labels improve performance.
By Hassan Saeed Hassan Albattra, Mazen Mohammed Bahgat, Rahatara Ferdousi, Hana Essam Sayed Ahmed Amrya, Mariam Mousa
The paper adapts the compact PP‑OCRv6 recognizer for historical text recognition and compares it to a conventional CRNN across various training regimes, including generalized pretraining, domain‑specific training, corpus‑level fine‑tuning, and manuscript‑specific few‑shot adaptation on multilingual Latin and Arabic scripts. While PP‑OCRv6 does not always beat the CRNN when trained from scratch, heterogeneous pretraining significantly improves its generalization. Additionally, fine‑tuned PP‑OCRv6 can surpass a large vision‑language model (Qwen3.5‑based Medusa) that is specifically tailored for historical Latin‑script handwriting recognition.
By Benjamin Kiessling (ALMAnaCH)
The paper evaluates twelve financial sentiment models—including dictionary-based methods, finance-specific transformers, and open-source large language models—using linguistic and economic validity criteria. General-purpose LLMs match finance-specific transformers in classification performance but do not yield stronger economic relationships. While several models correlate with earnings surprises, none shows a significant link to next‑day stock returns, and performance is strongest for large earnings beats or misses.
By Arslan Bisharat, Oudom Hean
oMeBench is a large-scale, expert-curated benchmark designed to evaluate large language models (LLMs) on organic mechanism reasoning. It contains over 10,000 annotated mechanistic steps, including reaction type labels, intermediate structures, and difficulty ratings, and introduces the oMeS scoring framework to assess logical consistency and chemical structural similarity. Evaluation shows that while current LLMs display promising chemical intuition, they often fail to produce correct and consistent multi-step reasoning, though prompting and fine-tuning can bring smaller models up to the level of closed‑source frontier models.
By Ruiling Xu, Yifan Zhang
The paper introduces pre‑trained models for extracting variant‑phenotype relations from biomedical text, focusing on the SNPPhenA corpus. Fine‑tuning small BERT‑based models, especially DeBERTa, achieves performance close to the current state‑of‑the‑art. Moreover, careful fine‑tuning of Google’s Gemini Pro 1.0 surpasses existing benchmarks on both sentence‑level and abstract‑level relation extraction tasks.
By Claudiu Creanga, Liviu P. Dinu, Daniela Gifu
The paper introduces Live-ProsodyJudge (LPJ), a cost‑effective pairwise evaluator distilled from Gemini for assessing fine‑grained prosody in live streaming TTS. It identifies a flaw called verdict coupling, where multi‑dimensional scores collapse into a single preference, and proposes Decoupled‑Live‑ProsodyJudge (D‑LPJ) to eliminate this issue through masking and a span‑local GRPO strategy. Experiments show LPJ outperforms a single Gemini call in accuracy, and D‑LPJ provides independent dimension judgments, achieving high alignment with human top‑3 selections in a TTS candidate tournament.
By Zifan Guan, Longyu Lu, Junan Zhang, Zhizheng Wu, Meiguang Jin, Junfeng Ma
The paper investigates what gradient similarity measures in data attribution for large language models. By independently varying task and answer format in supervised fine‑tuning benchmarks, the authors show that gradient alignment is driven by answer format rather than task semantics, with strong alignment for shared formats and none for differing formats. This pattern persists across training stages, model sizes, and families, and is evident in the selections of the LESS data‑selection method, which over‑represents its own answer format.
By Sunwoo Kim, Seokwon Jung, Sohyung Kim, Seong Joon Oh, Alice Oh
EmbodiedMind introduces a three-stage training paradigm for embodied foundation models that tackles inefficient sample use, task imbalance, and credit assignment in long-horizon planning. The stages—Rejection Sampling-based Fine‑Tuning, Iterative Rejection GRPO, and Trie‑GRPO—filter low‑informative data, balance task difficulty, and use action prefix trees for step‑level advantage estimation. This approach yields a state‑of‑the‑art average performance of 70.02% across 18 benchmarks, notably improving long‑horizon task planning accuracy.
By Feifan Wang, Zongbing Zhang, Yu Zhang, Lingfeng Wang, Yurui Zhu, Jin Deng, Mingliang Zhang, Zhengguang Gao, Yongcheng Wang, Jin Xu, Ri Yang
CellRFT is a reinforcement fine‑tuning framework designed to improve single‑cell perturbation modeling by directly optimizing biological evaluation metrics. It employs policy‑gradient methods to learn from non‑differentiable biological rewards and aggregates multiple rewards hierarchically. Experiments show that CellRFT enhances perturbation prediction across various pretrained models and reveals interactions between different biological criteria, suggesting new ways to shape model behavior and evaluation design.
By Jie Yan, Li Liu, Hanze Guo, Jiaxin Hu, Houxin He, Xiaoning Qi, Haoran Wang, Cong Li, Zhong-Yuan Zhang, Yong Wang
This study evaluates the classification accuracy of six modern deep‑learning architectures—VGG19, ResNet50, GoogleNet, ConvNeXt, EfficientNet, and Vision Transformers—on breast ultrasound images categorized by BI‑RADS. Using 2,945 training images and 936 validation images from 1,540 patients, the models were tested in full fine‑tuning, linear evaluation, and training‑from‑scratch settings. The best performance was achieved with full fine‑tuning, yielding 76.39 % accuracy and a 67.94 % F1 score.
By Malitha Gunawardhana, Norbert Zolek
SlugTrails is a new egocentric benchmark for floor‑plan‑based indoor visual localization in large buildings, featuring 30 Hz Aria glasses recordings across three campus buildings and six floors. The dataset includes CAD‑derived floor plans with semantic classes, circulation masks, and laser‑surveyed anchors for trajectory alignment. Five representative systems were evaluated, showing that stock checkpoints perform poorly while fine‑tuning on SlugTrails significantly improves performance and cross‑dataset generalization, indicating that data scarcity limits current methods.
By Yunqian Cheng, Roberto Manduchi
The paper introduces PreDE, a policy‑calibrated framework that predicts how post‑training quantization will degrade task performance in world action models (WAMs) before deployment. By calibrating two thresholds on a small development set, PreDE can accept, reject, or defer new quantization configurations based on offline action deviations, achieving 75% coverage of decisions that match closed‑loop outcomes. Experiments on five WAMs and real‑robot trials show that PreDE accurately identifies high‑deviation configurations and enables significant speedups and memory reductions without compromising performance.
By Jiuyi Xu, Jinjia Guo, Meida Chen, Jing Du, Yangming Shi