ArGuard is a shared task that evaluates harmful content detection in Arabic memes and LLM prompts, featuring two tracks: Track A for multimodal hate detection in memes and Track B for harmful prompt detection in Arabic LLM safety evaluation. Fifty‑eight teams registered, 35 reached the final evaluation, and 27 submitted system‑description papers, with participants experimenting with models such as AraBERT, Jais, and Qwen3‑VL. The top systems achieved macro‑F1 scores of 0.823 on A1, 0.419 on A2, 0.984 on B1, and 0.790 on B2, with fine‑grained meme classification in A2 proving the most challenging due to sparse labels and distribution shifts.
By Firoj Alam, Md. Rafiul Biswas, Mohamed Bayan Kmainasi, Ali Ezzat Shahroor, Hamdy Mubarak, George Mikros, Abul Hasnat, Wajdi Zaghouani
Agentic-GER is an LLM-based agent designed to improve terminology accuracy in long‑form speech transcription. It leverages global context from the entire transcript to flag suspicious terms, selectively re‑transcribes the source audio to verify candidate corrections, and uses accepted edits to inform future decisions. Experiments on GigaSpeechBench with four LLMs and two ASR systems show consistent terminology improvements in both Chinese and English, achieving up to a 36.8% relative reduction in biased character error rate over the Whisper baseline for Chinese speech.
By Yanqiao Zhu, Wupeng Wang, Zhifu Gao, Xiangang Li, Xie Chen
The paper "Less is More: Encoder-only Audio-Visual Segmentation" introduces EASE, an encoder-only model for Audio‑Visual Semantic Segmentation (AVSS). EASE achieves state‑of‑the‑art accuracy while running at up to 365 FPS—about three times faster than previous Transformer‑based AVSS models—and trains in under 11 GPU‑hours. The authors demonstrate that simpler, faster architectures can match or exceed the performance of more complex models across various backbones and resolutions.
By Ilpo Viertola, Vladimir Iashin, Sophie T\"otterstr\"om, Esa Rahtu
The paper introduces Selective Probability Mass Concentration (sPMC), a training framework that strengthens implicit visual grounding in multimodal large language models by selectively regularizing attention heads most responsive to visual evidence. sPMC treats attention over visual tokens as a spatial probability distribution and encourages mass to concentrate on semantically relevant regions using segmentation-derived priors, while leaving other heads unconstrained. Across six multimodal benchmarks, sPMC yields an average zero‑shot improvement of 3% and gains up to 11.3% for various models by regularizing only 3%–15% of their attention heads.
By Jiaqi Deng, Zonghan Wu, Zhan Heng, Xiaoshui Huang, Huan Huo, Guandong Xu
This paper introduces an Edge AI system that classifies sleep and wake states on constrained devices using a multimodal pipeline on an ESP32‑S3 microcontroller. It fuses inertial head‑movement sensing with visual pose classification, running in parallel under FreeRTOS to meet real‑time constraints. The two‑stage detection achieves 96.5 % accuracy for motion‑based detection and 89 % for pose classification, proving robust binary sleep‑wake classification in mobile scenarios.
By Stefan Reitmann, Lena Oden
The paper investigates how providing execution traces to multimodal judges in agentic video‑generation systems can bias their verdicts. On a benchmark of 109 two‑event clips, traces that falsely report successful tool calls cause large‑language‑model judges to incorrectly accept 78–90 % of failures, while contradictory traces lead to 100 % rejection of correct clips. The effect persists even when judges are instructed to consider only the video frames, indicating that the vulnerability stems from the judges’ learned trust in tool logs rather than the visual content itself.
By Jian Xu
PTC-Bias is a two-stage framework that improves contextual biasing in speech large language models by using phoneme-level temporal competition. In the first stage, PTC Retrieval performs frame-synchronous phoneme decoding to generate a compact shortlist of bias words and their speech intervals. The second stage, PTC Correction, applies a local competition between retrieved candidates and mismatched transcript spans within those intervals, reducing near-homophone and word-segmentation errors without extra SpeechLLM passes. Experiments on LibriSpeech demonstrate consistent gains across two SpeechLLMs, with PTC-Bias reducing B-WER by up to 23.9% relative to CTC-Filter while keeping U-WER nearly unchanged.
By Zhiqi Ai, Han Cheng, Shiyi Mu, Yongjin Zhou, Shugong Xu
BanglaKontho is a newly released 20‑hour single‑speaker Bangla text‑to‑speech corpus derived from professional audiobook recordings, comprising 7,050 segmented utterances with verified transcripts at 24 kHz. The project also provides a reusable Bangla text normalizer that handles Bangladeshi‑style digit grouping, currency and date expressions, Danda punctuation, and Unicode normalization, along with the full preprocessing pipeline. A baseline MB‑iSTFT‑VITS model trained from scratch on this corpus achieves a 9.5 % WER and 4.46 naturalness MOS, outperforming the same architecture retrained on the smaller 12‑hour IndicTTS‑Bn corpus.
By Mizbaul Haque Maruf
YODAS v3 is a weakly‑labeled speech corpus that offers more than 1.1 million hours of 48 kHz multi‑channel audio across 147 languages, making it the largest open speech dataset available and the first large‑scale collection with high‑fidelity stereo audio. The authors detail a new collection methodology that balances language representation, achieving 22 languages with over 10 k hours and 73 languages with over 5 k hours of data. They also analyze language, audio, and transcription quality, and demonstrate the dataset’s utility by training baseline speech‑recognition and neural‑codec models.
By William Chen, Shinnosuke Takamichi, Sayaka Shiota, Satoru Fukayama, Samuele Cornell, Shinji Watanabe
The paper investigates how the order of generating explanations—whether a rationale is produced before or after the answer—affects vision‑language reasoning. By conducting controlled experiments on knowledge‑intensive QA, visual entailment, and compositional grounding tasks, the authors show that larger models are required for reliable rationale‑first generation, while answer‑first generation is less susceptible to format errors. The study concludes that explanation ordering, model scale, pre‑training knowledge, fine‑tuning, and task structure jointly influence prediction accuracy and reasoning faithfulness.
By Siting Liang, Luca Rippe, Omar Adjali, Daniel Sonntag
The paper introduces task-informed parameter-efficient fine-tuning methods for low-resource speech recognition by applying Fisher-Whitened Cross-Covariance Analysis (FCCA) to Whisper and Qwen3-ASR. Two extensions—Asymmetric-Coupled FCCA (AC‑FCCA) and Adaptive‑Rank FCCA (AR‑FCCA)—are proposed to exploit cross‑layer sharing and adapt rank allocation within a fixed parameter budget. Experiments on multilingual datasets show that standard FCCA matches or surpasses LoRA, while AR‑FCCA consistently improves performance across models without increasing trainable parameters.
By Asmee Mishra, Mengjie Qian, Brechtje Post, Kate Knill
The paper introduces a token‑level extension of Omni‑Temporal Classification (OTC) for automatic speech recognition, allowing unsupported tokens to be bypassed while preserving supervision for the rest of the word. Across 19 languages and three corpora, this token‑level OTC consistently outperforms standard CTC, achieving the lowest mean word error rate on every dataset and a 9.45% average relative WER reduction. A predictive‑entropy‑indexed schedule replaces epoch‑based relaxation, reducing training‑length dependence while maintaining performance.
By Saurabh Kumar, Diptiman Mohanta, Prasanta Kumar Ghosh
The paper introduces a perception interface that separates vision from language in vision‑language models. A frozen perception stack detects objects, a deterministic semantic serializer converts the perceived state into text, and a standard text‑only large language model (LLM) answers questions. Experiments on a campus‑robot benchmark show that this serialized interface outperforms a zero‑shot VLM of the same language‑model size, especially as the language model shrinks, and that the advantage persists under paraphrase and different supervision regimes.
By Cong Xu, Ravi Sankar
M3GD introduces a multimodal representation that fuses pre‑trained 2D image and 3D LiDAR foundation models for robotic novel view synthesis, avoiding the need for a separate cross‑modal translator. By projecting LiDAR onto the image latent grid and injecting the resulting geometry‑aware packets via a lightweight residual adapter, the method enhances both RGB and depth synthesis on the GrandTour dataset compared to an image‑only baseline. Ablation studies confirm that pixel‑aligned LiDAR content drives the performance gains, and real‑world deployment on a ground robot demonstrates a tunable quality–cost trade‑off.
By Yang Zhou, Jiuhong Xiao, Shizhao Ye, Long Quang, Carlos Nieto-Granda, Giuseppe Loianno
OptiSAR-Net++ introduces a new cross‑domain remote sensing visual grounding task (CD‑RSVG) and the first large‑scale benchmark dataset, OptSAR‑RSVG. The framework replaces Transformer decoding with a CLIP‑based contrastive approach, employing a patch‑level Low‑Rank Adaptation Mixture of Experts for efficient cross‑domain feature decoupling and a text‑guided dual‑gate fusion module for improved semantic‑visual alignment. Experiments show state‑of‑the‑art performance on OptSAR‑RSVG and DIOR‑RSVG, with notable gains in localization accuracy and computational efficiency.
By Xiaoyu Tang, Jun Dong, Jintao Cheng, Rui Fan
The paper introduces LowBridge, a method for cross‑modal medical image segmentation that leverages shared low‑level features such as edges between MRI and CT scans. It trains a generative model to reconstruct source‑modality images from edge maps and then trains a segmentation network on these generated images. At test time, edge features from target‑modality images are fed into the generative model to produce source‑style images, which are segmented by the pretrained network, achieving state‑of‑the‑art results on multiple public datasets.
By Pengfei Lyu, Pak-Hei Yeung, Jing Xia, De Hu, Xiaosheng Yu, Jianning Chi, Chengdong Wu, Jagath C. Rajapakse
The paper presents a 3D foundation model for light sheet fluorescence microscopy (LSM) that is pretrained on a large curated set of 3D images from various organisms, stains, and imaging protocols. By jointly optimizing for masked reconstruction and image‑text alignment, the model learns transferable volumetric representations that dramatically reduce the need for annotated data. The pretrained backbone enables efficient few‑shot adaptation to downstream tasks such as segmentation, classification, and deblurring, consistently outperforming baselines according to standard metrics and expert evaluation.
By Adina Scheinfeld, Haotan Zhang, Shang Mu, Rudolf L. M. van Herten, Lucas Stoffl, Ali Erturk, Zhuhao Wu, Johannes C. Paetzold
VietPrism is a newly released, large‑scale Vietnamese speech corpus that combines 993.4 hours of real utterances from 1,262 verified speakers with 3.1 k hours of synthetic spoof speech. It uniquely offers transcripts, consistent speaker identities, five dialect groups, and extensive Vietnamese‑English code‑switching—nearly half of the corpus—while pairing each spoof with a matched bona fide utterance. The dataset enables controlled evaluation of deep‑fake detection models, revealing significant variability in detector performance across dialects and speaker similarity.
By Minh Hoang, Thai Le
The paper introduces Smol‑VL‑BLV, a compact vision‑language model designed for blind and low‑vision users. It employs a 500M decoder transformer with teacher‑student distillation and Group Relative Policy Optimization to add spatial detail, directional cues, and hazard detection to post‑training. After a lightweight finetuning step, the model achieves significant gains on spatial, social, OCR, and VQA benchmarks while remaining under 450 MB and running entirely offline on a mid‑range Android phone.
By Rishabh Choudhary, Shreyansh Raj, Umesh Goyal, Shubh Kashyap, Shrestha Kumar, Sushovan Jena, Komal Kumar, Hisham Cholakkal, Aditya Nigam
RotVLA introduces a Vision‑Language‑Action framework that replaces discrete latent action encoding with a continuous rotational latent action representation on the group SO(n). This design provides continuity, compositionality, and structured geometry that better capture real‑world action dynamics, and a triplet frame learning scheme enforces meaningful temporal dynamics while preventing degeneration. Trained with 1.7 B parameters on large cross‑embodiment datasets, RotVLA achieves state‑of‑the‑art performance on LIBERO and RoboTwin2.0 benchmarks and shows strong real‑world manipulation results.
By Qiwei Li, Xicheng Gong, Xinghang Li, Peiyan Li, Quanyun Zhou, Hangjun Ye, Jiahuan Zhou, Yadong Mu