EXPL-FR is a lightweight adapter that aligns a vision‑language model’s image encoder with a frozen face‑recognition (FR) embedding space, enabling the FR model to be explained using semantic attribute prompts without any text training. By mapping 978 attribute prompts across 22 categories into the FR space, the method identifies the most detectable concepts—forming a readable semantic signature that better separates identities than the full vocabulary. The approach is evaluated on four FR backbones and two VLM encoders, providing identity‑level, per‑image, and differential explanations, and demonstrates that prompt‑driven audits can rank FR models by per‑ethnicity error and attribute‑change verification cost without requiring labeled data.
By Guray Ozgur, Mustafa Efe Tamyapar, Naser Damer, Fadi Boutros
The paper introduces LaST, a large-small collaborative framework for zero-shot surgical phase recognition. It combines a foundation model that generates frame-level phase priors with a lightweight model that refines predictions through iterative temporal refinement, dynamic quality control, and dual-model cross-learning. Experiments show LaST outperforms baseline and state-of-the-art methods, achieving significant accuracy gains on unseen clinical domains.
By Yiyi Zhang, Ying Zheng, Wenxin Fan, Yu Zhu, Yuchen Yuan, Litao Zhao, Zheng Li, Pheng-Ann Heng
VISTA is a gradient‑based test‑time alignment framework designed for next‑scale visual autoregressive (VAR) image generation. It optimizes intermediate representations within the frozen transformer to enforce compositional constraints, without altering model weights or requiring extra training. Experiments on two benchmarks and two model scales show that VISTA improves compositional accuracy by up to 20% on a 2B backbone and 6% on an 8B backbone, while preserving image quality and enabling a smaller model to outperform a larger one.
By Hossein Shahabadi, Niki Sepasian, Mahdieh Soleymani Baghshah
arXiv:2604.09746v2 Announce Type: replace-cross
Abstract: As large language models (LLMs) are increasingly deployed as autonomous agents, understanding how strategic behavior emerges in multi-agent e...
By Aarush Sinha, Arion Das, Soumyadeep Nag, Charan Karnati, Shravani Nag, Chandra Vadhan Raj, Aman Chadha, Vinija Jain, Suranjana Trivedy, Amitava Das
arXiv:2608.22788v1 Announce Type: new
Abstract: Large-scale rollouts have become a core component of modern LLM systems, spanning reinforcement learning (RL) post-training, on-policy distillation (OP...
By Tianqi Xu, Lu Lv, Haoyang Huang, Wenjie Huang, Zhanming Shen, Yuhao Shen, Baolin Zhang, Xinyi Hu, Shuang Ge, Jun Dai, Tianyu Liu, Suorong Yang, Zhikai Li, Ye Bai, Jun Zhang, Lei Chen, Yue Li, Mingchen Wan
The paper introduces Protected QL Memory, a dual‑path document classifier that separates a causal local pathway from an associative matrix writer and a finalized memory reader. It evaluates three empirical properties—capability preservation, diagnostic access, and task performance sensitivity—across nine dataset‑seed conditions, finding that writer capability is fully preserved while diagnostic accuracy gaps are large but macro‑F1 scores change minimally. The study concludes that preserving associative capability and diagnostic access does not guarantee downstream reliance on memory, urging separate evaluation of these objectives.
By Isaac Kofi Nti
The paper introduces a finite‑width geometric framework that explains how learned feature geometries are organized, transported, and selectively aligned in deep neural networks. It quantifies incompatibilities among weight‑generated covariance, gates, and backward sensitivities using three families of commutators, and provides exact layerwise identities that decompose these commutators into sources such as downstream transport, adjacent‑layer imbalance, and nonlinear gate‑covariance interactions. The study demonstrates that spectral alignment is a layer‑ and scale‑dependent compatibility phenomenon governed by transport, interaction, cancellation, and possible damping, rather than a universal consequence of training.
By Kaj Nystr\"om
arXiv:2608.22960v1 Announce Type: new
Abstract: Coding agents are increasingly evaluated not only by whether they solve a task, but also by how they execute it. However, existing process-level evalua...
By Jiawei He, Mengyu Shi, Jie jia, Xikai Yang, Dong Sun
arXiv:2607.27910v2 Announce Type: replace
Abstract: Inference time defences against vision language model jailbreaks often subtract a calibrated direction from the residual stream at a chosen decoder...
By Xiangyu Yin, Tora Bodin, Rohan Menon, Chih-Hong Cheng
arXiv:2503.05684v2 Announce Type: replace-cross
Abstract: Pre-trained foundation models can be efficiently adapted for specific tasks using Low-Rank Adaptation (LoRA), but the fairness properties of...
By Parameswaran Kamalaruban, Mark Anderson, Stuart Burrell, Maeve Madigan, Piotr Skalski, David Sutton
arXiv:2608.21501v1 Announce Type: new
Abstract: Credit assignment in large-language-model reinforcement learning (LLM RL) can be separated into three objects: evidence about success, a transport oper...
By Qifan Shi, Zhaolu Kang, Chenghua Zhu
arXiv:2608.21985v1 Announce Type: new
Abstract: As the adoption of large language models (LLMs) grows in Arabic-speaking regions, ensuring their safety and cultural alignment is increasingly critical...
By Fidaa Abed, Haidar Khan, M Saiful Bari, Babar Khan, Abdalghani Abujabal
arXiv:2605.11533v4 Announce Type: replace
Abstract: Routine clinical check-up reports combine laboratory measurements, physiological assessments, imaging findings and visually structured information,...
By Sike Xiang, Shuang Chen, Kevin Qinghong Lin, Jialin Yu, Yijia Sun, Philip Torr, Amir Atapour-Abarghouei
SketchFlow is a new generative framework for creating high‑quality vector sketches from text prompts. It uses a Gaussian Mixture Model prior in the CLIP latent space and an Optimal Transport Conditional Flow Matching model to map this prior to sketch features, which are then decoded by a Hybrid Diffusion Decoder combining 1D U‑Net and Transformer architectures. The approach achieves superior visual quality and human‑like drawing styles, and supports zero‑shot synthesis for unseen concepts and smooth semantic interpolation.
By Jin Zhou, Hongliang Yang, Pengfei Xu, Hui Huang
arXiv:2608.23476v1 Announce Type: new
Abstract: Narrow fine-tuning on small, domain-specific datasets can produce broad and surprising changes in model behavior-a phenomenon called weird generalizati...
By Miriam Wanner, Mark Dredze, William Walden
The paper investigates how annotator-style rebuttals can manipulate large language model (LLM) moderation systems, either by whitewashing hateful content as normal or smearing normal content as hateful. Using a rejudge protocol that adds decision‑boundary perturbations and adversarial rationales, the authors show that such rebuttals significantly degrade moderation performance, especially in multi‑turn settings. The study finds consistent, model‑specific asymmetries between the two manipulation directions and demonstrates that explicit reasoning prompts and defensive instructions mitigate but do not eliminate the vulnerability.
By Junyu Lu, Kaiyuan Liu, Jingyi Kang, Deyi Ji, Hailong Zhang, Lanyun Zhu, Qi Zhu, Bo Xu, Liang Yang, Hongfei Lin
arXiv:2512.17730v2 Announce Type: replace
Abstract: Detectors of AI-generated images tend to inherit the biases of the data they are trained on: models fitted to GAN imagery learn to treat GAN-specif...
By Yichen Jiang, Mohammed Talha Alam, Sohail Ahmed Khan, Duc-Tien Dang-Nguyen, Fakhri Karray
arXiv:2608.21832v1 Announce Type: new
Abstract: Computer-use agents ground natural-language instructions in screenshots to locate interface elements, yet existing benchmarks do not isolate whether mo...
By Md Abrar Jahin, Md Rizwan Parvez
The paper introduces CLAW-4L, a benchmark of 300 pairs of English and non‑English Wikipedia biographies (French, Chinese, Azerbaijani) focused on women from non‑English contexts, complete with claim annotations and a fine‑grained claim‑pair relation corpus. It proposes a claim‑based enrichment framework that extracts claims from both biographies, aligns them to identify enrichment evidence from the non‑English version, and rewrites the English biography accordingly. Experiments demonstrate that non‑English Wikipedia biographies can improve English biography coverage, though lower‑resource settings still pose challenges.
By Yifei Song, Ziyang Chen, Emil Sayilov, Claire Gardent
arXiv:2606.21596v2 Announce Type: replace
Abstract: Recent image-to-3D scene methods recover high-fidelity 3D objects with plausible arrangements, but often leave floatings and interpenetrations that...
By Haodong Li, Lulu Shao, Haolin Lu, Yu Fu, Yen-Ru Chen, Seemandhar Jain, Manmohan Chandraker