The paper reports a new phenomenon in CLIP embeddings where human and AI‑generated paintings naturally separate along dominant principal directions without any supervised training. The authors investigate this separation by linking embedding directions back to image features using interpretable representations and gradient‑based inversion, finding that the separation is driven by distributed multiscale image structure rather than simple global or local statistics. They also show that small, imperceptible image perturbations can cause large displacements along these directions, highlighting a mismatch between CLIP’s visual evidence and human perception.
By Andrea Asperti
The paper compares encoder‑based and generative decoder‑based large language models for evaluating automatic speech recognition (ASR). It examines BERTScore and SemDist across various LLMs, layers, and pooling strategies, finding that both metrics can strongly correlate with human judgments when properly configured. For generative LLMs, the study explores pairwise hypothesis selection via prompting and direct error classification, showing that while encoder‑based metrics remain competitive, generative models excel in hypothesis comparison and enhance interpretability of ASR evaluation.
By Thibault Ba\~neras-Roux, Shashi Kumar, Driss Khalil, Sergio Burdisso, Petr Motlicek, Shiran Liu, Mickael Rouvier, Jane Wottawa, Richard Dufour
The paper introduces AWM, a framework that treats the terminal working memory of long‑document VQA agents as an answerable evidence artifact. It proposes a memory‑only answerability diagnostic and incorporates this signal into the GRPO reward, giving higher advantage to trajectories whose final memory can answer the question alone. Experiments on MMLongBench‑Doc and LongDocURL show that AWM‑GRPO boosts final‑answer accuracy by up to 11.9 points and reduces the rate of correct answers that cannot be supported by memory alone.
By Dongzhuoran Zhou, Yuqicheng Zhu, Yule Liu, Zhen Yang, Rui Lu, Yuxiao Dong, Jie Tang, Evgeny Kharlamov
Corpus2Skill is a retrieval architecture that transforms an enterprise knowledge base into a hierarchical skill directory, enabling an LLM agent to navigate from high-level summaries to specific documents and backtrack when necessary. On an enterprise customer‑support benchmark, it outperforms single‑shot dense, hybrid, hierarchical‑retrieval, and agentic RAG baselines in answer quality and grounding, with a moderate cost tradeoff. An eleven‑dataset study shows that corpus navigation excels on single‑domain corpora with a recoverable topical taxonomy but is less effective on open‑domain factoid pools or homogeneous‑tabular corpora, providing a design guideline for knowledge‑grounded systems.
By Yiqun Sun, Pengfei Wei, Lawrence B. Hsieh
The paper evaluates privacy risks in graph neural networks enhanced by large language models (LLMs). Using a five‑stage framework, the authors test six real‑world text‑attributed graph datasets with 42 model configurations and six privacy attack methods across link, label, and membership inference threats. Results show that LLM‑enhanced GNNs are more vulnerable than shallow baselines, with semantic enrichment amplifying exploitable signals, and that differential privacy can reduce risk but at a significant cost to utility.
By Longzhu He, Zelang Wen, Chaozhuo Li, Sen Su
The paper introduces VDAR-Router, a routing framework for large language models that uses verbalized query difficulty analysis to guide model selection. It first generates an explicit difficulty profile for each query, retrieves historical examples with similar profiles, and then estimates model suitability to choose a model based on a reward function balancing performance and cost. Experiments on three datasets show that VDAR-Router consistently outperforms existing baselines in cost‑performance trade‑offs, and case studies confirm that explicit difficulty analysis improves example relevance and routing reliability.
By Yu-Chien Tang, Jun-Chen Hung, Wen-Chih Peng, An-Zi Yen
The paper examines a multimodal approach that combines a self‑supervised GNN encoder with an alternating optimization scheme involving a language‑model teacher. Despite the expectation that this joint strategy would enhance predictive performance, the authors find that the combined model fails to deliver significant gains. They identify six key factors—ranging from anchor strength trade‑offs to misaligned representation spaces—that explain why the integration of text knowledge does not fully benefit graph learning.
By Fumiaki Kimino (SOKENDAI), Ryoma Sato (SOKENDAI, National Institute of Informatics)
E2Rank (Efficient Embedding-based Ranking) is a unified framework that extends a single text embedding model to perform both retrieval and listwise reranking. By treating the listwise prompt—constructed from the query and its top‑K candidates—as a pseudo‑relevance feedback query, E2Rank reranks via cosine similarity against precomputed document embeddings, avoiding costly autoregressive decoding. The approach achieves state‑of‑the‑art results on BEIR, competitive performance on the reasoning‑intensive BRIGHT benchmark, lower latency than existing LLM‑based rerankers, and improved embedding performance on MTEB—all within a single model.
By Qi Liu, Yanzhao Zhang, Mingxin Li, Dingkun Long, Pengjun Xie, Jiaxin Mao
The paper introduces a model‑agnostic open‑set detection framework for air‑to‑air visual object detection on UAVs, addressing the limitations of closed‑set detectors under domain shifts and flight data corruption. It estimates semantic uncertainty through entropy modeling in the embedding space and employs spectral normalization and temperature scaling to improve open‑set discrimination. Experiments on the AOT aerial benchmark and real‑world flight tests show up to a 10% relative AUROC improvement over standard YOLO detectors, with background rejection further enhancing robustness without sacrificing accuracy.
By Spyridon Loukovitis, Anastasios Arsenos, Vasileios Karampinis, Athanasios Voulodimos
The paper introduces Hessian-free high-resolution (HFHR) dynamics, an extension of underdamped Langevin dynamics that incorporates reversible position diffusion for sampling in machine learning. It provides an explicit quantitative contraction rate under a position Poincaré inequality, weighted Hessian and Laplacian bounds, and a compact Sobolev embedding, even when the potential is non‑convex. For the HFHR Monte Carlo algorithm, a path‑space Girsanov argument yields a non‑asymptotic convergence bound and an explicit iteration complexity in total variation distance, improving on previous HFHR results and demonstrating benefits of a positive diffusion parameter through numerical experiments.
By Wujun Lv, Xiaoyu Wang, Yingli Wang, Lingjiong Zhu
Randomized sketch-and-solve algorithms accelerate overconstrained $\ell_2$ regression by replacing the input with a smaller problem. Standard subspace embeddings guarantee that the cost of the regress...
Bike-sharing systems are an important component of low-carbon urban mobility, but continued expansion creates challenges in both cold-start prediction and equitable resource allocation. Newly deployed...
This paper proposes a framework for constructing a classifier as a safeguard layer, and for developing a complementary diagnostic that identifies which of the classifier's confident decisions can be t...
Every memory-based knowledge editor in the SERAC lineage depends on a scope decision: given a query, does a stored edit apply? We report that current knowledge-editing benchmarks cannot measure this d...
Traffic Anomaly Understanding (TAU) requires models and systems to detect, reason about, and explain anomalous events in transportation videos. To address this challenge, we propose TAU-Agent, an agen...
Hosted retrieval-augmented generation (RAG) and semantic search allow users to query valuable provider-held corpora, raising two competing demands: to hide each query and chosen result, yet reveal onl...
The article "How Does a RAG Reranker Really Work?" explores the inner workings of Retrieval-Augmented Generation (RAG) rerankers, focusing on how data scientists explain the model’s operations behind the scenes. It discusses the impact of these insights on architecture decisions within enterprise document intelligence, specifically in the context of Enterprise Document Intelligence Vol.1 #2D. The piece highlights the importance of transparent model explanations for effective enterprise RAG implementation.
By Kezhan Shi
Finger vein verification is a promising biometric modality for secure authentication because vascular patterns are internal, difficult to observe externally, and relatively resistant to presentation a...
Multi-vector representations have emerged as an effective paradigm for multimodal retrieval, representing each sample with multiple complementary embeddings to capture fine-grained cross-modal informa...
The study compares a custom capital gains calculation engine with a retrieval‑augmented generation (RAG) vector store of market advisory reports in a multi‑agent financial advisory system. A 2x2 factorial experiment showed that the tax‑optimization engine significantly reduced tax savings, while the RAG component had no significant effect. The RAG‑only condition yielded the highest tax savings, suggesting that pretrained language model knowledge may suffice for tax‑loss harvesting without specialized tooling.
By Aryan Brar, Justin Du, Avery Lor, Kylie Seto, Eric Taylor