The paper introduces Dual-Embedding Watermarking (DEW), a semantic watermarking technique for large language models that combines contextual and token-level embeddings. DEW applies algebraic vector-space operations to generate a watermark signal that remains robust to paraphrasing and translation, while obfuscating the signal with pseudo-random matrices seeded by a secret key. Experiments demonstrate state‑of‑the‑art robustness, especially against translation, with minimal computational overhead and preserved text quality at lower watermark strengths.
By Jonas Sch\"afer, Cezary Pilaszewicz, Gerhard Wunder
The paper introduces FailSAE, a method that uses Sparse Autoencoders to predict failures in vision‑language models (VLMs) such as CLIP. By treating failure prediction as a classification over sparse SAE latent activations and employing a three‑stage training pipeline, the approach yields higher prediction accuracy than existing confidence‑score or auxiliary‑classifier baselines. Analysis shows that the SAE captures class‑specific concepts and reveals a shift toward ambiguous or style‑related concepts during failures, offering insights for runtime failure recovery.
By Jie Ma, Zongxi Liu, Yi Zhu
BeaconKV is a training‑free key‑value cache compression technique for Large Reasoning Models that uses beacon queries—compact representatives of query clusters—to predict which KV pairs will be revisited during long‑horizon reasoning. By focusing on Thought Revisiting Tokens that re‑attend distant context, BeaconKV reduces memory usage up to 5.8× and improves throughput by over 4.3× while largely preserving cache accuracy across multiple open‑source LRMs and reasoning benchmarks.
By Janghyeon Kim, Minsoo Kim, Kyuhong Shim, Jungwook Choi
The paper introduces Prompt2Box, a method that embeds prompts into a box embedding space to capture both semantic similarity and specificity relations, addressing the limitation of traditional vector embeddings that conflate topical similarity with specificity. Using a trained encoder on existing and synthesized datasets, Prompt2Box achieves significant improvements, reducing specificity prediction error by 45% over a prompt-length baseline and identifying 13.5% more LLM weaknesses in hierarchical clustering compared to vector baselines. The authors also present a novel dimension‑reduction technique for visualizing and comparing box embeddings, and provide the code on GitHub.
By Neeladri Bhuiya, Shib Sankar Dasgupta, Andrew McCallum, Haw-Shiuan Chang
CrossDepth introduces geometry-constrained attention for multi-view surround depth estimation, addressing cross-image inconsistencies caused by varying camera intrinsics and limited receptive fields. The method conditions features on per-pixel camera-aware ray embeddings and extends pixel context via cross-image attention limited to geometrically plausible regions. Trained self-supervised with photometric consistency, it achieves better depth accuracy and consistency on DDAD and nuScenes compared to existing self-supervised approaches.
By Samer Abualhanud, Max Mehltretter
The paper introduces a Nepali Question‑Answer dataset focused on passport‑related FAQs to support information retrieval in a low‑resource language. The authors fine‑tune transformer‑based embedding models for semantic similarity and compare them against the BM25 baseline. Their experiments show that fine‑tuned SBERT models outperform BM25, while multilingual E5 embeddings achieve the best overall retrieval performance.
By Funghang Limbu Begha, Praveen Acharya, Bal Krishna Bal
AngelFingerprint introduces a watermarking framework for text-guided image editing that embeds the editing prompt’s CLIP text embedding directly into the diffusion model’s weights via a LoRA module. The watermark is recoverable from image pixels alone, providing an explainable trace of the edit while remaining stealthy even under full white-box access. Experiments on the MagicBrush dataset show the extractor achieves 86% top‑1 accuracy in 200‑way prompt retrieval, outperforming prompt inversion methods.
By Bo-Han Kung, Futa Waseda, Ching-Chun Chang, Isao Echizen, Shang-Tse Chen
The paper introduces Intrinsic Temporal Adaptation (ITA) for Partially Relevant Video Retrieval (PRVR), a task that seeks untrimmed videos containing moments relevant to a text query. ITA employs a Backbone-Internal Temporal Adaptation that lets the final visual transformer layers attend to neighboring frames, creating temporally aware embeddings while keeping CLIP frozen. Additionally, an Affinity-Weighted Gradient Propagation technique softly aggregates top‑k frames based on text‑frame affinities to better handle the weakly supervised nature of PRVR, leading to state‑of‑the‑art performance and more accurate frame‑level evidence retrieval.
By Hyun Seok Seong, Woojin Jun, SuBeen Lee, Jae-Pil Heo
The paper develops a theory for relocating a finite number of compact sets in ℝ^n to arbitrary target domains using diffeomorphisms of ℝ^n. It proves that any such collection can be embedded differentiably into ℝ^{n+1} so that the images become linearly separable. The authors apply this result to show that compact datasets in ℝ^n can be made linearly separable by width‑n deep neural networks with Leaky‑ReLU, ELU, or SELU activations, and that mutually disjoint compact datasets can be separated in ℝ^{n+1} by a width‑(n+1) DNN.
By Xiao-Song Yang, Xuan Zhou, Qi Zhou
GLASS is a graph‑level anomaly detection framework that aligns graph and language representations on a unit hypersphere to achieve cross‑domain transferability. It constructs a Graph Descriptor Prompt to encode local, global, and semantic graph properties, and uses a multi‑slice soft cosine objective to unify graph and text embeddings. Anomaly scoring is performed via spherical density estimation with von Mises‑Fisher kernels, enabling zero‑shot detection and few‑shot adaptation across twelve benchmarks and three meta‑domains, outperforming recent GLAD baselines.
By Xudong Wang, Chris Ding, Tongxin Li, Jicong Fan
The BIT.UA team from the University of Aveiro participated in the 14th BioASQ Task B challenge, presenting a refactored modular pipeline for biomedical question answering. They replaced the PyTerrier PISA index with PostgreSQL-based pg_textsearch for BM25 retrieval and adopted Qdrant for dense embedding indexing, while also exploring HyDE-based query expansion and a Context-1 retrieval strategy. For answer generation, they introduced an LLM-as-a-judge framework and an agent quorum mechanism that allows multiple agents with diverse prompts to debate and converge on a consensus answer, achieving competitive MAP ranks of 5 in Phase A batches.
By Andr\'e Ribeiro, R\'uben Garrido, Alexander Christiansen, Richard A. A. Jonker, S\'ergio Matos
ConfRAG introduces a confidence-guided approach to reduce hallucinations in large language models and selectively trigger Retrieval-Augmented Generation (RAG) only when the model is uncertain. The ConfQA fine‑tuning strategy trains the model to answer correctly or respond with "I am unsure," achieving a drop in hallucination rates from 20‑40% to below 5% across factuality benchmarks. Building on ConfQA, ConfRAG limits external retrievals by more than 30% while maintaining over 95% accuracy in ideal scenarios.
By Yin Huang, Yifan Ethan Xu, Kai Sun, Vera Yan, Alicia Sun, Haidar Khan, Jimmy Nguyen, Jingxiang Chen, Mohammad Kachuee, Zhaojiang Lin, Yue Liu, Aaron Colak, Anuj Kumar, Wen-tau Yih, Xin Luna Dong
The paper introduces a multi‑resolution framework for interpreting human behavioral traces in workplace agents, arguing that a single summary is insufficient. It defines a hierarchy of semantic operators, motifs, episodes, and day‑level rhythms, and applies this to 667 million events from a commercial productivity suite, yielding 120 operator types, thousands of motifs, 25 episode types, and five rhythm archetypes. Validation shows the taxonomy is structurally stable and improves next‑episode prediction by 17% macro‑F1 over a flat baseline, while ablation experiments demonstrate that different questions require different temporal resolutions.
By Lin Ai, Scott Counts
The paper investigates the energy landscape of the t‑SNE algorithm, highlighting its non‑convexity and the resulting difficulty in understanding its behavior. It demonstrates that for a broad class of t‑SNE‑related energies and symmetric data densities, there exist infinite families of distinct critical points that preserve discrete symmetries in both feature and embedding spaces. These critical configurations explain empirical observations such as topology breaking and spurious clustering, and the authors support their claims with numerical and analytical examples.
By Nakul Haridas, Ryan Murray
AutoLR is an autonomous harness designed to streamline the iterative research‑and‑engineering cycle for industrial recommender systems, exemplified by NetEase’s gaming‑community app DASHEN. It integrates a multi‑expert council for adversarial review, a deterministic evidence‑weighted selector to allocate trial budgets, and a layered knowledge system that fuses external research with domain‑specific insights and empirical evidence. Large language model agents handle semantic reasoning and code generation, while deterministic controllers maintain control over execution, metrics, guardrails, and state management.
By Qi Zhang, Yanlin Chen, Wenchao Xiao
The paper investigates whether pretrained image models can generalize to unseen datasets by clustering their embeddings. Using encoders trained only on ImageNet‑1k, both supervised and self‑supervised, the authors evaluate clustering performance on out‑of‑domain images. They find that supervised encoders perform better within the training domain, while self‑supervised encoders excel far outside it, and that fine‑tuning self‑supervised models reverses this trend. Additionally, the study shows that the silhouette score in UMAP‑reduced space correlates strongly with clustering accuracy, offering a proxy metric when labels are unavailable.
By Scott C. Lowe, Joakim Bruslund Haurum, Sageev Oore, Thomas B. Moeslund, Graham W. Taylor
UniMate is a unified foundation model that generates articulated motion for any skeleton from a rigged 3D asset and a text prompt, eliminating the need for test‑time optimization or per‑skeleton retraining. It uses a topology‑aware diffusion transformer that incorporates skeletal topology through graph‑aware attention bias, spectral rotary position embedding, and a global topological conditioner. Trained on the newly curated UniML3D dataset of 13,006 diverse motion sequences, UniMate outperforms existing baselines in quality, generalization, and efficiency, and supports zero‑shot cross‑topology transfer, in‑betweening, expansion, and text‑guided editing.
By Linzhan Mou, Jiahui Lei, Zhiyang Dou, Chenyue Cai, Chaoyue Song, Adam Finkelstein, Szymon Rusinkiewicz
The paper introduces CGM-Rec, a continual graph memory framework designed for adaptive recommendation in the presence of intent drift. CGM-Rec treats the knowledge graph as writable memory, comprising a conservative Semantic Graph Memory for stable relational knowledge and a fast-reactive Episodic Lesson Memory for recent outcomes and corrective hints. Experiments show that, with frozen model parameters and one-pass reranking, CGM-Rec outperforms neural and LLM-based baselines across multiple recommendation settings, achieving significant gains such as a 29.58% improvement in HR@1 over the strongest LLM baseline on Bundle.
By Hao Nguyen Ngoc, Tung Nguyen, Nguyen Thi Hanh, Hoang Thai Dinh, Nguyen Xuan Tung
The paper introduces FedIoC, a federated learning framework that embeds structured threat indicators into gradient updates using a supervised contrastive loss. By aligning gradients from clients that share indicators for the same attack campaign, the server can cluster updates via cosine similarity to recover global campaign patterns without transmitting sensitive indicators. Experiments on two public threat‑detection benchmarks show that the server successfully identifies cross‑organizational campaign cohorts from fragmented local data.
By Manuel R\"oder, Bibin Babu, Frank-Michael Schleif
NS-ST-GraphRAG is a neuro‑symbolic spatio‑temporal GraphRAG framework designed to process long‑form literary narratives by integrating ontology‑guided extraction, deterministic constraint checking, dual temporal coordinates, spatial scene attributes, and dynamic sub‑graph retrieval. It selects the appropriate graph state based on the temporal and spatial scope of a query, grounding generated answers in traceable evidence. The authors also introduce Red‑Chamber‑QA, an open multi‑hop question‑answering benchmark for classical Chinese literature, and report that NS‑ST‑GraphRAG outperforms a frozen‑window baseline and a closed‑book model on a held‑out 120‑question split.
By Zheng Kui Lin