Retrieval pipelines, vector search, chunking and reranking: how models are grounded in a corpus instead of their weights.
arXiv:2609.14245v1 Announce Type: cross Abstract: Context compression reduces generator input in retrieval-augmented generation, but answer quality alone does not characterize citation attribution. W...
arXiv:2609.14437v1 Announce Type: cross Abstract: Deepfake detection systems often exhibit significant performance degradation when deployed on unseen manipulation methods, limiting their reliability...
The paper introduces DiCEf, a variant of the DiCE counterfactual example generator that incorporates expert knowledge expressed as a fuzzy linguistic vocabulary. By embedding continuous data into this linguistic domain, DiCEf personalises counterfactual explanations to be linguistically perceptible while maintaining minimal cost, sparsity, and diversity. Experimental results on a real‑world dataset demonstrate that this approach yields semantically meaningful counterfactuals for the explainee.
The paper presents a lightweight framework for reliable egocentric action anticipation that tackles two failure modes caused by temporal corruption: unreliable temporal evidence during encoding and implausible verb‑noun compositions during decoding. It introduces Temporal Reliability Suppression (TRS), which predicts per‑frame suppression scores to penalize encoder attention and weight temporal pooling, and Robust Verb‑Noun Graph (RVG), which re‑ranks verb‑noun pairs using a PMI‑based compatibility graph derived from training labels. Experiments with corruption‑augmented training show that TRS+CA+RVG achieves 29.1% average corrupted accuracy and 88.2% relative robustness across six corruptions, while dramatically reducing rare verb‑noun predictions from 15.4% to 1.1%. whyItMatters":"The approach improves the robustness of wearable action anticipation systems to real‑world challenges such as missing frames, masking, and sensor noise, which are common in egocentric video settings."
arXiv:2609.13595v1 Announce Type: cross Abstract: Predicting potential dangers is core to safety. Forecasting the presence of other traffic agents is core to danger prediction. Occluded traffic agent...
arXiv:2609.15150v1 Announce Type: new Abstract: Unpaired cross-modal distillation transfers grade structure from histopathology into a micro-ultrasound (micro-US) encoder by aligning a pooled needle-...
arXiv:2609.15285v1 Announce Type: new Abstract: Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative disease in which early assessment remains challenging, particularly in low-resou...
arXiv:2609.14279v1 Announce Type: cross Abstract: Temporal knowledge graph embedding (TKGE) models infer missing facts in knowledge graphs that evolve over time. Many existing models use a single geo...
arXiv:2609.13868v1 Announce Type: cross Abstract: Earth-observation (EO) foundation models have become exceptionally effective at learning se mantic, high-dimensional geospatial embeddings, while mod...
Dynamic Learning Solutions is an automated pipeline that transforms NCERT textbook PDFs into interactive video explanations. Users upload a PDF and ask a question; the system retrieves relevant content, generates a multi-scene script, creates images with Stable Diffusion, animates them with DynamiCrafter, and adds synchronized narration via Google Text‑to‑Speech. The result is a coherent, textbook‑aligned video that turns static material into an engaging learning experience.
Traditional multimodal representation learning and generation are two stages: a contrastive or self-supervised visual encoder is trained first, followed by a separate downstream generative model. This...
Self-supervised learning (SSL) can reduce the need for labelled medical images, but the choice of pretext objective remains unclear for lung ultrasound (LUS). Contrastive learning, masked reconstructi...
Scientific research increasingly relies on large, heterogeneous data sources, motivating interest in retrieval-augmented generation (RAG) systems that provide natural language access to scientific kno...
Research data repositories are essential infrastructure for scientific inquiry and for ensuring that datasets follow FAIR (Findable, Accessible, Interoperable, and Reusable) principles. However, repos...
Audio deepfake detectors need to transfer to languages absent from training, as multilingual speech synthesis outpaces labeled anti-spoofing resources. While detectors increasingly rely on self-superv...
Despite the rapid progress of LLM-based agents for planning, code generation, and debugging, their practical value for tabular machine learning remains underexplored. In this paper, we investigate a c...
Multilingual Large Language Models (LLMs) traditionally rely on a single vocabulary shared by all supported languages, which can lead to uneven compression across them. Moreover, their large embedding...
Physics-Informed Conformal Prediction (PI‑CP) embeds PDE residuals into the nonconformity score of split conformal prediction, yielding distribution‑free prediction intervals with provable coverage that adapt spatially to physics violations. The method demonstrates consistent 89‑91% coverage across six physics scenarios, outperforming MC Dropout and Deep Ensembles, while Fourier Neural Operators (FNO) achieve superior accuracy over CNN and DeepONet. Additionally, the authors prove that FNO’s translation equivariance limits its ability to solve PDEs with Dirichlet boundary conditions, and show that adding coordinate channels can reduce error by up to 63×.
The paper introduces ASCIL, a post‑ASR correction framework that re‑evaluates wake‑up intent by combining acoustic embeddings, linguistic cues, device context, and past misclassifications. ASCIL interprets both implicit (hesitation, disengagement, silence) and explicit (cancellation, repetition) signals as noisy indicators of misclassification, enabling online pattern updates without manual annotation. On a proprietary dataset of 3,667 interactions, ASCIL reduces errors by up to 54.27% relative on a session‑disjoint subset and 24.39% at a 0.90 threshold, while adding less than 60 ms of latency and improving intentional acceptance rates.
The study benchmarks tokenization choices for generative medical event models, evaluating quantization granularity, reference-range anchoring, code–value fusion, numeric and temporal encodings, and native versus harmonized event representations. Using Llama and Qwen architectures, 156 models were trained and assessed on early hospitalization data, showing that fusing codes with value deciles and using event-order or admission-relative RoPE embeddings improved predictive performance. The Common Longitudinal Intensive Care Unit Data Format (CLIF) reduced token count by 30.8% while enhancing outcomes in most families.