Retrieval pipelines, vector search, chunking and reranking: how models are grounded in a corpus instead of their weights.
arXiv:2609.15997v1 Announce Type: cross Abstract: Improving safety at intersections requires identifying crash mechanisms and recommending appropriate countermeasures. However, this process tradition...
arXiv:2609.16309v1 Announce Type: new Abstract: Despite the rapid progress of LLM-based agents for planning, code generation, and debugging, their practical value for tabular machine learning remains...
The paper presents a multimodal machine learning framework that classifies Emirati residential architectural styles by combining visual features from images and textual descriptions using OpenAI's CLIP model. The unified 512‑dimensional embeddings are reduced with UMAP, clustered with K‑Means, and then used to train an SVM classifier, achieving a 98% accuracy across eight style clusters. This approach outperforms previous studies and demonstrates the value of integrating visual and textual data for cultural heritage analysis.
RAG-CT is a defense mechanism designed to protect Retrieval-Augmented Generation (RAG) systems from leaking personally identifiable information (PII). It works by detecting malicious queries through analysis of entropy and margin distributions, applying a score-based detection method. Experiments across four attack strategies and four baseline defenses on two datasets show that RAG-CT significantly reduces PII leakage while outperforming existing defenses, all without altering the underlying LLM or retriever.
arXiv:2602.06652v2 Announce Type: replace Abstract: The robustness of Vision Language Models (VLMs) is commonly assessed through output-level invariance, implicitly assuming that stable predictions r...
arXiv:2609.17061v1 Announce Type: cross Abstract: Message-passing Graph Neural Networks (GNNs) iteratively propagate and aggregate local neighborhood information followed by global readout to learn g...
arXiv:2609.16163v1 Announce Type: new Abstract: The sharp increase in mass shootings underscores an urgent need for systems that guide victims to safety in real time. An effective evacuation system m...
arXiv:2503.05794v4 Announce Type: replace-cross Abstract: Speaker verification models are trained on large-scale public datasets whose licenses usually prohibit unauthorized commercial use, yet such...
FLAT (Flexible‑Length Aligned Transmodal representations) is a joint multimodal pre‑training framework that learns a shared encoder for images and text, producing 1‑D continuous embeddings that can be directly used by downstream generative decoders. By combining contrastive alignment with bidirectional cross‑modal generative objectives, FLAT yields representations that are both discriminative and generative, enabling cross‑modal retrieval and generation with a single pre‑training stage. The model achieves strong performance on T2I generation (GenEval 71.1), image captioning (BLEU‑4 40.5, CIDEr 138.6), and retrieval tasks (Recall@5 86.8/75.8 on MS‑COCO, 98.3/93.6 on Flickr30K), and supports linear interpolation, latent space arithmetic, and zero‑shot composed retrieval.
The paper proposes a method called language orthogonalization to improve zero‑shot cross‑lingual audio deepfake detection. By removing language‑dependent variation from self‑supervised speech models using a target‑free ridge map on language‑identification embeddings, the approach consistently lowers equal error rates across six languages and six model backbones. The gains are larger when the target language is more distant in the language‑identification space.
The paper re‑examines the impact of removing the encoder from Action Chunking Transformers (ACT), a model used for robot manipulation learning. Contrary to the original claim that encoder removal drops success rates from 35% to 2%, the authors find no such dramatic effect in their re‑runs, though minor variations remain uncertain. They attribute discrepancies to training length and checkpoint selection, and note that the encoder’s latent variable offers little reconstruction benefit on the tested benchmark, while its removal speeds up training.
The paper investigates intersectional biases in multimodal clinical predictions using Electronic Healthcare Records (EHR). It introduces datasets MIMIC-Eye1 and MIMIC-IV ED, applies unified text representations from pre‑trained clinical language models, and benchmarks bias mitigation at the intersectional subgroup level. Results show that subgroup‑specific mitigation is robust across datasets, subgroups, and embeddings, effectively addressing intersectional biases in multimodal settings.
arXiv:2609.16179v1 Announce Type: new Abstract: Z-loss has been widely applied to the logits of language-model output heads and sparse mixture-of-experts routers. Z-loss constrains the softmax log-no...
arXiv:2609.17509v1 Announce Type: cross Abstract: Neural audio codecs are a key component in speech language modeling. However, their high frame rates lead to long sequence lengths, increasing comput...
Lit3R is a system developed by tus-nlp for the LitTraceQA shared task, which focuses on evidence-grounded question answering over scientific literature. The system combines off-the-shelf retrieval, reranking, and large language model components without task-specific training, using an iterative retrieval process that merges BM25-based sparse and dense retrieval, cross-encoder reranking, and LLM verification, along with paper-to-paper expansion. In the official test set, Lit3R achieved a 4th place ranking on the leaderboard.
arXiv:2609.16032v1 Announce Type: cross Abstract: Diffusion-based artifact detectors score whole-slide image patches by reconstruction error under a model fine-tuned on clean tissue. We show that con...
InfoTaxa presents an information‑calibrated, label‑free clustering approach for fine‑grained visual taxonomy, using frozen pretrained visual embeddings and DNA as an audit signal. On the BIOSCAN‑5M dataset, the method achieves 0.79 AMI at family and 0.67 at genus, outperforming prior image baselines and matching oracle‑K and graph‑based methods. The study shows that while clustering efficiency recovers most image‑available information at higher taxonomic ranks, species‑level performance remains limited by both clustering and representation, with DNA adding significant predictive value.
The paper introduces the Multi-modal Knowledge Preserving Adapter (MKP-Adapter), an adapter-only approach that enables backward compatible training for multi-modal large language models without updating the backbone. It employs a multi-level preservation loss to maintain embedding geometry and a focal re-weighting strategy to focus on difficult samples. Experiments show strong backward compatibility across image, text, visual document, and video retrieval tasks with minimal latency overhead.
The paper introduces NepKANUN, an AI-powered legal assistant designed specifically for Nepali legal texts. Built on a fine‑tuned large language model and integrated into a Retrieval‑Augmented Generation (RAG) framework, it delivers precise answers to natural language legal queries. Evaluated with BERTScore, the system achieved strong F1 scores of 0.82, 0.77, and 0.71 across simple, moderate, and complex question categories, and expert reviews confirm its usability.
EviScope is a new paired counterfactual benchmark that evaluates grounded language models by fixing the question while manipulating evidence—adding, removing, distracting, or contradicting it. The v1.1 dataset includes 40 four‑condition quartets with repaired counterfactual claims and span‑level support labels for automated assessment. Experiments on Qwen2.5‑7B, Llama 3.1 8B, and Gemini 3.5 Flash show that paired metrics reveal grounding behaviors hidden by simple answer accuracy, such as unsupported answers, conflict blindness, and incorrect non‑answer actions.