Retrieval pipelines, vector search, chunking and reranking: how models are grounded in a corpus instead of their weights.
arXiv:2609.00513v1 Announce Type: new Abstract: Retrieval-Augmented Generation (RAG) mitigates large language models (LLMs) hallucinations, yet conventional dense retrieval struggles with the complex...
The paper investigates the use of Multi-Agent Debate (MAD) for creative generation tasks such as narrative writing and scientific ideation. It finds that MAD’s convergence-driven design suppresses output diversity across independent runs, creating a trade-off with creative tasks. To address this, the authors propose Creative-MAD, which introduces Cognitive Lens Assignment and Embedding-based Peer Selection to preserve agent divergence, and demonstrate that it improves lexical and semantic diversity while maintaining quality.
The paper introduces the Unified Memory Agent (UMA), a system that builds a query‑agnostic external memory from a data stream and reuses it across multiple question‑answering sessions. UMA employs a single policy to manage a structured Memory Bank via CRUD operations and uses Task‑Stratified GRPO to supervise memory maintenance based on QA trajectory rewards. The authors also present Ledger‑QA, a benchmark for long‑horizon state tracking, and demonstrate that UMA outperforms other methods on test‑time learning and accurate‑retrieval tasks, with UMA‑Specialist further improving performance after task adaptation.
The paper introduces SAM3-LoRA, a parameter‑efficient adaptation of the SAM3 segmentation foundation model using Low‑Rank Adaptation for multi‑class structural defect segmentation. It presents a supervision method that trains the model directly from COCO‑style instance segmentation by using category names as prompts, eliminating the need for prompt templates or learned embeddings. The authors also identify a failure mode where the model responds to any prompt due to positive‑prompt‑only annotations and resolve it with exhaustive hard‑negative prompting, achieving significant improvements in pixel IoU and instance recall on both a tunnel lining dataset and the Structural Defects Dataset.
The paper introduces a method for generating multilingual reasoning traces for medical question answering using large language models. It creates 500,000 reasoning traces in English, Italian, and Spanish by retrieving medical information from Wikipedia and applies them to MedQA and MedMCQA datasets extended into Italian and Spanish. The approach improves performance in both few‑shot in‑context learning and supervised fine‑tuning, achieving state‑of‑the‑art results for 8B‑parameter LLMs and releasing all resources for further research.
The paper "Evaluating Style-Personalized Text Generation: Challenges and Directions" examines the difficulties of assessing text that is tailored to individual users’ styles. It critiques common metrics such as BLEU, embeddings, and LLM-as-judges, and introduces a style discrimination benchmark covering domain discrimination, authorship attribution, and LLM-generated personalized versus non-personalized discrimination across eight writing tasks. The study finds that ensembles of diverse evaluation metrics outperform single-evaluator approaches and offers guidance for reliable assessment of style-personalized generation.
Inspicio is an open‑vocabulary pipeline that links tokens in historical or low‑resource languages to synsets in the Open English WordNet without needing a source‑language sense inventory. It uses an instruction‑tuned LLM to generate two English translations, candidate dictionary definitions, and English lemmas, then performs hybrid retrieval combining dense definition similarity, sparse lemma matching, and Maximal Marginal Relevance re‑ranking. Evaluated on Latin, Ancient Greek, PREMOVE, and Italian data, the best configuration achieves 96% Recall@50 on a perception‑verb test set and remains competitive in out‑of‑domain and cross‑lingual scenarios.
arXiv:2609.00231v1 Announce Type: new Abstract: Existing research on object hallucination in multimodal large language models (MLLMs) predominantly attributes the problem to language priors such as o...
arXiv:2509.05084v2 Announce Type: replace Abstract: The primary paradigm in Neural Combinatorial Optimization (NCO) consists of construction methods, where a neural network is trained to sequentially...
arXiv:2609.00521v1 Announce Type: cross Abstract: From horizon detection to fibre structures in X-ray imaging, many vision tasks recover lines via peak detection in Hough space $H=S^1\times\mathbb{R}...
arXiv:2609.00396v1 Announce Type: new Abstract: Histopathological whole slide images (WSIs) are central to cancer diagnosis, but their gigapixel scale, tissue heterogeneity, weak slide-level supervis...
REFACTOR‑VLA is a wake/sleep framework that learns reusable motor‑program abstractions for vision‑language‑action models. In its sleep phase, it clusters motor‑program fragments using a Behavioral‑Equivalence Kernel derived from rollouts of a latent world model, while its wake phase emits typed lambda terms that a library‑conditioned action decoder consumes. Experiments on LIBERO show that adding an auxiliary supervised contrastive loss during world‑model warm‑up significantly improves clustering quality, outperforming the strongest published baseline across all four task suites.
arXiv:2603.02974v2 Announce Type: replace Abstract: DINO models provide rich patch-level representations that have recently enabled strong performance in unsupervised anomaly detection (UAD). Most ex...
The paper introduces a retrieval‑augmented generation framework that combines fine‑tuned open‑weight language models with a hybrid dense‑lexical retrieval pipeline to synthesize corrosion knowledge. Applied to magnesium alloy corrosion, the system achieves significant improvements in token‑level accuracy and high faithfulness and context recall, while an additional Reason Map framework constructs evidence graphs to detect causal errors that standard factuality metrics miss. The modular design is positioned as a general blueprint for trustworthy AI‑assisted knowledge synthesis across engineering domains.
The paper introduces PLANET, a multi‑object tracker that transcends traditional image‑plane limitations by incorporating 3D scene geometry into its query formation. By lifting 2D tracking datasets into 3D and embedding reconstructed geometry into features and positional encodings, PLANET encourages queries to encode object positions. An auxiliary 3D location prediction task and a dual‑resolution temporal memory further enhance performance, enabling state‑of‑the‑art results on three diverse benchmarks.
EvoFlint is an evolutionary atlas that maps multi‑turn vulnerabilities in large language models by treating attack discovery as a search problem rather than a generation task. It uses evolutionary quality‑diversity search to evolve phased conversation plans, employing Pareto fitness for success rate and severity, novelty search for diversity, and a generation‑level memory to incorporate model insights. The resulting risk‑indexed archive, tested on HarmBench, shows high attack success rates across models such as Claude Sonnet, GPT‑5, and Qwen3, revealing which harm categories each model’s safety training covers or misses.
SpecMind is a Multi-Agent Retrieval-Augmented Generation system designed to support spectrum intelligence by reasoning over heterogeneous data sources such as policy proceedings, legal regulations, and license databases. The system coordinates specialized sub-agents that retrieve and synthesize knowledge, enabling autonomous agents to handle diverse query types. SpecBench, a Q&A dataset based on real-world license records and policy proceedings, was created to evaluate RAG systems in the spectrum domain, and experimental results show SpecMind outperforms traditional RAG systems with over an 80% win rate against strong baselines.
arXiv:2609.00691v1 Announce Type: new Abstract: Post-hoc out-of-distribution detectors are fitted on a finite reference set, so every score they produce is an estimate. If we had chosen a different s...
arXiv:2609.00411v1 Announce Type: new Abstract: Modern face recognition (FR) owes much of its success to deep neural networks that learn to extract compact identity embeddings from face images. These...
The paper introduces QXymb, a framework for building observational declarative twins of quantum circuits, and presents its first complete order‑0 specialization, QILP‑0. QILP‑0 transforms observed circuit behavior into a finite multi‑valued propositional logic program by incrementally traversing a declared family of quantum observables, quantifying progress via reference‑relative coverage, and preserving observational semantics through deterministic mapping back to original observable columns. Validation on Bars & Stripes and MNIST quantum machine learning settings shows that the induced QILP‑0 theory achieves perfect, conflict‑free reconstruction of the discrete relations, with logical exactness separated from numerical and discretization uncertainties. whyItMatters":"The work demonstrates a method to construct exact observational declarative twins of quantum circuits, enabling precise logical reconstruction of quantum behavior independent of numerical uncertainties."