Retrieval pipelines, vector search, chunking and reranking: how models are grounded in a corpus instead of their weights.
arXiv:2609.40362v1 Announce Type: new Abstract: We present Multimodal Flow, a fully continuous generative model of language and vision. Most unified multimodal models either model both language and q...
arXiv:2605.17630v3 Announce Type: replace Abstract: Frozen segmentation foundation models often fail when the target class appears in a form that is weakly represented during pretraining. To address...
The paper introduces VHOP, a data generation framework and benchmark for visual agentic search, and VHOP-Router, an end‑to‑end training pipeline that turns a standard embedding model into an autoregressive multi‑step retriever operating directly in visual latent space. VHOP-Router eliminates the need for intermediate text queries, boosting retrieval accuracy from under 5% to 76.3% and improving task success rates by 52.7% while dramatically reducing token usage and API payloads. The approach generalizes to unseen difficulty levels and realistic test sets, offering an efficient solution that preserves native LLM capabilities.
Gromov-Wasserstein multidimensional scaling (GW-MDS) learns low-dimensional representations from relational data but remains transductive, providing no explicit mapping for unseen samples. We introduc...
Digital image watermarking is increasingly critical in media contexts, as emerging regulations and industry practices require marking AI-generated content and ensuring traceable sources to prevent man...
Distinguishing periodic from chaotic dynamics in a time series is a fundamental challenge in both physics and engineering. Yet, end-to-end learned architectures must discover both a representation and...
Recent diffusion-based pipelines have achieved promising progress in image-to-3D synthesis. However, generating high-fidelity details remains challenging, especially when the input image contains rich...
This work proposes a mutual feedback architecture, MEQ, that refines the two inputs, of possibly different modalities, into a pair of coupled embeddings such that each embedding reflects the informati...
Human intelligence relies heavily on learned intuition: recognising patterns and judging situations without explicitly unfolding every intermediate step. We introduce Bongard, an open-weight System On...
The paper introduces Zero-Training Fisher Geometry Alignment (ZFGA), a closed‑form test‑time adaptation technique that aligns the Fisher geometry of test‑time feature embeddings with a reference geometry from clean data, improving robustness to covariate shift without altering model parameters. ZFGA is evaluated on CIFAR‑10‑C and ImageNet‑C with ResNet‑50, DINO ViT‑S/16, and CLIP ViT‑B/32, consistently outperforming zero‑shot inference and remaining stable across all three model families. Compared to other training‑free and gradient‑based methods, ZFGA uniquely avoids substantial performance degradation and requires only forward passes and matrix operations at inference time, making it a lightweight, deterministic alternative to optimization‑based test‑time adaptation.
The paper introduces Aperture, a method for storing Fourier moments of a token’s weighted support at rotary embedding frequencies to preserve positional information during token compression. It demonstrates that these moments provide the minimal real dimension necessary for maintaining expected rotary interactions, enabling additive updates and separate attention normalization. Experiments show that Aperture’s stored support achieves 65.63% accuracy in a video question‑answering task, slightly below the 67.12% accuracy of a standard merging rule, highlighting the trade‑off between exact positional preservation and downstream performance.
The paper introduces HyBrain, a model that uses spatiotemporal hyperedges to capture coordinated EEG seizure activity more efficiently than pairwise edge approaches. By generating one token per channel per second and pooling them into shared hyperedge embeddings, HyBrain performs well across detection and prediction tasks on TUSZ and CHB‑MIT datasets, achieving top AUROC scores and competitive training resources. A qualitative example shows a single hyperedge tracking the preictal‑ictal‑postictal progression of a real seizure.
CORE-BREW is a new multi‑bit watermarking method for large language models that uses log‑likelihood ratios for soft‑decision decoding, targeting a fixed hit rate to calibrate the watermark channel. It introduces entropy‑aware erasures to reduce perturbations in low‑entropy contexts and combines likelihood‑based scoring with soft‑decision list decoding to better exploit token‑level reliability. Experiments on open‑source LLMs show that CORE‑BREW improves detection robustness and payload recovery compared to the BREW baseline while keeping false‑positive rates low and maintaining translation quality metrics close to unwatermarked text.
arXiv:2609.36527v1 Announce Type: new Abstract: Recovering complete physical fields from sparse observations is challenging because the measurements may not uniquely determine the underlying state. D...
arXiv:2609.37888v1 Announce Type: cross Abstract: Class-Incremental Learning (CIL) requires models to recognize new classes over time without forgetting previously learned ones. With the rise of visi...
arXiv:2609.35868v1 Announce Type: new Abstract: Is human readability necessary for effective fine-tuning of large language models? We investigate whether model-conditioned training representations ca...
This paper introduces a Byzantine‑robust federated retrieval‑augmented generation (RAG) framework that uses aligned calibration and fixed‑membership conformal prediction to ensure that the answer set contains the correct answer with a chosen probability, even when some nodes are compromised. By having all nodes score the same calibration questions and retaining only candidates that could be kept by a plausible group of honest nodes, the method guarantees correctness in finite samples and produces smaller answer sets than simpler approaches. Experiments on medical exam question‑answering tasks with language‑model nodes demonstrate that the method meets the target coverage whenever the number of misbehaving nodes does not exceed the declared bound, while plain averaging often fails.
MUGEN is a large‑scale real‑world dataset of over 1,300 hours of 4K panoramic videos with rich semantic and geometric annotations, designed to support interactive 360° world exploration. Wan360 is a camera‑controllable panoramic video generation model built on MUGEN, featuring ERP‑aware components (periodic longitude RoPE, ERP‑aware padding, random roll yaw) and a panoramic Plücker embedding for camera motion. Together, they address gaps in data and model support for immersive, temporally coherent 360° video generation along user‑specified camera trajectories.
The paper introduces Federation of Experts (FoE), a new architecture that reorganizes the mixture-of-experts (MoE) block in transformer layers into multiple MoE clusters. Each cluster handles a single KV head, and expert parallelism is applied within clusters while a sum operation synchronizes post‑attention residuals across clusters. FoE eliminates all‑to‑all communication on a single GPU and limits it to intra‑node communication in multi‑node setups, leading to significant reductions in inference latency and throughput improvements on LongBench.
The paper introduces SACReg, a spectral anti-collapse regularizer that enforces λ-balance across weight matrices to prevent dimensional collapse in the backbone of joint-embedding self-supervised learning models. Applied to JEPA, the resulting λ-JEPA improves ImageNet-1k classification and linear-probe transfer on eight image datasets, and also outperforms prior video self-supervised methods on Something-Something-v2 and Kinetics-400.