Skip to main content
Search

Mohamed bin Zayed University of Artificial Intelligence

19 related articles

Subscribe to updates
  1. arXiv

    Finance agents meet an open world: none of 150 sessions passes, and all 35 the simulated user approved fail a program check

    FinCosmos opens all three uncertainties of financial agent evaluation at once: a simulated investor reveals a hidden strategy only when asked, the agent must source its own Web data and backtest under a data cutoff, and programs verify the delivery instead of matching a reference answer. Across 150 sessions with four coding agents and a tool-free model, none is accepted, and all 35 sessions the simulated user signed off fail at least one program veto.

  2. arXiv

    Few-step text-to-image gains a feature-moment constraint: Z-Image-Turbo hits 94.25% on Chinese LongText-Bench at two evaluations

    Few-step text-to-image generation must render prompt details with very few sampling steps, and distribution matching distillation (DMD) relies on learned score differences whose statistics need not align. Intrinsic Distribution Matching keeps the full DMD correction and adds a moment constraint on the means and per-coordinate standard deviations of the generator's exponential-moving-average features, with no external feature encoder. On Z-Image-Turbo 6B at two network evaluations it reaches 94.25% and 90.20% on Chinese and English LongText-Bench, the highest scores among the compared methods at that budget.

  3. arXiv

    Tabular foundation model rankings shuffle across benchmarks: on 160 classification and 81 regression datasets, accurate point predictions do not guarantee trustworthy intervals

    Across 160 classification and 81 regression datasets, tabular foundation model rankings change with the benchmark and the reliability metric, and regression models that already sit near nominal marginal coverage still under-cover in their worst feature slab, so model selection cannot rest on aggregate predictive accuracy alone.

  4. arXiv

    Visual reasoning drops coordinate regression: a probe token samples evidence over visual tokens, hitting 78.0% on 4K with about 1.2% of the tokens

    TiCO replaces coordinate regression in thinking-with-images with a probe token that samples evidence over visual tokens, reaching 78.0% overall accuracy on HR-Bench 4K while generating only 6.2 tokens per query.

  5. Fact-checking articles can now be drafted automatically: QRAFT chains evidence, drafting and simulated editing, but experts still want a human in the loop

    QRAFT splits fact-checking article writing into evidence gathering, drafting and a simulated editorial review run by multiple LLM agents, beating prior text-generation methods on automatic metrics across 1,227 examples while still scoring well below expert-written articles with professional fact-checkers.

  6. arXiv

    AutoSynth generates editable Helm synthesizer programs directly with an audio-conditioned autoregressive model, achieving competitive results in both synthesizer inversion and text-driven generation

    AutoSynth serializes MIDI performance events, fixed synthesizer parameters, and variable-length modulation routes into a unified sequence and learns their dependencies with an audio-conditioned autoregressive model, trained first by supervised learning on large-scale automatically constructed audio-program pairs and then by group-relative policy optimization with a mixed reward, so that it predicts a synthesizer program directly from reference audio and converts text into a program via a pretrained audio generation model, achieving competitive results in both synthesizer inversion and text-driven generation.

  7. arXiv

    Adapting language-dominance probing from cross-lingual to 26 Arabic varieties finds no MSA internal dominance, with dialect representations densely overlapping

    Adapting the language-dominance probing framework of Shani and Basirat (2025) from cross-lingual to 25 Arabic dialects plus MSA, the study analyzes hidden representations layer by layer in BLOOM, mGPT and three further models and finds markedly lower normalized mutual information than in the cross-lingual case, a densely overlapping dialect space, and layer-wise dominance scores near the uniform baseline, giving no evidence that MSA or any dialect acts as a dominant internal reference point.

  8. arXiv

    Study reveals two fusion pathways in multimodal LLMs: concatenation models go text-first then vision, native models show early visual-textual co-adaptation

    The study runs three connected analyses (alignment decoupling, attention routing and entropy, intrinsic dimensionality) plus causal intervention on representative concatenation and native multimodal LLMs, finding that concatenation models follow a text-first, vision-later fusion pathway while native models show earlier visual-textual co-adaptation and feature-space reorganization, offering a mechanistic view of multimodal fusion and architecture-aware diagnostics.

  9. arXiv

    RDTU pairs a residual diffusion model with a retained-set kernel predictor to produce pseudo-labels, yielding unlearned time-series forecasters that most closely match exact retraining

    The work presents RDTU, a Residual Diffusion framework for time-series unlearning: it first uses a retained-set neural tangent kernel predictor to obtain a deletion-compatible base forecast, then quantifies the global and local structural support of each affected window via the volume contribution of the retained-reference data, and finally has a diffusion model generate a residual correction that estimates the counterfactual forecast, yielding a pseudo-label field that guides a lightweight model update; experiments show RDTU consistently produces unlearned models that most closely match exact retraining.

  10. arXiv

    Omni-Embed-Mini binds speech, audio, image, video and visual documents into one cosine space via a frozen text backbone, holding text retrieval at 49.57 nDCG@10 with 0.9B parameters

    MBZUAI presents Omni-Embed-Mini, which recasts cross-modal alignment as self-distillation through a shared frozen backbone: each media sample is paired with a dense cascaded caption and the teacher target is simply that same frozen backbone's embedding of the caption, so only projectors and phased LoRA adapters are trained; the 0.9B variant maps text, speech, audio, images, video and visually-rich documents into one cosine space while keeping text weights bit-identical, scoring 49.57 nDCG@10 on MTEB-v2 BEIR-8, and the 2.3B variant edges ahead of the closed gemini-embedding-2 on the overall-modality average, 51.39 to 49.51.