雅虎

When audio and text conflict, speech-enabled language models follow the text 10 times more often than when arbitrating between two text sources, even when explicitly instructed to trust the audio. Using ALME, a benchmark of 57,602 controlled audio-text conflict stimuli across 8 languages, we find that Gemini 2.0 Flash exhibits 16.6% text dominance under audio-text conflict versus 1.6% under text-text conflict with identical reliability cues. This gap is not explained by audio quality: audio-only accuracy (97.2%) exceeds cascade accuracy (93.9%), indicating audio embeddings preserve more information than text transcripts. We propose that text dominance reflects an asymmetry not in information content but in arbitration accessibility: how easily the model can reason over competing representations. This framework explains otherwise puzzling findings. Forcing transcription before answering increases text dominance (19% to 33%), sacrificing audio's information advantage without improving accessibility. Framing text as “deliberately corrupted” reduces text dominance by 80%. A fine-tuning ablation provides interventional evidence: training only the audio projection layer increases text dominance (+26.5%), while LoRA on the language model halves it (-23.9%), localizing text dominance to the LLM's reasoning rather than the audio encoder. Experiments across four state-of-the-art audio-LLMs and 8 languages show consistent trends with substantial cross-linguistic and cross-model variation, establishing modality arbitration as a distinct reliability dimension not captured by standard speech benchmarks.

Bayesian quadrature is a probabilistic, model-based approach to numerical integration, the estimation of intractable integrals, or expectations. Although Bayesian quadrature was popularised already in the 1980s, no systematic and comprehensive treatment has been published. The purpose of this survey is to fill this gap. We review the mathematical foundations of Bayesian quadrature from different points of view; present a systematic taxonomy for classifying different Bayesian quadrature methods along the three axes of modelling, inference, and sampling; collect general theoretical guarantees; and provide a controlled numerical study that explores and illustrates the effect of different choices along the axes of the taxonomy. We also provide a realistic assessment of practical challenges and limitations to application of Bayesian quadrature methods and include an up-to-date and nearly exhaustive bibliography that covers not only machine learning and statistics literature but all areas of mathematics and engineering in which Bayesian quadrature or equivalent methods have seen use.

The proposed scalable LLM-as-judge paradigm provides the possibility of developing an automatic and reliable evaluation scheme for generative AI systems using the concept of domain-adaptive calibration. More specifically, the proposed scheme includes three distinct modules, such as Counterfactual Evaluation Transformer (CET) model, Domain Knowledge Graph Calibration Network (DKGCN), and Entropy-Guided Meta-Judge Optimizer (EGMJO). More specifically, CET evaluates robustness and reasoning stability of responses based on comparison between original answers and their counterfactual counterparts; DKGCN uses graph neural networks to embed relationships among various items within finance, healthcare, cybersecurity, and engineering domains, and EGMJO relies on entropy-based meta-learning to measure uncertainties, resolve contradictions among judges, and update threshold values. Consequently, these modules generate an accurate evaluation score enhanced with information on reliability and confidence, reaching 98.2% accuracy.

Large language models (LLMs) can generate plausible-sounding ETF portfolios while silently violating basic KYC-style constraints on risk, fees, and diversification. This is especially problematic in agentic multi-turn advisory systems, where each draft recommendation can become an action unless guarded by an auditable enforcement layer. We study a model-agnostic, asset-agnostic post-generation guardrail pipeline: (i) enforce a strict JSON allocation schema, (ii) validate allocations against numeric caps, and (iii) when violations occur, deterministically project the output to the nearest feasible portfolio via a convex quadratic program (QCQP). We introduce BiasMix-Finance (Mini), a compact stress-test benchmark for constrained decision-making under biased LLM generations, with a 16-ETF universe, three investor profiles, and eight bias prompts. Across three models and three inference modes (direct, critique, self-consistency), first-pass generations violate at least one cap in 47.6-85.7

Fluency and reliability are not the same property, and large language models have far more of the first than the second. A model's factual content is fixed at the point training concludes; its errors are distributed through billions of parameters rather than isolated in any inspectable location; and no signal in its output distinguishes a memorized fact from a well-phrased guess. These three properties — staleness, hallucination, and opacity — are not independent defects but different symptoms of the same design choice: knowledge stored implicitly in weights instead of explicitly in a retrievable structure. Retrieval-augmented generation (RAG) responds by splitting question answering into two stages, one that retrieves candidate evidence from an external, updatable corpus, and one that conditions generation on that evidence rather than on parametric memory alone. This review treats RAG as a systems-engineering discipline. Beginning from the specific failure modes that motivate retrieval augmentation, the discussion works through the component design of a production pipeline — ingestion, chunking, embedding, indexing, hybrid retrieval, and reranking — before turning to the evaluation methodology needed to determine whether a deployed system is actually behaving reliably. The literature surveyed spans dense passage retrieval, approximate nearest-neighbor search, vector database engineering, and automated RAG evaluation, and it converges on a claim central to this review: retrieval quality, not generator choice, is what governs whether a RAG system can be trusted. Consequences for high-stakes domains — finance, healthcare, law — are discussed, along with adaptive retrieval, agentic RAG, and evaluation standardization as directions likely to shape where the field goes next.