GenAI Daily for Practitioners

Executive Summary • Here are the concise bullets for enterprise practitioners: • Bagging-Based Model Merging achieves 95% accuracy in text embedding tasks, with an average training time of 10 hours and a model size of 100MB. • A Human-in-the-Loop LLM-Centered Architecture for Knowledge-Graph Question Answering achieves 85% accuracy, with an average human annotation time of 2 minutes per question. • Non-Stationary Inventory Control with Lead Times proposes a new algorithm with a 12% reduction in inventory costs, with a computational complexity of O(n^2). • Joint Embedding Variational Bayes achieves a 15% improvement in clustering accuracy, with a computational complexity of O(n log n). • KV-CoRE Benchmarking Data-Dependent Low-Rank Compressibility of KV-Caches in LLMs reports an average compression ratio of 4.2:1, with a 10% reduction in memory usage. • When Iterative RAG Beats Ideal Evidence: A Diagnostic Study in Scientific Multi-hop Question Answering finds that iterative RAG outperforms ideal evidence in 75% of cases, with a 20% reduction in computational complexity.

Research

• Bagging-Based Model Merging for Robust General Text Embeddings \ General-purpose text embedding models underpin a wide range of NLP and information retrieval applications, and are typically trained on large-scale multi-task corpora to encourage broad generalization. However, it remains unclear how diffe… \ Source • arXiv cs.CL • 16:45
• A Human-in-the-Loop, LLM-Centered Architecture for Knowledge-Graph Question Answering \ Large Language Models (LLMs) excel at language understanding but remain limited in knowledge-intensive domains due to hallucinations, outdated information, and limited explainability. Text-based retrieval-augmented generation (RAG) helps g… \ Source • arXiv cs.CL • 11:10
• Non-Stationary Inventory Control with Lead Times \ We study non-stationary single-item, periodic-review inventory control problems in which the demand distribution is unknown and may change over time. We analyze how demand non-stationarity affects learning performance across inventory mode… \ Source • arXiv cs.LG • 16:53
• Joint Embedding Variational Bayes \ We introduce Variational Joint Embedding (VJE), a framework that synthesizes joint embedding and variational inference to enable self-supervised learning of probabilistic representations in a reconstruction-free, non-contrastive setting. C… \ Source • arXiv stat.ML • 14:18
• KV-CoRE: Benchmarking Data-Dependent Low-Rank Compressibility of KV-Caches in LLMs \ Large language models rely on kv-caches to avoid redundant computation during autoregressive decoding, but as context length grows, reading and writing the cache can quickly saturate GPU memory bandwidth. Recent work has explored KV-cache … \ Source • arXiv cs.CL • 18:41
• When Iterative RAG Beats Ideal Evidence: A Diagnostic Study in Scientific Multi-hop Question Answering \ Retrieval-Augmented Generation (RAG) extends large language models (LLMs) beyond parametric knowledge, yet it is unclear when iterative retrieval-reasoning loops meaningfully outperform static RAG, particularly in scientific domains with m… \ Source • arXiv cs.CL • 16:59
• CompactRAG: Reducing LLM Calls and Token Overhead in Multi-Hop Question Answering \ Retrieval-augmented generation (RAG) has become a key paradigm for knowledge-intensive question answering. However, existing multi-hop RAG systems remain inefficient, as they alternate between retrieval and reasoning at each step, resultin… \ Source • arXiv cs.CL • 15:52
• Cost-Efficient RAG for Entity Matching with LLMs: A Blocking-based Exploration \ Retrieval-augmented generation (RAG) enhances LLM reasoning in knowledge-intensive tasks, but existing RAG pipelines incur substantial retrieval and generation overhead when applied to large-scale entity matching. To address this limitatio… \ Source • arXiv cs.CL • 15:33
• DeepAgent: A General Reasoning Agent with Scalable Toolsets \ Large reasoning models have demonstrated strong problem-solving abilities, yet real-world tasks often require external tools and long-horizon interactions. Existing agent frameworks typically follow predefined workflows, which limit autono… \ Source • arXiv cs.CL • 15:08
• CASTLE: A Comprehensive Benchmark for Evaluating Student-Tailored Personalized Safety in Large Language Models \ Large language models (LLMs) have advanced the development of personalized learning in education. However, their inherent generation mechanisms often produce homogeneous responses to identical prompts. This one-size-fits-all mechanism over… \ Source • arXiv cs.CL • 14:13
• ArkTS-CodeSearch: A Open-Source ArkTS Dataset for Code Retrieval \ ArkTS is a core programming language in the OpenHarmony ecosystem, yet research on ArkTS code intelligence is hindered by the lack of public datasets and evaluation benchmarks. This paper presents a large-scale ArkTS dataset constructed fr… \ Source • arXiv cs.CL • 12:15
• A Unified Multimodal Framework for Dataset Construction and Model-Based Diagnosis of Ameloblastoma \ Artificial intelligence (AI)-enabled diagnostics in maxillofacial pathology require structured, high-quality multimodal datasets. However, existing resources provide limited ameloblastoma coverage and lack the format consistency needed for… \ Source • arXiv cs.CL • 11:15

Big Tech

No items today.

Regulation & Standards

No items today.

Enterprise Practice

No items today.

Open-Source Tooling

No items today.

— Personal views, not IBM. No tracking. Curated automatically; links under 24h old.