Despite notable advancements in Multimodal Large Language Models (MLLMs), most state-of-the-art models have not undergone thorough alignment with human preferences. This gap exists because current alignment research has primarily achieved progress in specific areas (e.g., hallucination reduction), while the broader question of whether aligning models with human preferences can systematically enhance MLLM capability remains largely unexplored. To this end, we introduce MM-RLHF, a dataset containing $\mathbf{120k}$ fine-grained, human-annotated preference comparison pairs. This dataset represents a substantial advancement over existing resources, offering superior size, diversity, annotation granularity, and quality. Leveraging this dataset, we propose several key innovations to improve both the quality of reward models and the efficiency of alignment algorithms. Notably, we introduce a Critique-Based Reward Model, which generates critiques of model outputs before assigning scores, offering enhanced interpretability and more informative feedback compared to traditional scalar reward mechanisms. Additionally, we propose Dynamic Reward Scaling, a method that adjusts the loss weight of each sample according to the reward signal, thereby optimizing the use of high-quality comparison pairs. Our approach is rigorously evaluated across $\mathbf{10}$ distinct dimensions and $\mathbf{27}$ benchmarks, with results demonstrating significant and consistent improvements in model performance. Specifically, fine-tuning LLaVA-ov-7B with MM-RLHF and our alignment algorithm leads to a $\mathbf{19.5}$% increase in conversational abilities and a $\mathbf{60}$% improvement in safety. We have open-sourced the preference dataset, reward model, training and evaluation code, as well as reward modeling and safety benchmarks. For more details, please visit our project page: https://mm-rlhf.github.io.

Recent advancements in Generative AI have significantly improved the efficiency and adaptability of natural language processing (NLP) systems, particularly through Retrieval-Augmented Generation (RAG), Low-Rank Adaptation (LoRA), and Weight-Decomposed Low-Rank Adaptation (DoRA). RAG integrates external knowledge to enhance factual consistency in generative outputs, while LoRA enables parameter-efficient fine-tuning of large language models (LLMs). DoRA further refines this process by optimizing fine-tuning through adaptive parameter ranking and domain-aware weight adjustments, improving learning efficiency while maintaining inference performance. This paper presents a large-scale empirical evaluation of RAG, LoRA, and DoRA, with model fine-tuning and generation performance assessed on 20,000 FAQ-based queries, while the knowledge base spans 400,000 entries. The study analyzes key performance metrics such as accuracy, relevance, and inference latency. Experimental results demonstrate that DoRA achieves the highest accuracy (90.1%), relevance score (0.88), and lowest latency (110 ms per query), outperforming both LoRA and RAG in real-world, domain-specific generative AI applications. Furthermore, this study examines the trade-offs between fine-tuning efficiency, computational cost, and real-time adaptability across different models. Findings highlight RAG's effectiveness in knowledge grounding, LoRA's cost-efficient domain adaptation, and DoRA's ability to balance fine-tuning efficiency with model precision. These insights provide practical guidance for deploying AI-driven generative systems in accuracy-critical domains such as healthcare, finance, and legal services, ensuring scalability, reliability, and optimal performance in dynamic environments.

While watermarks for closed LLMs have matured and have been included in large-scale deployments, these methods are not applicable to open-source models, which allow users full control over the decoding process. This setting is understudied yet critical, given the rising performance of open-source models. In this work, we lay the foundation for systematic study of open-source LLM watermarking. For the first time, we explicitly formulate key requirements, including durability against common model modifications such as model merging, quantization, or finetuning, and propose a concrete evaluation setup. Given the prevalence of these modifications, durability is crucial for an open-source watermark to be effective. We survey and evaluate existing methods, showing that they are not durable. We also discuss potential ways to improve their durability and highlight remaining challenges. We hope our work enables future progress on this important problem.

Recent advancements in large language models (LLMs) have demonstrated extraordinary comprehension capabilities with remarkable breakthroughs on various vision-language tasks. However, the application of LLMs in generating reliable medical diagnostic reports remains in the early stages. Currently, medical LLMs typically feature a passive interaction model where doctors respond to patient queries with little or no involvement in analyzing medical images. In contrast, some ChatBots simply respond to predefined queries based on visual inputs, lacking interactive dialogue or consideration of medical history. As such, there is a gap between LLM-generated patient-ChatBot interactions and those occurring in actual patient-doctor consultations. To bridge this gap, we develop an LLM-based dialogue system, namely proactive multi-round vision-language interactions for computer-aided diagnosis (ProMRVL-CAD), to generate patient-friendly disease diagnostic reports. The proposed ProMRVL-CAD system allows proactive dialogue to provide patients with constant and reliable medical access via an integration of knowledge graph into a recommendation system. Specifically, we devise two generators: a Proactive Question Generator (Pro-Q Gen) to generate proactive questions that guide the diagnostic procedure and a Multi-Vision Patient-Text Diagnostic Report Generator (MVP-DR Gen) to produce high-quality diagnostic reports. Evaluating two real-world publicly available datasets, MIMIC-CXR and IU-Xray, our model has better quality in generating medical reports. We further demonstrate the performance of ProMRVL achieves robust under the scenarios with low image quality. Moreover, we have created a synthetic medical dialogue dataset that simulates proactive diagnostic interactions between patients and doctors, serving as a valuable resource for training LLM.

Conventional predictive modeling of parametric relationships in manufacturing processes is limited by the subjectivity of human expertise and intuition on the one hand and by the cost and time of experimental data generation on the other hand. This work addresses this issue by establishing a new Large Language Model (LLM) framework. The novelty lies in combining automatic extraction of process-relevant knowledge embedded in the literature with iterative model refinement based on a small amount of experimental data. This approach is evaluated on three distinct manufacturing processes that are based on machining, deformation, and additive principles. The results show that for the same small experimental data budget the models derived by our framework have unexpectedly high extrapolative performance, often surpassing the capabilities of conventional Machine Learning. Further, our approach eliminates manual generation of initial models or expertise-dependent interpretation of the literature. The results also reveal the importance of the nature of the knowledge extracted from the literature and the significance of both the knowledge extraction and model refinement components.

Elon Musk says he will abandon his 97.4bn offer to buy the non-profit behind OpenAI if the ChatGPT maker drops its plan to convert into a for-profit company. "If OpenAI, Inc's Board is prepared to preserve the charity's mission and stipulate to take the'for sale' sign off its assets by halting its conversion, Musk will withdraw the bid," lawyers for the billionaire said in a filing to a California court on Wednesday. "Otherwise, the charity must be compensated by what an arms-length buyer will pay for its assets." Musk and a group of investors made their offer earlier this week, in the latest twist to a dispute with the artificial intelligence company that he helped found a decade ago. OpenAI is controlled by a non-profit board bound to its original mission of safely building "better-than-human" AI for public benefit.

OpenAI's CEO Sam Altman wrote a social media post with an update on the roadmap for ChatGPT. In it, he explained that they've halted the launch of its upcoming o3 reasoning model to instead focus more on a streamlined yet monolithic version of GPT-5. "We want AI to'just work' for you; we realize how complicated our model and product offerings have gotten. We hate the model picker as much as you do and want to return to magic unified intelligence. We will next ship GPT-4.5, the model we called Orion internally, as our last non-chain-of-thought model.

A new version of ChatGPT is rumored to be released early this year, and while that's fine and dandy for people who know how to use AI, it could create a serious knowledge gap for beginner users who don't start learning now. If you don't want to get left behind in the Wild West, it's time to saddle up, cowboy. Luckily, learning ChatGPT and other AI tools is easier (and less painful) than riding a horse. We have an online training bundle with 12 courses that'll teach you the ropes in just a couple of weeks if you can dedicate an hour each day to studying. Get lifetime access for only 19.97 for a limited time (reg.

This story was originally published on the author's substack, Field Notes with Alexander C Kaufman, to which you can subscribe here. Artificial intelligence is driving up demand for electricity--the only question is how much, and what provides the power. Over the next three years, the Lawrence Berkeley National Laboratory estimates, AI's thirst for power will double or triple. Last month, OpenAI unveiled its Stargate Project, a plan to invest 500 billion in the infrastructure for artificial intelligence over the next four years that includes adding 25 gigawatts of new electricity capacity. Right now, the most likely source of electricity to power those data centers is gas.

Stargate, the massive artificial intelligence (AI) infrastructure project recently unveiled by President Donald Trump, has begun production in Texas -- with data center construction in other states expected to be announced in the coming months. OpenAI, Softbank, Oracle and other partners' total investment of 500 million in the project will produce a large-scale network of campuses. Each campus will be designed in the roughly 1 gigawatt (GW) or greater range, a measurement of electricity that can power a minimum of 750,000 homes. During a recent press briefing on The Stargate Project attended by Fox News Digital, OpenAI announced that construction on the first site is underway in Abilene, Texas. Significant progress has been made in identifying additional locations.