The example shows the learned IT map with high w =2 0 approximating the ET map.

The example shows the learned IT map with high w =2 0 approximating the ET map.

Despite recent progress in video large language models (VideoLLMs), a key open challenge remains: how to equip models with chain-of-thought (CoT) reasoning abilities grounded in fine-grained object-level video understanding. Existing instruction-tuned models, such as the Qwen and LLaVA series, are trained on high-level video-text pairs, often lacking structured annotations necessary for compositional, step-by-step reasoning. We propose CoTasks: Chain-of-Thought based Video Instruction Tuning Tasks, a new framework that decomposes complex video questions of existing datasets (e.g., NeXT-QA, STAR) into four entity-level foundational tasks: frame localization, entity tracking, spatial and temporal relation extraction. By embedding these intermediate CoT-style reasoning steps into the input, CoTasks enables models to explicitly perform object-centric spatiotemporal reasoning. Experiments on the NeXT-QA benchmark show that CoTasks significantly enhance inference performance: LLaVA-video-7B improves by +3.3 points in average GPT-4 evaluation score, and Qwen2.5-VL-3B gains +17.4, with large boosts in causal (+14.6), temporal (+10.9), and descriptive (+48.1) subcategories. These results demonstrate the effectiveness of CoTasks as a structured CoT-style supervision framework for improving compositional video reasoning.

Since the video was posted earlier this month, millions of TikTok users have watched as a young Chinese man in a blue T-shirt sits beside a traditional tea set and speaks directly to the camera in accented English: "Let's expose luxury's biggest secret." He stands and lifts what looks like an Hermès Birkin bag, one of the world's most exclusive and expensive handbags, before gesturing toward the shelves filled with more bags behind him. "You recognize them: Hermès, Louis Vuitton, Prada, Gucci--all crafted in our workshops." He ends by urging viewers to buy directly from his factory. Video "exposés" like this--where a sales agent breaks down the material cost of luxury goods, from handbags to perfumes to appliances--are everywhere on TikTok right now.

Multimodal Large Language Models (MLLMs) are commonly evaluated using costly annotated multimodal benchmarks. However, these benchmarks often struggle to keep pace with the rapidly advancing requirements of MLLM evaluation. We propose GenCeption, a novel and annotation-free MLLM evaluation framework that merely requires unimodal data to assess inter-modality semantic coherence and inversely reflects the models' inclination to hallucinate. Analogous to the popular DrawCeption game, GenCeption initiates with a non-textual sample and undergoes a series of iterative description and generation steps. Semantic drift across iterations is quantified using the GC@T metric. Our empirical findings validate GenCeption's efficacy, showing strong correlations with popular MLLM benchmarking results. GenCeption may be extended to mitigate training data contamination by utilizing ubiquitous, previously unseen unimodal data.

"I'm going to show you a fake bag so you can see how spectacular an imitation can be." Lucía De La Calle, is an appraiser at the second-hand luxury handbags boutique Del Páramo Vintage in Valladolid, in northwest Spain. She shows EL PAÍS a Louis Vuitton Twist MM bag, a classic model that the brand wants to make a basic, betting on its reissue year after year in different colors. "I think my eye is quite trained but, often, even I am unable to perceive the difference," says De La Calle. The woman who bought this bag for €2,500 on the second-hand market brought it to the store with the brand's characteristic orange box and authentic dust cover, as well as a copy of the invoice from its previous owner.

Counterfeiting is a big business. Nearly $509 billion of fake and pirated products were sold internationally in 2016. In that year, the latest for which data was available, counterfeit goods made up 3.3% of international trade, up from 2.5% three years earlier, according to the Organization for Economic Cooperation and Development. That figure, which does not include domestic trade in fakes, not only means companies are losing revenue and consumers are not getting their money's worth; counterfeiting also helps fund organized crime. Because it skirts safety regulations, makers of counterfeits could use toxic materials or produce unsafe products.

As a data science intern, I have come to the realization that the amount of value you provide is directly proportional to the price tag you get. The number of skills you possess only matters if those skills can translate to added value to your customer in the form of increased sales or decreases costs. The more value you provide, the more valuable you are as a Data Scientist. In today's article, I will be going over a significant topic that I have come across repeatedly during my time as an internee and that topic is Price Elasticity. Price Elasticity tells us how sensitive sales of a particular product are to a unit change in its price.

It normally takes a user three to five minutes to go through the authentication process, but she is faster because the store, Opulent Habits, in Madison, N.J., has been using the app since 2018. "I can do it in less than a minute at this point," Ms. Matthaei says. A look at how innovation and technology are transforming the way we live, work and play. Increasingly, the role of spotting counterfeits is being filled by artificial-intelligence algorithms that have studied every angle of tens of thousands of bags, shoes and other items that are often knocked off. Inc. are developing machine-learning tools to help protect shoppers.

Image classification is not a hard topic anymore. Tensorflow has all the inbuilt functionalities that take care of the complex mathematics for us. Without knowing the details of the neural network, we can use a neural network now. In today's project, I used a Convolutional Neural Network (CNN) which is an advanced version of the neural network. If you worked with the FashionMNIST dataset that contains shirts, shoe handbags, etc., CNN will figure out important portions of the images.