Large language models (LLMs) are increasingly studied in the context of multi-turn reasoning, where models iteratively refine their outputs based on user-provided feedback. Such settings are crucial for tasks that require complex reasoning, yet existing feedback paradigms often rely on issuing new messages. LLMs struggle to integrate these reliably, leading to inconsistent improvements. In this work, we introduce in-place feedback, a novel interaction paradigm in which users directly edit an LLM's previous response, and the model conditions on this modified response to generate its revision. Empirical evaluations on diverse reasoning-intensive benchmarks reveal that in-place feedback achieves better performance than conventional multi-turn feedback while using $79.1\%$ fewer tokens. Complementary analyses on controlled environments further demonstrate that in-place feedback resolves a core limitation of multi-turn feedback: models often fail to apply feedback precisely to erroneous parts of the response, leaving errors uncorrected and sometimes introducing new mistakes into previously correct content. These findings suggest that in-place feedback offers a more natural and effective mechanism for guiding LLMs in reasoning-intensive tasks.

Diffusion models have achieved remarkable advancements in text-to-image generation. However, existing models still have many difficulties when faced with multiple-object compositional generation. In this paper, we propose RealCompo, a new training-free and transferred-friendly text-to-image generation framework, which aims to leverage the respective advantages of text-to-image models and spatial-aware image diffusion models (e.g., layout, keypoints and segmentation maps) to enhance both realism and compositionality of the generated images. An intuitive and novel balancer is proposed to dynamically balance the strengths of the two models in denoising process, allowing plug-and-play use of any model without extra training. Extensive experiments show that our RealCompo consistently outperforms state-of-the-art text-to-image models and spatial-aware image diffusion models in multiple-object compositional generation while keeping satisfactory realism and compositionality of the generated images. Notably, our RealCompo can be seamlessly extended with a wide range of spatial-aware image diffusion models and stylized diffusion models.

This started out as a challenge. With a friend of mine, we wanted to see if it was possible to build something from scratch and push it to production. In this post, we'll go through the necessary steps to build and deploy a machine learning application. This starts from data collection to deployment and the journey, as you'll see it, is exciting and fun . Before we begin, let's have a look at the app we'll be building: As you see, this web app allows a user to evaluate random brands by writing reviews. While writing, the user will see the sentiment score of his input updating in real-time along with a proposed rating from 1 to 5. The user can then change the rating in case the suggested one does not reflect his views, and submit. You can think of this as a crowd sourcing app of brand reviews with a sentiment analysis model that suggests ratings that the user can tweak and adapt afterwards. To build this application we'll follow these steps: All the code is available in our github repository and organized in independant directories, so you can check it, run it and improve it. Disclaimer: The scripts below are meant for educational purposes only: scrape responsibly. In order to train a sentiment classifier, we need data. We can sure download open source datasets for sentiment analysis tasks such as Amazon Polarity or IMDB movie reviews but for the purpose of this tutorial, we'll build our own dataset.