CADMorph: Geometry-Driven Parametric CAD Editing via a Plan-Generate-Verify Loop

A new framework called CIP has been introduced to mitigate hallucinations in large language models (LLMs) when processing long and noisy contexts. By constructing a causal relation sequence among entities and actions, CIP enhances reasoning quality and factual grounding across various models, including GPT-4o and Gemini 2.0 Flash.

Recent advancements in retrieval-augmented generation (RAG) have led to the introduction of MedBioRAG, a model designed to enhance biomedical question-answering (QA) by integrating semantic and lexical search with document retrieval and supervised fine-tuning. This model has demonstrated superior performance compared to previous state-of-the-art models across various benchmark datasets.

A new benchmark named SmokeBench has been introduced to assess the capabilities of multimodal large language models (MLLMs) in detecting and localizing wildfire smoke in images. The benchmark includes four tasks: smoke classification, tile-based and grid-based smoke localization, and smoke detection, evaluating models such as Idefics2, Qwen2.5-VL, and GPT-4o. Results indicate that while some models can identify smoke over large areas, they struggle with precise localization, particularly in early detection stages.

The introduction of UFVideo marks a significant advancement in video understanding by utilizing multi-modal Large Language Models (LLMs) to achieve unified fine-grained cooperative understanding across various video contexts. This model integrates visual-language guided alignment to enhance video comprehension at global, pixel, and temporal scales, addressing limitations in existing specialized video understanding tasks.