Search Results for author:
Found 3 papers, 2 papers with code
Reason from Fallacy: Enhancing Large Language Models' Logical Reasoning through Logical Fallacy Understanding
Large Language Models (LLMs) have demonstrated good performance in many reasoning tasks, but they still struggle with some complicated reasoning tasks including logical reasoning.
StableLLaVA: Enhanced Visual Instruction Tuning with Synthesized Image-Dialogue Data
However, these datasets often exhibit domain bias, potentially constraining the generative capabilities of the models.
Ranked #56 on
Visual Question Answering
on MM-Vet
Disentangled Pre-training for Image Matting
The pre-training task is designed in a similar manner as image matting, where random trimap and alpha matte are generated to achieve an image disentanglement objective.
Cannot find the paper you are looking for? You can
Submit a new open access paper.
