AI Breakdown

arxiv Preprint - Summarization is (Almost) Dead

In this episode we discuss Summarization is (Almost) Dead
by Xiao Pu, Mingqi Gao, Xiaojun Wan. The paper investigates the capabilities of large language models (LLMs) in summary generation. Through new datasets and human evaluation experiments, the authors find that LLM-generated summaries are preferred by evaluators compared to human-written summaries and fine-tuned model summaries. LLM-generated summaries exhibit improved factual consistency and fewer instances of extrinsic hallucinations, leading the authors to suggest that traditional text summarization methods may no longer be necessary. However, the authors emphasize the need for further exploration in areas such as dataset creation and evaluation methods.

arxiv Preprint - LLM-Grounder: Open-Vocabulary 3D Visual Grounding with Large Language Model as an Agent

In this episode we discuss LLM-Grounder: Open-Vocabulary 3D Visual Grounding with Large Language Model as an Agent
by Jianing Yang, Xuweiyi Chen, Shengyi Qian, Nikhil Madaan, Madhavan Iyengar, David F. Fouhey, Joyce Chai. The paper introduces a new approach called LLM-Grounder for grounding 3D visual scenes using natural language queries. It utilizes a Large Language Model (LLM) to break down complex queries and a visual grounding tool to identify objects in the scene. The method does not require labeled training data and achieves state-of-the-art accuracy on the ScanRefer benchmark.

Neurips 2023 spotlight - Unified Embedding: Battle-Tested Feature Representations for Web-Scale ML Systems

In this episode we discuss Unified Embedding: Battle-Tested Feature Representations for Web-Scale ML Systems
by Benjamin Coleman, Wang-Cheng Kang, Matthew Fahrbach, Ruoxi Wang, Lichan Hong, Ed H. Chi, Derek Zhiyuan Cheng. The paper introduces a framework called Feature Multiplexing, which allows for the use of a single representation space across multiple categorical features in web-scale machine learning systems. This framework addresses the high parameter count issue that arises from representing each feature value as a d-dimensional embedding. The paper also proposes a practical approach called Unified Embedding, which simplifies feature configuration, adapts to dynamic data distributions, and is compatible with modern hardware. The effectiveness of Unified Embedding is demonstrated in improving offline and online metrics across various web-scale systems.

arxiv Preprint - Chain-of-Verification Reduces Hallucination in Large Language Models

In this episode we discuss Chain-of-Verification Reduces Hallucination in Large Language Models
by Shehzaad Dhuliawala, Mojtaba Komeili, Jing Xu, Roberta Raileanu, Xian Li, Asli Celikyilmaz, Jason Weston. The paper proposes the Chain-of-Verification (COVE) method to address the issue of factual hallucination in large language models. COVE involves generating an initial response, planning independent fact-checking questions, and generating a final verified response. The experiments demonstrate that COVE reduces hallucinations in various tasks and variations of the method further improve performance.

arxiv Preprint - Language Modeling Is Compression

In this episode we discuss Language Modeling Is Compression
by Grégoire Delétang, Anian Ruoss, Paul-Ambroise Duquenne, Elliot Catt, Tim Genewein, Christopher Mattern, Jordi Grau-Moya, Li Kevin Wenliang, Matthew Aitchison, Laurent Orseau, Marcus Hutter, Joel Veness. The authors argue that large language models can be seen as powerful compressors due to their predictive capabilities. They demonstrate that these models outperform specific compressors like PNG and FLAC. The paper explores the implications of the prediction-compression equivalence and discusses the use of any compressor to build a conditional generative model.

arxiv Preprint - From Sparse to Dense: GPT-4 Summarization with Chain of Density Prompting

In this episode we discuss From Sparse to Dense: GPT-4 Summarization with Chain of Density Prompting
by Griffin Adams, Alexander Fabbri, Faisal Ladhak, Eric Lehman, Noémie Elhadad. The paper introduces a method called "Chain of Density" (CoD) for generating summaries with varying levels of information density. Using GPT-4, the authors generate entity-sparse summaries and then iteratively add missing salient entities without increasing the length. CoD summaries are found to be more abstractive, exhibit more fusion, and have less lead bias compared to GPT-4 summaries generated by a vanilla prompt, with human preference favoring denser GPT-4 summaries.

ICCV 2023 - Multimodal Garment Designer: Human-Centric Latent Diffusion Models for Fashion Image Editing

In this episode we discuss Multimodal Garment Designer: Human-Centric Latent Diffusion Models for Fashion Image Editing
by Alberto Baldrati, Davide Morelli, Giuseppe Cartella, Marcella Cornia, Marco Bertini, Rita Cucchiara. This paper introduces a framework for generating fashion images using multimodal prompts such as text, body poses, and garment sketches. The proposed architecture utilizes latent diffusion models, a novel approach in the fashion domain. The effectiveness of the framework is demonstrated through experiments using extended fashion datasets, showing realistic and coherent results.

arxiv Preprint - GPT Can Solve Mathematical Problems Without a Calculator

In this episode we discuss GPT Can Solve Mathematical Problems Without a Calculator
by Zhen Yang, Ming Ding, Qingsong Lv, Zhihuan Jiang, Zehai He, Yuyi Guo, Jinfeng Bai, Jie Tang. The paper challenges the belief that large language models cannot perform arithmetic operations accurately without calculator tools. The researchers present MathGLM, a 2 billion-parameter language model, which achieves nearly 100% accuracy in multi-digit arithmetic operations, surpassing GPT-4. They demonstrate the model's capability by training it on a dataset containing multi-step arithmetic operations and math problems described in text, and it performs similarly to GPT-4 on a Chinese math problem test set. The results suggest that language models can excel in mathematical problem-solving without the need for calculators, given sufficient training data.

ICCV 2023 - Adding Conditional Control to Text-to-Image Diffusion Models

In this episode we discuss Adding Conditional Control to Text-to-Image Diffusion Models
by Lvmin Zhang, Anyi Rao, Maneesh Agrawala. The paper introduces ControlNet, a neural network architecture that enhances spatial control in text-to-image diffusion models. It incorporates additional images, such as edge maps and human pose skeletons, as conditioning factors to specify desired image composition. ControlNet utilizes pretrained encoding layers and gradually adjusts parameters for improved spatial control. Experimental results with different conditioning controls and datasets demonstrate the effectiveness of ControlNet, suggesting its potential to broaden the applications of image diffusion models.

arxiv Preprint - Neurons in Large Language Models: Dead, N-gram, Positional

In this episode we discuss Neurons in Large Language Models: Dead, N-gram, Positional
by Elena Voita, Javier Ferrando, Christoforos Nalmpantis. In this paper, the authors analyze a family of language models called OPT models and focus on the activation of neurons in the feedforward blocks. They find that there are many inactive "dead" neurons in the early part of the network and that active neurons in this region primarily act as token and n-gram detectors. The authors also identify positional neurons that are activated based on position rather than textual data.