Modular vision-language models (Vision-LLMs) align pretrained image encoders with frozen large language models (LLMs), representing a computationally much more efficient alternative to end-to-end training of large vision-language models from scratch, which is prohibitively expensive for most researchers and practitioners. Vision-LLMs instead post-hoc condition LLMs to `understand' the output of an image encoder. With the abundance of readily available high-quality English image-text data as well as monolingual English LLMs, the research focus has been on English-only Vision-LLMs. Multilingual vision-language models are still predominantly obtained via expensive end-to-end pretraining, resulting in comparatively smaller models, trained on limited multilingual image data supplemented with text-only multilingual corpora. In this work, we present mBLIP, the first multilingual Vision-LLM, which we obtain in a computationally efficient manner -- on consumer hardware and using only a few million training examples -- by leveraging a pretrained multilingual LLM. To this end, we \textit{re-align} an image encoder previously tuned to an English LLM to a new, multilingual LLM -- for this, we leverage multilingual data from a mix of vision-and-language tasks, which we obtain by machine-translating high-quality English data to 95 languages. On the IGLUE benchmark, mBLIP yields results competitive with state-of-the-art models. Moreover, in image captioning on XM3600, mBLIP (zero-shot) even outperforms PaLI-X (a model with 55B parameters). Compared to these very large multilingual vision-language models trained from scratch, we obtain mBLIP by training orders of magnitude fewer parameters on magnitudes less data. We release our model and code at \url{https://github.com/gregor-ge/mBLIP}.

Large language models (LLMs) have achieved remarkable success in the field of natural language processing, enabling better human-computer interaction using natural language. However, the seamless integration of speech signals into LLMs has not been explored well. The "decoder-only" architecture has also not been well studied for speech processing tasks. In this research, we introduce Speech-LLaMA, a novel approach that effectively incorporates acoustic information into text-based large language models. Our method leverages Connectionist Temporal Classification and a simple audio encoder to map the compressed acoustic features to the continuous semantic space of the LLM. In addition, we further probe the decoder-only architecture for speech-to-text tasks by training a smaller scale randomly initialized speech-LLaMA model from speech-text paired data alone. We conduct experiments on multilingual speech-to-text translation tasks and demonstrate a significant improvement over strong baselines, highlighting the potential advantages of decoder-only models for speech-to-text conversion.

Most state-of-the-art techniques for medical image segmentation rely on deep-learning models. These models, however, are often trained on narrowly-defined tasks in a supervised fashion, which requires expensive labeled datasets. Recent advances in several machine learning domains, such as natural language generation have demonstrated the feasibility and utility of building foundation models that can be customized for various downstream tasks with little to no labeled data. This likely represents a paradigm shift for medical imaging, where we expect that foundation models may shape the future of the field. In this paper, we consider a recently developed foundation model for medical image segmentation, UniverSeg [6]. We conduct an empirical evaluation study in the context of prostate imaging and compare it against the conventional approach of training a task-specific segmentation model. Our results and discussion highlight several important factors that will likely be important in the development and adoption of foundation models for medical image segmentation.

Large Language Models (LLMs) have made extraordinary progress in the field of Artificial Intelligence and have demonstrated remarkable capabilities across a large variety of tasks and domains. However, as we venture closer to creating Artificial General Intelligence (AGI) systems, we recognize the need to supplement LLMs with long-term memory to overcome the context window limitation and more importantly, to create a foundation for sustained reasoning, cumulative learning and long-term user interaction. In this paper we propose RecallM, a novel architecture for providing LLMs with an adaptable and updatable long-term memory mechanism. Unlike previous methods, the RecallM architecture is particularly effective at belief updating and maintaining a temporal understanding of the knowledge provided to it. We demonstrate through various experiments the effectiveness of this architecture. Furthermore, through our own temporal understanding and belief updating experiments, we show that RecallM is four times more effective than using a vector database for updating knowledge previously stored in long-term memory. We also demonstrate that RecallM shows competitive performance on general question-answering and in-context learning tasks.

We introduce phi-1, a new large language model for code, with significantly smaller size than competing models: phi-1 is a Transformer-based model with 1.3B parameters, trained for 4 days on 8 A100s, using a selection of ``textbook quality" data from the web (6B tokens) and synthetically generated textbooks and exercises with GPT-3.5 (1B tokens). Despite this small scale, phi-1 attains pass@1 accuracy 50.6% on HumanEval and 55.5% on MBPP. It also displays surprising emergent properties compared to phi-1-base, our model before our finetuning stage on a dataset of coding exercises, and phi-1-small, a smaller model with 350M parameters trained with the same pipeline as phi-1 that still achieves 45% on HumanEval.

By providing external information to large language models (LLMs), tool augmentation (including retrieval augmentation) has emerged as a promising solution for addressing the limitations of LLMs' static parametric memory. However, how receptive are LLMs to such external evidence, especially when the evidence conflicts with their parametric memory? We present the first comprehensive and controlled investigation into the behavior of LLMs when encountering knowledge conflicts. We propose a systematic framework to elicit high-quality parametric memory from LLMs and construct the corresponding counter-memory, which enables us to conduct a series of controlled experiments. Our investigation reveals seemingly contradicting behaviors of LLMs. On the one hand, different from prior wisdom, we find that LLMs can be highly receptive to external evidence even when that conflicts with their parametric memory, given that the external evidence is coherent and convincing. On the other hand, LLMs also demonstrate a strong confirmation bias when the external evidence contains some information that is consistent with their parametric memory, despite being presented with conflicting evidence at the same time. These results pose important implications that are worth careful consideration for the further development and deployment of tool- and retrieval-augmented LLMs.

Our work addresses the critical issue of distinguishing text generated by Large Language Models (LLMs) from human-produced text, a task essential for numerous applications. Despite ongoing debate about the feasibility of such differentiation, we present evidence supporting its consistent achievability, except when human and machine text distributions are indistinguishable across their entire support. Drawing from information theory, we argue that as machine-generated text approximates human-like quality, the sample size needed for detection increases. We establish precise sample complexity bounds for detecting AI-generated text, laying groundwork for future research aimed at developing advanced, multi-sample detectors. Our empirical evaluations across multiple datasets (Xsum, Squad, IMDb, and Kaggle FakeNews) confirm the viability of enhanced detection methods. We test various state-of-the-art text generators, including GPT-2, GPT-3.5-Turbo, Llama, Llama-2-13B-Chat-HF, and Llama-2-70B-Chat-HF, against detectors, including oBERTa-Large/Base-Detector, GPTZero. Our findings align with OpenAI's empirical data related to sequence length, marking the first theoretical substantiation for these observations.

Recent endeavors in video editing have showcased promising results in single-attribute editing or style transfer tasks, either by training text-to-video (T2V) models on text-video data or adopting training-free methods. However, when confronted with the complexities of multi-attribute editing scenarios, they exhibit shortcomings such as omitting or overlooking intended attribute changes, modifying the wrong elements of the input video, and failing to preserve regions of the input video that should remain intact. To address this, here we present a novel grounding-guided video-to-video translation framework called Ground-A-Video for multi-attribute video editing. Ground-A-Video attains temporally consistent multi-attribute editing of input videos in a training-free manner without aforementioned shortcomings. Central to our method is the introduction of Cross-Frame Gated Attention which incorporates groundings information into the latent representations in a temporally consistent fashion, along with Modulated Cross-Attention and optical flow guided inverted latents smoothing. Extensive experiments and applications demonstrate that Ground-A-Video's zero-shot capacity outperforms other baseline methods in terms of edit-accuracy and frame consistency. Further results and codes are provided at our project page (http://ground-a-video.github.io).

Multi-modal learning (Ngiam et al., 2011) integrates information from a variety of data types, resulting in AI systems that are both robust and precise. Recently, CLIP (Radford et al., 2021) emerged as a milestone work that leverages vision-language contrastive pretraining to jointly learn image and text embeddings, using the vast amounts of image-text data available on the web. During the training process, CLIP considers image-text data that appear together as positive pairs and other combinations as negative pairs. The goal is to maximize the embedding similarity for the positive pairs while minimizing it for the negative pairs. Remarkably, this approach has achieved significant success in zero-shot transfer (Lei Ba et al., 2015), indicating the model's ability to handle a great variety of tasks without prior exposure to any of their training data. Inspired by CLIP's groundbreaking zero-shot capabilities, subsequent studies (Yao et al., 2022; Li et al., 2022; Mu et al., 2022; Goel et al., 2022; Zhai et al., 2022; Alayrac et al., 2022) emerged with the primary objective of further enhancing CLIP's zero-shot performance. Despite the empirical success of CLIP in zero-shot transfer, the theoretical understanding of how it works remains elusive. An intriguing inquiry is thus: How does CLIP learn representations that are transferable to the various downstream tasks?

Over the last decades, excellent computational chemistry tools have been developed. Integrating them into a single platform with enhanced accessibility could help reaching their full potential by overcoming steep learning curves. Recently, large-language models (LLMs) have shown strong performance in tasks across domains, but struggle with chemistry-related problems. Moreover, these models lack access to external knowledge sources, limiting their usefulness in scientific applications. In this study, we introduce ChemCrow, an LLM chemistry agent designed to accomplish tasks across organic synthesis, drug discovery, and materials design. By integrating 18 expert-designed tools, ChemCrow augments the LLM performance in chemistry, and new capabilities emerge. Our agent autonomously planned and executed the syntheses of an insect repellent, three organocatalysts, and guided the discovery of a novel chromophore. Our evaluation, including both LLM and expert assessments, demonstrates ChemCrow's effectiveness in automating a diverse set of chemical tasks. Surprisingly, we find that GPT-4 as an evaluator cannot distinguish between clearly wrong GPT-4 completions and Chemcrow's performance. Our work not only aids expert chemists and lowers barriers for non-experts, but also fosters scientific advancement by bridging the gap between experimental and computational chemistry.