Nous utilisons des témoins pour analyser le trafic et l’utilisation de notre site web, afin de personnaliser votre expérience. Vous pouvez désactiver ces technologies à tout moment, mais cela peut restreindre certaines fonctionnalités du site. Consultez notre Politique de protection de la vie privée pour en savoir plus.
Paramètre des cookies
Vous pouvez activer et désactiver les types de cookies que vous souhaitez accepter. Cependant certains choix que vous ferez pourraient affecter les services proposés sur nos sites (ex : suggestions, annonces personnalisées, etc.).
Cookies essentiels
Ces cookies sont nécessaires au fonctionnement du site et ne peuvent être désactivés. (Toujours actif)
Cookies analyse
Acceptez-vous l'utilisation de cookies pour mesurer l'audience de nos sites ?
Multimedia Player
Acceptez-vous l'utilisation de cookies pour afficher et vous permettre de regarder les contenus vidéo hébergés par nos partenaires (YouTube, etc.) ?
Publications
A Deep Dive into the Trade-Offs of Parameter-Efficient Preference Alignment Techniques
Large language models are first pre-trained on trillions of tokens and then instruction-tuned or aligned to specific preferences. While pre-… (voir plus)training remains out of reach for most researchers due to the compute required, fine-tuning has become affordable thanks to parameter-efficient methods such as LoRA and QLoRA. Alignment is known to be sensitive to the many factors involved, including the quantity and quality of data, the alignment method, and the adapter rank. However, there has not yet been an extensive study of their effect on downstream performance. To address this gap, we conduct an in-depth investigation of the impact of popular choices for three crucial axes: (i) the alignment dataset (HH-RLHF and BeaverTails), (ii) the alignment technique (SFT and DPO), and (iii) the model (LLaMA-1, Vicuna-v1.3, Mistral-7b, and Mistral-7b-Instruct). Our extensive setup spanning over 300 experiments reveals consistent trends and unexpected findings. We observe how more informative data helps with preference alignment, cases where supervised fine-tuning outperforms preference optimization, and how aligning to a distinct preference boosts performance on downstream tasks. Through our in-depth analyses, we put forward key guidelines to help researchers perform more effective parameter-efficient LLM alignment.
Large language models are first pre-trained on trillions of tokens and then instruction-tuned or aligned to specific preferences. While pre-… (voir plus)training remains out of reach for most researchers due to the compute required, fine-tuning has become affordable thanks to parameter-efficient methods such as LoRA and QLoRA. Alignment is known to be sensitive to the many factors involved, including the quantity and quality of data, the alignment method, and the adapter rank. However, there has not yet been an extensive study of their effect on downstream performance. To address this gap, we conduct an in-depth investigation of the impact of popular choices for three crucial axes: (i) the alignment dataset (HH-RLHF and BeaverTails), (ii) the alignment technique (SFT and DPO), and (iii) the model (LLaMA-1, Vicuna-v1.3, Mistral-7b, and Mistral-7b-Instruct). Our extensive setup spanning over 300 experiments reveals consistent trends and unexpected findings. We observe how more informative data helps with preference alignment, cases where supervised fine-tuning outperforms preference optimization, and how aligning to a distinct preference boosts performance on downstream tasks. Through our in-depth analyses, we put forward key guidelines to help researchers perform more effective parameter-efficient LLM alignment.
Lifelong Learning of Video Diffusion Models From a Single Video Stream
Jinsoo Yoo
Yingchen He
Saeid Naderiparizi
Dylan Green
Gido M. van de Ven
Geoff Pleiss
F. Wood
This work demonstrates that training autoregressive video diffusion models from a single, continuous video stream is not only possible but r… (voir plus)emarkably can also be competitive with standard offline training approaches given the same number of gradient steps. Our demonstration further reveals that this main result can be achieved using experience replay that only retains a subset of the preceding video stream. We also contribute three new single video generative modeling datasets suitable for evaluating lifelong video model learning: Lifelong Bouncing Balls, Lifelong 3D Maze, and Lifelong PLAICraft. Each dataset contains over a million consecutive frames from a synthetic environment of increasing complexity.
Continual Reinforcement Learning (CRL) aims to develop algorithms that adapt to non-stationary sequences of tasks. A promising recent approa… (voir plus)ch utilizes Recurrent Neural Networks (RNNs) to learn contextual Markov Decision Process (MDP) embeddings. This enables a reinforcement learning (RL) agent to discern the optimality of actions across diverse tasks. In this study, we examine two critical failure modes in the learning of these contextual MDP embeddings. Specifically, we find that RNNs are prone to catastrophic forgetting, manifesting in two distinct ways: (i) embedding collapse---where agents initially learn a contextual task structure that later collapses to a single task, and (ii) embedding drift---where learning embeddings for new MDPs interferes with embeddings the RNN outputs for previous MDPs in the sequence, leading to suboptimal performance of downstream policy networks conditioned on stale embeddings. We explore the effects of various objective functions and network architectures concerning these failure modes, revealing that one of these modes consistently emerges across different setups.
Generating novel active molecules for a given protein is an extremely challenging task for generative models that requires an understanding … (voir plus)of the complex physical interactions between the molecule and its environment. In this paper, we present a novel generative model, BindGPT which uses a conceptually simple but powerful approach to create 3D molecules within the protein's binding site. Our model produces molecular graphs and conformations jointly, eliminating the need for an extra graph reconstruction step. We pretrain BindGPT on a large-scale dataset and fine-tune it with reinforcement learning using scores from external simulation software. We demonstrate how a single pretrained language model can serve at the same time as a 3D molecular generative model, conformer generator conditioned on the molecular graph, and a pocket-conditioned 3D molecule generator. Notably, the model does not make any representational equivariance assumptions about the domain of generation. We show how such simple conceptual approach combined with pretraining and scaling can perform on par or better than the current best specialized diffusion models, language models, and graph neural networks while being two orders of magnitude cheaper to sample.
The combination of unoccupied aerial vehicles (UAVs) and artificial intelligence to map vegetation represents a promising new approach to im… (voir plus)prove the detection of invasive alien plant species (IAPS). The high spatial resolution achievable with UAVs and recent innovations in computer vision, especially with convolutional neural networks, suggest that early detection of IAPS could be possible, thus facilitating their management. In this study, we evaluated the suitability of this approach for mapping the location of common reed (Phragmites australis subsp. australis) within a national park located in southern Quebec, Canada. We collected data on six distinct dates during the growing season, covering environments with different levels of reed invasion. Overall, model performance was high for the different dates and zones, especially for recall (mean of 0.89). The results showed an increase in performance, reaching a peak following the appearance of the inflorescence in September (highest F1-score at 0.98). Furthermore, a decrease in spatial resolution negatively affected recall (18% decrease between a spatial resolution of 0.15 cm pixel−1 and 1.50 cm pixel−1) but did not have a strong impact on precision (2% decrease). Despite challenges associated with common reed mapping in a post-treatment monitoring context, the use of UAVs and deep learning shows great potential for IAPS detection when supported by a suitable dataset. Our results show that, from an operational point of view, this approach could be an effective tool for speeding up the work of biologists in the field and ensuring better management of IAPS.
With the growing popularity of text-to-image generative models, there has been increasing focus on understanding their risks and biases. Rec… (voir plus)ent work has found that state-of-the-art models struggle to depict everyday objects with the true diversity of the real world and have notable gaps between geographic regions. In this work, we aim to increase the diversity of generated images of common objects such that per-region variations are representative of the real world. We introduce an inference time intervention, contextualized Vendi Score Guidance (c-VSG), that guides the backwards steps of latent diffusion models to increase the diversity of a sample as compared to a"memory bank"of previously generated images while constraining the amount of variation within that of an exemplar set of real-world contextualizing images. We evaluate c-VSG with two geographically representative datasets and find that it substantially increases the diversity of generated images, both for the worst performing regions and on average, while simultaneously maintaining or improving image quality and consistency. Additionally, qualitative analyses reveal that diversity of generated images is significantly improved, including along the lines of reductive region portrayals present in the original model. We hope that this work is a step towards text-to-image generative models that reflect the true geographic diversity of the world.
External audits of AI systems are increasingly recognized as a key mechanism for AI governance. The effectiveness of an audit, however, depe… (voir plus)nds on the degree of system access granted to auditors. Recent audits of state-of-the-art AI systems have primarily relied on black-box access, in which auditors can only query the system and observe its outputs. However, white-box access to the system's inner workings (e.g., weights, activations, gradients) allows an auditor to perform stronger attacks, more thoroughly interpret models, and conduct fine-tuning. Meanwhile, outside-the-box access to its training and deployment information (e.g., methodology, code, documentation, hyperparameters, data, deployment details, findings from internal evaluations) allows for auditors to scrutinize the development process and design more targeted evaluations. In this paper, we examine the limitations of black-box audits and the advantages of white- and outside-the-box audits. We also discuss technical, physical, and legal safeguards for performing these audits with minimal security risks. Given that different forms of access can lead to very different levels of evaluation, we conclude that (1) transparency regarding the access and methods used by auditors is necessary to properly interpret audit results, and (2) white- and outside-the-box access allow for substantially more scrutiny than black-box access alone.
2024-06-05
The 2024 ACM Conference on Fairness, Accountability, and Transparency (publié)