The performance of a large language model (LLM) depends heavily on the quality and size of its pretraining dataset. However, the pretraining datasets for state-ofthe-art open LLMs like Llama 3 and Mixtral are not publicly available and very little is known about how they were created. In this work, we introduce FineWeb, a 15-trillion token dataset derived from 96 Common Crawl snapshots that produces better-performing LLMs than other open pretraining datasets. To advance the understanding of how best to curate high-quality pretraining datasets, we carefully document and ablate all of the design choices used in FineWeb, including indepth investigations of deduplication and filtering strategies. In addition, we introduce FineWeb-Edu, a 1.3-trillion token collection of educational text filtered from FineWeb.

For all authors... (a) Do the main claims made in the abstract and introduction accurately reflect the paper's contributions and scope? If you ran experiments... (a) Did you include the code, data, and instructions needed to reproduce the main experimental results (either in the supplemental material or as a URL)? [Yes] (b) Did you specify all the training details (e.g., data splits, hyperparameters, how they were chosen)? If you used crowdsourcing or conducted research with human subjects... (a) Did you include the full text of instructions given to participants and screenshots, if applicable? [N/A] (b) Did you describe any potential participant risks, with links to Institutional Review Board (IRB) approvals, if applicable? [N/A] (c) Did you include the estimated hourly wage paid to participants and the total amount spent on participant compensation? Graph Convolutional Networks (GCNs) could be crucial tools for a broad range of applications, including social networks, computer vision, natural language processing, traffic prediction, chemistry, protein design, recommendation system and so on [64, 58]. Any of these applications may have a different social effect. The use of GCNs could improve protein design efficiency and lead to the development of new medicines, but it could also result in job losses.

Table 6 provides summary statistics of domain coverage. Overall, the benchmark covers 8,637 biology images and 8,678 pathology images across 12 subdomains. Similarly, Table 7 shows summary statistics of microscopy modalities covered by Micro-Bench perception, including 10,864 images for light microscopy, 5,618 for fluorescence microscopy, and 833 images for electron microscopy across 8 microscopy imaging submodalities and 25 unique microscopy staining techniques (see Table 8). Micro-Bench Perception (Coarse-grained): Hierarchical metadata for each of the 17,235 perception images and task-specific templates (shown in Table 23) are used to create 5 coarse-grained questions and captions regarding microscopy modality, submodality, domain, subdomain, and staining technique. The use of hierarchical metadata enables the generation of options within each hierarchical level.

The problem of (point) forecasting univariate time series is considered. Most approaches, ranging from traditional statistical methods to recent learning-based techniques with neural networks, directly operate on raw time series observations. As an extension, we study whether local topological properties, as captured via persistent homology, can serve as a reliable signal that provides complementary information for learning to forecast. To this end, we propose topological attention, which allows attending to local topological features within a time horizon of historical data. Our approach easily integrates into existing end-to-end trainable forecasting models, such as N-BEATS, and, in combination with the latter, exhibits state-of-the-art performance on the large-scale M4 benchmark dataset of 100,000 diverse time series from different domains. Ablation experiments, as well as a comparison to a broad range of forecasting methods in a setting where only a single time series is available for training, corroborate the beneficial nature of including local topological information through an attention mechanism.

We compare with Indic and Non-Indic datasets. A.1 Comparison with existing datasets In this section, we compare our proposed MACD with existing datasets in detail in Table 10. We note that large scale datasets containing more than 50K samples exist for some non-Indic languages like English, Greek and Turkish language. These datasets enable large-scale study of abuse detection for these languages. However, for other languages, presence of large-scale datasets is still lacking. Next, we compare with Indic datasets and note that Indic datasets are small-scale as compared to non-Indic datasets. This shows that there is an immediate requirement for a dataset like MACD to fill this gap and foster advancements in abuse detection in Indic languages. Overall and at language level, MACD is one of the largest dataset for studying Indic languages. A.2 MACD dataset Explicit warning: We want to urge the community to be mindful of the fact that our dataset MACD contains comments which express abusive behaviour towards religion, region, gender etc. that might be abusive and depressing to the researchers.

Social media platforms were conceived to act as online'town squares' where people could get together, share information and communicate with each other peacefully. However, harmful content borne out of bad actors are constantly plaguing these platforms slowly converting them into'mosh pits' where the bad actors take the liberty to extensively abuse various marginalised groups. Accurate and timely detection of abusive content on social media platforms is therefore very important for facilitating safe interactions between users. However, due to the small scale and sparse linguistic coverage of Indic abusive speech datasets, development of such algorithms for Indic social media users (one-sixth of global population) is severely impeded.

A current remarkable improvement of unsupervised visual representation learning is based on heavy networks with large-batch training. While recent methods have greatly reduced the gap between supervised and unsupervised performance of deep models such as ResNet-50, this development has been relatively limited for small models. In this work, we propose a novel unsupervised learning framework for small networks that combines deep self-supervised representation learning and knowledge distillation within one-phase training. In particular, a teacher model is trained to produce consistent cluster assignments between different views of the same image. Simultaneously, a student model is encouraged to mimic the prediction of on-the-fly self-supervised teacher. For effective knowledge transfer, we adopt the idea of domain classifier so that student training is guided by discriminative features invariant to the representational space shift between teacher and student. We also introduce a network driven multi-view generation paradigm to capture rich feature information contained in the network itself. Extensive experiments show that our student models surpass state-of-the-art offline distilled networks even from stronger self-supervised teachers as well as top-performing self-supervised models. Notably, our ResNet-18, trained with ResNet-50 teacher, achieves 68.3% ImageNet Top-1 accuracy on frozen feature linear evaluation, which is only 1.5% below the supervised baseline.

Unless otherwise specified all algorithms were timed on single core versions even though, for instance, the proposed method is in places trivially parallelizable (i.e. during forest build). An exception was the grid search across meta-parameters to find the best (optimal) reference model where parallelization was used when required as this stage does not form part of the time comparisons. Hosted on Google Colaboratory they enable the use of hosted or local runtime environments. When tested hosted runtimes were running Python 3.6.9 Please note that while a hosted runtime can be used for ease of replication, all timings reported in the paper were based on using a local runtime environment as previously indicated NOT a hosted environment. The notebooks, when run in the hosted environment will automatically install the required packages developed as part of this work.

Efficient fine-tuning of large language models for task-specific applications is imperative, yet the vast number of parameters in these models makes their training increasingly challenging. Despite numerous proposals for effective methods, a substantial memory overhead remains for gradient computations during updates. Can we fine-tune a series of task-specific small models and transfer their knowledge directly to a much larger model without additional training? In this paper, we explore weak-to-strong specialization using logit arithmetic, facilitating a direct answer to this question. Existing weak-to-strong methods often employ a static knowledge transfer ratio and a single small model for transferring complex knowledge, which leads to suboptimal performance.

We consider the imitation learning problem of learning a policy in a Markov Decision Process (MDP) setting where the reward function is not given, but demonstrations from experts are available. Although the goal of imitation learning is to learn a policy that produces behaviors nearly as good as the experts' for a desired task, assumptions of consistent optimality for demonstrated behaviors are often violated in practice. Finding a policy that is distributionally robust against noisy demonstrations based on an adversarial construction potentially solves this problem by avoiding optimistic generalizations of the demonstrated data.