Skywork-Reward: Bag of Tricks for Reward Modeling in LLMs

In this report, we introduce a collection of methods to enhance reward modeling for LLMs, focusing specifically on data-centric techniques. We propose effective data selection and filtering strategies for curating high-quality open-source preference datasets, culminating in the Skywork-Reward data collection, which contains only 80K preference pairs -- significantly smaller than existing datasets. Using this curated dataset, we developed the Skywork-Reward model series -- Skywork-Reward-Gemma-27B and Skywork-Reward-Llama-3.1-8B -- with the former currently holding the top position on the RewardBench leaderboard. Notably, our techniques and datasets have directly enhanced the performance of many top-ranked models on RewardBench, highlighting the practical impact of our contributions in real-world preference learning applications.

Skywork-Math: Data Scaling Laws for Mathematical Reasoning in Large Language Models -- The Story Goes On

In this paper, we investigate the underlying factors that potentially enhance the mathematical reasoning capabilities of large language models (LLMs). We argue that the data scaling law for math reasoning capabilities in modern LLMs is far from being saturated, highlighting how the model's quality improves with increases in data quantity. To support this claim, we introduce the Skywork-Math model series, supervised fine-tuned (SFT) on common 7B LLMs using our proposed 2.5M-instance Skywork-MathQA dataset. Skywork-Math 7B has achieved impressive accuracies of 51.2% on the competition-level MATH benchmark and 83.9% on the GSM8K benchmark using only SFT data, outperforming an early version of GPT-4 on MATH. The superior performance of Skywork-Math models contributes to our novel two-stage data synthesis and model SFT pipelines, which include three different augmentation methods and a diverse seed problem set, ensuring both the quantity and quality of Skywork-MathQA dataset across varying difficulty levels. Most importantly, we provide several practical takeaways to enhance math reasoning abilities in LLMs for both research and industry applications.

Skywork-MoE: A Deep Dive into Training Techniques for Mixture-of-Experts Language Models

In this technical report, we introduce the training methodologies implemented in the development of Skywork-MoE, a high-performance mixture-of-experts (MoE) large language model (LLM) with 146 billion parameters and 16 experts. It is initialized from the pre-existing dense checkpoints of our Skywork-13B model. We explore the comparative effectiveness of upcycling versus training from scratch initializations. Our findings suggest that the choice between these two approaches should consider both the performance of the existing dense checkpoints and the MoE training budget. We highlight two innovative techniques: gating logit normalization, which improves expert diversification, and adaptive auxiliary loss coefficients, allowing for layer-specific adjustment of auxiliary loss coefficients. Our experimental results validate the effectiveness of these methods. Leveraging these techniques and insights, we trained our upcycled Skywork-MoE on a condensed subset of our SkyPile corpus. The evaluation results demonstrate that our model delivers strong performance across a wide range of benchmarks.

LongSkywork: A Training Recipe for Efficiently Extending Context Length in Large Language Models

We introduce LongSkywork, a long-context Large Language Model (LLM) capable of processing up to 200,000 tokens. We provide a training recipe for efficiently extending context length of LLMs. We identify that the critical element in enhancing long-context processing capability is to incorporate a long-context SFT stage following the standard SFT stage. A mere 200 iterations can convert the standard SFT model into a long-context model. To reduce the effort in collecting and annotating data for long-context language modeling, we develop two novel methods for creating synthetic data. These methods are applied during the continual pretraining phase as well as the Supervised Fine-Tuning (SFT) phase, greatly enhancing the training efficiency of our long-context LLMs. Our findings suggest that synthetic long-context SFT data can surpass the performance of data curated by humans to some extent. LongSkywork achieves outstanding performance on a variety of long-context benchmarks. In the Needle test, a benchmark for long-context information retrieval, our models achieved perfect accuracy across multiple context spans. Moreover, in realistic application scenarios, LongSkywork-13B demonstrates performance on par with Claude2.1, the leading long-context model, underscoring the effectiveness of our proposed methods.

KeNet:Knowledge-enhanced Doc-Label Attention Network for Multi-label text classification

Multi-Label Text Classification (MLTC) is a fundamental task in the field of Natural Language Processing (NLP) that involves the assignment of multiple labels to a given text. MLTC has gained significant importance and has been widely applied in various domains such as topic recognition, recommendation systems, sentiment analysis, and information retrieval. However, traditional machine learning and Deep neural network have not yet addressed certain issues, such as the fact that some documents are brief but have a large number of labels and how to establish relationships between the labels. It is imperative to additionally acknowledge that the significance of knowledge is substantiated in the realm of MLTC. To address this issue, we provide a novel approach known as Knowledge-enhanced Doc-Label Attention Network (KeNet). Specifically, we design an Attention Network that incorporates external knowledge, label embedding, and a comprehensive attention mechanism. In contrast to conventional methods, we use comprehensive representation of documents, knowledge and labels to predict all labels for each single text. Our approach has been validated by comprehensive research conducted on three multi-label datasets. Experimental results demonstrate that our method outperforms state-of-the-art MLTC method. Additionally, a case study is undertaken to illustrate the practical implementation of KeNet.

MolKD: Distilling Cross-Modal Knowledge in Chemical Reactions for Molecular Property Prediction

How to effectively represent molecules is a long-standing challenge for molecular property prediction and drug discovery. This paper studies this problem and proposes to incorporate chemical domain knowledge, specifically related to chemical reactions, for learning effective molecular representations. However, the inherent cross-modality property between chemical reactions and molecules presents a significant challenge to address. To this end, we introduce a novel method, namely MolKD, which Distills cross-modal Knowledge in chemical reactions to assist Molecular property prediction. Specifically, the reaction-to-molecule distillation model within MolKD transfers cross-modal knowledge from a pre-trained teacher network learning with one modality (i.e., reactions) into a student network learning with another modality (i.e., molecules). Moreover, MolKD learns effective molecular representations by incorporating reaction yields to measure transformation efficiency of the reactant-product pair when pre-training on reactions. Extensive experiments demonstrate that MolKD significantly outperforms various competitive baseline models, e.g., 2.1% absolute AUC-ROC gain on Tox21. Further investigations demonstrate that pre-trained molecular representations in MolKD can distinguish chemically reasonable molecular similarities, which enables molecular property prediction with high robustness and interpretability.

Copy-Pasting Coherent Depth Regions Improves Contrastive Learning for Urban-Scene Segmentation

In this work, we leverage estimated depth to boost self-supervised contrastive learning for segmentation of urban scenes, where unlabeled videos are readily available for training self-supervised depth estimation. We argue that the semantics of a coherent group of pixels in 3D space is self-contained and invariant to the contexts in which they appear. We group coherent, semantically related pixels into coherent depth regions given their estimated depth and use copy-paste to synthetically vary their contexts. In this way, cross-context correspondences are built in contrastive learning and a context-invariant representation is learned. For unsupervised semantic segmentation of urban scenes, our method surpasses the previous state-of-the-art baseline by +7.14% in mIoU on Cityscapes and +6.65% on KITTI. For fine-tuning on Cityscapes and KITTI segmentation, our method is competitive with existing models, yet, we do not need to pre-train on ImageNet or COCO, and we are also more computationally efficient. Our code is available on https://github.com/LeungTsang/CPCDR

ImDrug: A Benchmark for Deep Imbalanced Learning in AI-aided Drug Discovery

The last decade has witnessed a prosperous development of computational methods and dataset curation for AI-aided drug discovery (AIDD). However, real-world pharmaceutical datasets often exhibit highly imbalanced distribution, which is largely overlooked by the current literature but may severely compromise the fairness and generalization of machine learning applications. Motivated by this observation, we introduce ImDrug, a comprehensive benchmark with an open-source Python library which consists of 4 imbalance settings, 11 AI-ready datasets, 54 learning tasks and 16 baseline algorithms tailored for imbalanced learning. It provides an accessible and customizable testbed for problems and solutions spanning a broad spectrum of the drug discovery pipeline such as molecular modeling, drug-target interaction and retrosynthesis. We conduct extensive empirical studies with novel evaluation metrics, to demonstrate that the existing algorithms fall short of solving medicinal and pharmaceutical challenges in the data imbalance scenario. We believe that ImDrug opens up avenues for future research and development, on real-world challenges at the intersection of AIDD and deep imbalanced learning.

ImGCL: Revisiting Graph Contrastive Learning on Imbalanced Node Classification

Graph contrastive learning (GCL) has attracted a surge of attention due to its superior performance for learning node/graph representations without labels. However, in practice, unlabeled nodes for the given graph usually follow an implicit imbalanced class distribution, where the majority of nodes belong to a small fraction of classes (a.k.a., head class) and the rest classes occupy only a few samples (a.k.a., tail classes). This highly imbalanced class distribution inevitably deteriorates the quality of learned node representations in GCL. Indeed, we empirically find that most state-of-the-art GCL methods exhibit poor performance on imbalanced node classification. Motivated by this observation, we propose a principled GCL framework on Imbalanced node classification (ImGCL), which automatically and adaptively balances the representation learned from GCL without knowing the labels. Our main inspiration is drawn from the recent progressively balanced sampling (PBS) method in the computer vision domain. We first introduce online clustering based PBS, which balances the training sets based on pseudo-labels obtained from learned representations. We then develop the node centrality based PBS method to better preserve the intrinsic structure of graphs, which highlight the important nodes of the given graph. Besides, we theoretically consolidate our method by proving that the classifier learned by balanced sampling without labels on an imbalanced dataset can converge to the optimal balanced classifier with a linear rate. Extensive experiments on multiple imbalanced graph datasets and imbalance settings verify the effectiveness of our proposed framework, which significantly improves the performance of the recent state-of-the-art GCL methods. Further experimental ablations and analysis show that the ImGCL framework remarkably improves the representations of nodes in tail classes.

STformer: A Noise-Aware Efficient Spatio-Temporal Transformer Architecture for Traffic Forecasting

Traffic forecasting plays an indispensable role in the intelligent transportation system, which makes daily travel more convenient and safer. However, the dynamic evolution of spatio-temporal correlations makes accurate traffic forecasting very difficult. Existing work mainly employs graph neural netwroks (GNNs) and deep time series models (e.g., recurrent neural networks) to capture complex spatio-temporal patterns in the dynamic traffic system. For the spatial patterns, it is difficult for GNNs to extract the global spatial information, i.e., remote sensors information in road networks. Although we can use the self-attention to extract global spatial information as in the previous work, it is also accompanied by huge resource consumption. For the temporal patterns, traffic data have not only easy-to-recognize daily and weekly trends but also difficult-to-recognize short-term noise caused by accidents (e.g., car accidents and thunderstorms). Prior traffic models are difficult to distinguish intricate temporal patterns in time series and thus hard to get accurate temporal dependence. To address above issues, we propose a novel noise-aware efficient spatio-temporal Transformer architecture for accurate traffic forecasting, named STformer. STformer consists of two components, which are the noise-aware temporal self-attention (NATSA) and the graph-based sparse spatial self-attention (GBS3A). NATSA separates the high-frequency component and the low-frequency component from the time series to remove noise and capture stable temporal dependence by the learnable filter and the temporal self-attention, respectively. GBS3A replaces the full query in vanilla self-attention with the graph-based sparse query to decrease the time and memory usage. Experiments on four real-world traffic datasets show that STformer outperforms state-of-the-art baselines with lower computational cost.