Code datasets, often collected from diverse and uncontrolled sources such as GitHub, potentially suffer from quality issues, thereby affecting the performance and training efficiency of Large Language Models (LLMs) optimized for code generation. Previous studies demonstrated the benefit of using embedding spaces for data pruning, but they mainly focused on duplicate removal or increasing variety, and in other modalities, such as images. Our work focuses on using embeddings to identify and remove "low-quality" code data. First, we explore features of "low-quality" code in embedding space, through the use of synthetic corruptions. Armed with this knowledge, we devise novel pruning metrics that operate in embedding space to identify and remove low-quality entries in the Stack dataset. We demonstrate the benefits of this synthetic corruption informed pruning (SCIP) approach on the well-established HumanEval and MBPP benchmarks, outperforming existing embedding-based methods. Importantly, we achieve up to a 3% performance improvement over no pruning, thereby showing the promise of insights from synthetic corruptions for data pruning.
Vision-Language Models (VLMs) are pretrained on large, diverse, and noisy web-crawled datasets. This underscores the critical need for dataset pruning, as the quality of these datasets is strongly correlated with the performance of VLMs on downstream tasks. Using CLIPScore from a pretrained model to only train models using highly-aligned samples is one of the most successful methods for pruning.We argue that this approach suffers from multiple limitations including: 1) false positives due to spurious correlations captured by the pretrained CLIP model, 2) false negatives due to poor discrimination between hard and bad samples, and 3) biased ranking towards samples similar to the pretrained CLIP dataset. We propose a pruning method, SIEVE, that employs synthetic captions generated by image-captioning models pretrained on small, diverse, and well-aligned image-text pairs to evaluate the alignment of noisy image-text pairs. To bridge the gap between the limited diversity of generated captions and the high diversity of alternative text (alt-text), we estimate the semantic textual similarity in the embedding space of a language model pretrained on billions of sentences. Using DataComp, a multimodal dataset filtering benchmark, we achieve state-of-the-art performance on the large scale pool, and competitive results on the medium scale pool, surpassing CLIPScore-based filtering by 1.7% and 2.6% on average, on 38 downstream tasks.
An effective framework for learning 3D representations for perception tasks is distilling rich self-supervised image features via contrastive learning. However, image-to point representation learning for autonomous driving datasets faces two main challenges: 1) the abundance of self-similarity, which results in the contrastive losses pushing away semantically similar point and image regions and thus disturbing the local semantic structure of the learned representations, and 2) severe class imbalance as pretraining gets dominated by over-represented classes. We propose to alleviate the self-similarity problem through a novel semantically tolerant image-to-point contrastive loss that takes into consideration the semantic distance between positive and negative image regions to minimize contrasting semantically similar point and image regions. Additionally, we address class imbalance by designing a class-agnostic balanced loss that approximates the degree of class imbalance through an aggregate sample-to-samples semantic similarity measure. We demonstrate that our semantically-tolerant contrastive loss with class balancing improves state-of-the art 2D-to-3D representation learning in all evaluation settings on 3D semantic segmentation. Our method consistently outperforms state-of-the-art 2D-to-3D representation learning frameworks across a wide range of 2D self-supervised pretrained models.
Camera and LiDAR sensor modalities provide complementary appearance and geometric information useful for detecting 3D objects for autonomous vehicle applications. However, current fusion models underperform state-of-art LiDAR-only methods on 3D object detection benchmarks. Our proposed solution, Dense Voxel Fusion (DVF) is a sequential fusion method that generates multi-scale multi-modal dense voxel feature representations, improving expressiveness in low point density regions. To enhance multi-modal learning, we train directly with ground truth 2D bounding box labels, avoiding noisy, detector-specific, 2D predictions. Additionally, we use LiDAR ground truth sampling to simulate missed 2D detections and to accelerate training convergence. Both DVF and the multi-modal training approaches can be applied to any voxel-based LiDAR backbone without introducing additional learnable parameters. DVF outperforms existing sparse fusion detectors, ranking $1^{st}$ among all published fusion methods on KITTI's 3D car detection benchmark at the time of submission and significantly improves 3D vehicle detection performance of voxel-based methods on the Waymo Open Dataset. We also show that our proposed multi-modal training strategy results in better generalization compared to training using erroneous 2D predictions.