Improving semantic grounding in Vision-Language Models (VLMs) often involves collecting domain-specific training data, refining the network architectures, or modifying the training recipes. In this work, we venture into an orthogonal direction and explore self-correction in VLMs focusing on semantic grounding. We find that VLMs can correct their own semantic grounding mistakes when properly prompted and framed for the task, without any fine-tuning or even access to oracle feedback. We also introduce a self-correction framework in an iterative setting which consistently improves performance across all models investigated. Overall, we show that iterative self-correction consistently improves VLM performance in semantic grounding by up to 8.4 accuracy points across all models investigated, without requiring fine-tuning, additional architectural changes, or external data. Our exploration of self-correction also reveals that, even after several rounds of feedback, strong models like GPT-4V and GPT-4o retain limited capability in leveraging oracle feedback, suggesting promising directions for further research.

Modern deep learning systems require huge data sets to achieve impressive performance, but there is little guidance on how much or what kind of data to collect. Over-collecting data incurs unnecessary present costs, while under-collecting may incur future costs and delay workflows. We propose a new paradigm to model the data collection workflow as a formal optimal data collection problem that allows designers to specify performance targets, collection costs, a time horizon, and penalties for failing to meet the targets. This formulation generalizes to tasks with multiple data sources, such as labeled and unlabeled data used in semi-supervised learning, and can be easily modified to customized analyses such as how to introduce data from new classes to an existing model. To solve our problem, we develop Learn-Optimize-Collect (LOC), which minimizes expected future collection costs. Finally, we numerically compare our framework to the conventional baseline of estimating data requirements by extrapolating from neural scaling laws. We significantly reduce the risks of failing to meet desired performance targets on several classification, segmentation, and detection tasks, while maintaining low total collection costs.

Despite recent advances demonstrating vision- language models’ (VLMs) abilities to describe complex relationships among objects in images using natural language, their capability to quantitatively reason about object sizes and distances remains underexplored. In this work, we introduce a manually annotated benchmark of 241 questions across five categories specifically designed for quantitative spatial reasoning, and systematically investigate the performance of SoTA VLMs on this task. Our analysis reveals that questions involving reasoning about distances between objects are particularly challenging for SoTA VLMs; however, some VLMs perform significantly better at this task than others, with an almost 40 points gap between the two best performing models. We also make the surprising observation that the success rate of the top-performing VLM increases by 19 points when a reasoning path using a reference object emerges naturally in the response. Inspired by this observation, we develop a zero-shot prompting technique, SpatialPrompt, that encourages VLMs to answer quantitative spatial questions using references objects as visual cues. Specifically, we demonstrate that instruct- ing VLMs to use reference objects in their reasoning paths significantly improves their quantitative spatial reasoning performance, bypassing the need for external data, architectural modifications, or fine-tuning. Remarkably, by solely using SpatialPrompt, Gemini 1.5 Pro, GPT-4V, and GPT-4o improve by 56.2, 28.5, and 6.7 points on average in Q-Spatial Bench without the need for more data, model architectural modifications, or fine-tuning.

In object detection, varying annotation protocols across datasets can result in annotation mismatches, leading to inconsistent class labels and bounding regions. Addressing these mismatches typically involves manually identifying common trends and fixing the corresponding bounding boxes and class labels. To alleviate this laborious process, we introduce the label transfer problem in object detection. Here, the goal is to transfer bounding boxes from one or more source datasets to match the annotation style of a target dataset. We propose a data-centric approach, Label-Guided Pseudo-Labeling (LGPL), that improves downstream detectors in a manner agnostic to the detector learning algorithms and model architectures. Validating across four object detection scenarios, defined over seven different datasets and three different architectures, we show that transferring labels for a target task via LGPL consistently improves the downstream detection in every setting, on average by $1.88$ mAP and 2.65 AP$^{75}$. Most importantly, we find that when training with multiple labeled datasets, carefully addressing annotation mismatches with LGPL alone can improve downstream object detection better than off-the-shelf supervised domain adaptation techniques that align instance features.

Compared to classical machine learning (ML) models, generative models offer a new usage paradigm where (i) a single model can be used for many different tasks out-of-the-box; (ii) users interact with this model over a series of natural language prompts; and (iii) the model is ideally evaluated on binary user satisfaction with respect to model outputs. Given these characteristics, we explore the problem of how developers of new generative AI software can release and price their technology. We first develop a comparison of two different models for a specific task with respect to user cost-effectiveness. We then model the pricing problem of generative AI software as a game between two different companies who sequentially release their models before users choose their preferred model for each task. Here, the price optimization problem becomes piecewise continuous where the companies must choose a subset of the tasks on which to be cost-effective and forgo revenue for the remaining tasks. In particular, we reveal the value of market information by showing that a company who deploys later after knowing their competitor’s price can always secure cost-effectiveness on at least one task, whereas the company who is the first-to-market must price their model in a way that incentivizes higher prices from the latecomer in order to gain revenue. Most importantly, we find that if the different tasks are sufficiently similar, the first-to-market model may become cost-ineffective on all tasks regardless of how this technology is priced.

Given a small training data set and a learning algorithm, how much more data is necessary to reach a target validation or test performance? This question is of critical importance in applications such as autonomous driving or medical imaging where collecting data is expensive and time-consuming. Overestimating or underestimating data requirements incurs substantial costs that could be avoided with an adequate budget. Prior work on neural scaling laws suggest that the power-law function can fit the validation performance curve and extrapolate it to larger data set sizes. We find that this does not immediately translate to the more difficult downstream task of estimating the required data set size to meet a target performance. In this work, we consider a broad class of computer vision tasks and systematically investigate a family of functions that generalize the power-law function to allow for better estimation of data requirements. Finally, we show that incorporating a tuned correction factor and collecting over multiple rounds significantly improves the performance of the data estimators. Using our guidelines, practitioners can accurately estimate data requirements of machine learning systems to gain savings in both development time and data acquisition costs.

Modern deep learning systems require huge data sets to achieve impressive performance, but there is little guidance on how much or what kind of data to collect. Over-collecting data incurs unnecessary present costs, while under-collecting may incur future costs and delay workflows. We propose a new paradigm for modeling the data collection workflow as a formal optimal data collection problem that allows designers to specify performance targets, collection costs, a time horizon, and penalties for failing to meet the targets. Additionally, this formulation generalizes to tasks requiring multiple data sources, such as labeled and unlabeled data used in semi-supervised learning. To solve our problem, we develop Learn-Optimize-Collect (LOC), which minimizes expected future collection costs. Finally, we numerically compare our framework to the conventional baseline of estimating data requirements by extrapolating from neural scaling laws. We significantly reduce the risks of failing to meet desired performance targets on several classification, segmentation, and detection tasks, while maintaining low total collection costs.

The burst rank loss network is an extension of the burst erasure channel, where the channel matrix between the sender and receiver becomes rank-deficient for a certain period of consecutive time-slots. We study streaming communication over the burst rank loss channel, propose a new class of codes, ROBIN codes, and establish their optimality. Our construction uses the maximum rank distance and maximum sum rank codes from previous works as constituent codes and combines them in a layered fashion. Our results generalize previous works on both the single-link and multiple-parallel-link streaming setups over burst erasure channels. We perform simulations over statistical network models to show that ROBIN codes attain low packet loss rates in comparison with the existing codes.

In the classic burst erasure channel, packets are consecutively erased in bursts between gaps of perfect communication. For the Burst Rank Loss Network, instead of a single-link erasure channel, there is a channel matrix that is rank-deficient for a burst of time before returning to full-rank. We establish the streaming capacity of the Burst Rank Loss Network and construct a new family of layered codes referred to as Recovery Of Bursts In Networks (ROBIN) codes that achieve capacity. Our results generalize previous work on both the single-link and multiple-parallel-link streaming setups. Simulations over statistical network models show that ROBIN codes lose fewer packets than baseline codes.

The column Hamming distance of a convolutional code determines the error correction capability when streaming over a class of packet erasure channels. We introduce a metric known as the column sum rank that parallels the column Hamming distance when streaming over a network with link failures. We prove the rank analogues of several known column Hamming distance properties and introduce a new family of convolutional codes that maximize the column sum rank up to the code memory. Our construction involves finding a class of super-regular matrices that preserve this property after multiplication with non-singular block diagonal matrices in the ground field.

We study low-delay streaming codes for erasure channels in point-to-point and multicast scenarios. We consider a sliding window erasure channel which captures the temporal correlation in packet losses observed in real channels. This correlation is often modelled using statistical channels such as Gilbert-Elliott channel. In the point-to-point case, we provide a new class of codes, Embedded Maximum Distance Separable (EMDS) codes, which recovers from channels introducing a mixture of burst and isolated erasures. Moreover, we propose a technique that extends point-to-point codes for the multicast scenario with two receivers that tolerate different delays, T1and T2. The multicast codes opportunistically decode packets with short delay T1when the channel is relatively better and with long delay T2when the channel is worse. Simulations over multicast Gilbert-Elliott channels show that EMDS codes outperform other streaming codes for both users.

The column Hamming distance of a convolutional code determines the error correction capability when streaming over a class of packet erasure channels. We show that the column sum rank parallels column Hamming distance when streaming over a network with link failures. We prove rank analogues of several column distance properties and introduce a new family of convolutional codes that maximize the column sum rank up to the code memory. Our construction involves finding a class of super-regular matrices that preserve this property after multiplication with non-singular block diagonal matrices in the ground field.