While recent years have seen rapid progress in image-conditioned text generation, image captioning still suffers from the fundamental issue of hallucinations, namely, the generation of spurious details that cannot be inferred from the given image. Existing methods largely use closed-vocabulary object lists to mitigate or evaluate hallucinations in image captioning, ignoring the long-tailed nature of hallucinations that occur in practice. To this end, we propose a framework for addressing hallucinations in image captioning in the open-vocabulary setting. Our framework includes a new benchmark, OpenCHAIR, that leverages generative foundation models to evaluate open-vocabulary object hallucinations for image captioning, surpassing the popular and similarly-sized CHAIR benchmark in both diversity and accuracy. Furthermore, to mitigate open-vocabulary hallucinations without using a closed object list, we propose MOCHa, an approach harnessing advancements in reinforcement learning. Our multi-objective reward function explicitly targets the trade-off between fidelity and adequacy in generations without requiring any strong supervision. MOCHa improves a large variety of image captioning models, as captured by our OpenCHAIR benchmark and other existing metrics. We will release our code and models.

Cuneiform documents, the earliest known form of writing, are prolific textual sources of the ancient past. Experts publish editions of these texts in transliteration using specialized typesetting, but most remain inaccessible for computational analysis in traditional printed books or legacy materials. Off-the-shelf OCR systems are insufficient for digitization without adaptation. We present CuReD (Cuneiform Recognition-Documents), a deep learning-based human-in-the-loop OCR pipeline for digitizing scanned transliterations of cuneiform texts. CuReD has a character error rate of 9% on clean data and 11% on representative scans. We digitized a challenging sample of transliterated cuneiform documents, as well as lexical index cards from the University of Pennsylvania Museum, demonstrating the feasibility of our platform for enabling computational analysis and bolstering machine-readable cuneiform text datasets. Our result provide the first human-in-the-loop pipeline and interface for digitizing transliterated cuneiform sources and legacy materials, enabling the enrichment of digital sources of these low-resource languages.

Web-scale training on paired text-image data is becoming increasingly central to multimodal learning, but is challenged by the highly noisy nature of datasets in the wild. Standard data filtering approaches succeed in removing mismatched text-image pairs, but permit semantically related but highly abstract or subjective text. These approaches lack the fine-grained ability to isolate the most concrete samples that provide the strongest signal for learning in a noisy dataset. In this work, we propose a new metric, Image Caption Concreteness (ICC), that evaluates caption text without an image reference to measure its concreteness and relevancy for use in multimodal learning. Our unsupervised approach leverages strong foundation models for measuring visual-semantic information loss in multimodal representations. We demonstrate that this strongly correlates with human evaluation of concreteness in both single-word and caption-level texts. Moreover, we show that curation using ICC complements existing approaches: It succeeds in selecting the highest quality samples from multimodal web-scale datasets to allow for efficient training in resource-constrained settings.