Papers by Yezhou Yang
Video2Commonsense: Generating Commonsense Descriptions to Enrich Video Captioning (2020.emnlp-main)
Copied to clipboard
| Challenge: | Observable changes in the scene are reflected in captions, but actions are also linked to social aspects such as intentions, effects, and attributes that describe the agent. |
| Approach: | They propose to generate captions from videos that describe latent aspects of the human agent's actions. |
| Outcome: | The proposed model can be used to describe latent aspects of human actions in video clips and answer questions about videos. |
MUTANT: A Training Paradigm for Out-of-Distribution Generalization in Visual Question Answering (2020.emnlp-main)
Copied to clipboard
| Challenge: | Availability of large-scale datasets has enabled statistical machine learning in vision and language understanding. |
| Approach: | They propose a training paradigm that exposes models to perceptually similar mutations of input . they show a 10.57% improvement in the VQA-CP challenge . |
| Outcome: | The proposed training paradigm improves on the visual question answering challenge with 10.57% accuracy. |
Learning Action-Effect Dynamics for Hypothetical Vision-Language Reasoning Task (2022.findings-emnlp)
Copied to clipboard
| Challenge: | ‘Actions’ play a vital role in how humans interact with the world and therefore require the capability to perform ‘Reasoning about Actions & Change’ (RAC). |
| Approach: | They propose an encoder-decoder architecture to learn the representation of actions as vectors and combine it with existing modality parsers and a scene graph question answering model to evaluate the proposed system. |
| Outcome: | The proposed system performs better than baselines on the CLEVR_HYP dataset and compares with existing parsers and a scene graph question answering model. |
End-to-end Knowledge Retrieval with Multi-modal Queries (2023.acl-long)
Copied to clipboard
| Challenge: | a new task is proposed to learn knowledge retrieval with multimodal queries . a vision-language model can retrieve knowledge using images and text inputs . |
| Approach: | They propose a task for vision-language models to retrieve knowledge with multi-modal queries . they propose reViz, a model that integrates content from both text and image queries based on a multimodal query task . |
| Outcome: | The proposed task performs better under zero-shot settings than previous work on cross-modal retrieval. |
Lost in Translation? Translation Errors and Challenges for Fair Assessment of Text-to-Image Models on Multilingual Concepts (2024.naacl-short)
Copied to clipboard
| Challenge: | With growth in the popularity of text-to-image models has come interest in assessing their multilingual capabilities, including multilingual accessibility. |
| Approach: | They propose to correct translation errors in a concept list translated to seven languages and compare the outputs of the benchmark to those conditioned on the old. |
| Outcome: | The proposed benchmark contains translation errors in Spanish, Japanese, and Chinese. |
Semantically Distributed Robust Optimization for Vision-and-Language Inference (2022.findings-acl)
Copied to clipboard
| Challenge: | Existing methods to integrate linguistic knowledge into training pipelines are under-explored. |
| Approach: | They propose a model-agnostic method that leverages linguistic transformations to infer a set of linguistic phenomena. |
| Outcome: | The proposed method improves on benchmark datasets with images and video and is generalizable to other V&L tasks. |
Visuo-Linguistic Question Answering (VLQA) Challenge (2020.findings-emnlp)
Copied to clipboard
| Challenge: | Understanding images and text together is an important aspect of cognition and building advanced AI systems. |
| Approach: | They propose to derive joint inference about a given image-text modality and compile a question-answering corpus using an image and a reading passage. |
| Outcome: | The proposed method has better baseline performance but is still far behind human performance. |
SMURF: SeMantic and linguistic UndeRstanding Fusion for Caption Evaluation via Typicality Analysis (2021.acl-long)
Copied to clipboard
| Challenge: | Visual captioning is an open-ended area for evaluation, requiring specialized training to improve human-correlation. |
| Approach: | They propose a new evaluation framework rooted in information theory . they propose metric SPURTS and metric SMURF to measure fluency . |
| Outcome: | The proposed metrics achieve state-of-the-art correlation with human judgment compared with other evaluation metrics. |
TROPE: TRaining-Free Object-Part Enhancement for Seamlessly Improving Fine-Grained Zero-Shot Image Captioning (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing approaches to enhance zero-shot abilities in image captioning fail with fine-grained datasets. |
| Approach: | They propose a method to enhance captions with additional object-part details using object detector proposals and natural language processing techniques. |
| Outcome: | The proposed method improves performance on fine-grained datasets and improves on existing methods. |
Precision or Recall? An Analysis of Image Captions for Training Text-to-Image Generation Model (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Recent advances in text-to-image models have demonstrated remarkable capabilities in image synthesis. |
| Approach: | They analyze the critical role of caption precision and recall in text-to-image model training. |
| Outcome: | The proposed model trains with synthetic captions that show similar behavior to those trained on human-annotated captions. |
CRIPP-VQA: Counterfactual Reasoning about Implicit Physical Properties via Video Question Answering (2022.emnlp-main)
Copied to clipboard
| Challenge: | Videos often capture objects, their visible properties, their motion, and the interactions between different objects. |
| Approach: | They propose a video question answering dataset for reasoning about the implicit physical properties of objects in a scene. |
| Outcome: | The proposed dataset enables evaluation under several out-of-distribution settings – videos with objects with masses, coefficients of friction, and initial velocities that are not observed in the training distribution. |
AcT2I: Evaluating and Improving Action Depiction in Text-to-Image Models (2025.emnlp-main)
Copied to clipboard
| Challenge: | Text-to-Image (T2I) models have been successful in generating images from textual descriptions, but they struggle to capture nuanced and implicit attributes inherent in action depiction. |
| Approach: | They propose a benchmark to evaluate the performance of T2I models in generating images from action-centric prompts. |
| Outcome: | The proposed model achieves an increase of 72% on AcT2I. |
WeaQA: Weak Supervision via Captions for Visual Question Answering (2021.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for training visual question answering models rely on datasets with human-annotated image-quest-answer triplets. |
| Approach: | They propose a method to train models with synthetic Q-A pairs generated procedurally from captions. |
| Outcome: | The proposed method trains models with synthetic Q-A pairs generated from captions on three VQA benchmarks. |
CLEVR_HYP: A Challenge Dataset and Baselines for Visual Question Answering with Hypothetical Actions over Images (2021.naacl-main)
Copied to clipboard
| Challenge: | Existing research on visual question answering is limited to information explicitly present in an image or a video. |
| Approach: | They propose a vision-language question answering task based on a CLEVR dataset . they modify existing methods and propose baseline solvers for this task . |
| Outcome: | The proposed model motivates the development of better vision-language models . it provides insights about the capability of diverse architectures to perform joint reasoning over image-text modality. |
To Find Waldo You Need Contextual Cues: Debiasing Who’s Waldo (2022.acl-short)
Copied to clipboard
| Challenge: | Existing datasets for Person-centric visual grounding (PCVG) are biased and over-estimated . |
| Approach: | They propose to use a dataset for the Person-centric Visual Grounding task that requires pairing up a person’s name mentioned in a caption with a bounding box that points to the person in the image. |
| Outcome: | The proposed dataset offers a more practical benchmark for future improvements and better performance than the original dataset. |