1 code implementation • 14 Dec 2024 • Jihwan Oh, Jeonghwan Choi, Nicole Hee-Yeon Kim, Taewon Yun, Hwanjun Song
In this paper, we explore alternative way of leveraging Large Language Model (LLM) generated feedback to address the inherent limitation of using human-labeled data.
no code implementations • 29 Nov 2024 • Junyong Kang, Donghyun Son, Hwanjun Song, Buru Chang
In-context learning refers to the emerging ability of large language models (LLMs) to perform a target task without additional training, utilizing demonstrations of the task.
no code implementations • 17 Oct 2024 • Hwanjun Song, Taewon Yun, Yuho Lee, Jihwan Oh, Gihun Lee, Jason Cai, Hang Su
Developing effective text summarizers remains a challenge due to issues like hallucinations, key information omissions, and verbosity in LLM-generated summaries.
1 code implementation • 30 Sep 2024 • Yuho Lee, Taewon Yun, Jason Cai, Hang Su, Hwanjun Song
Existing benchmarks for summarization quality evaluation often lack diverse input scenarios, focus on narrowly defined dimensions (e. g., faithfulness), and struggle with subjective and coarse-grained annotation schemes.
1 code implementation • 16 Jul 2024 • Shunqi Mao, Chaoyi Zhang, Hang Su, Hwanjun Song, Igor Shalyminov, Weidong Cai
Contextualized Image Captioning (CIC) evolves traditional image captioning into a more complex domain, necessitating the ability for multimodal reasoning.
2 code implementations • 1 Jul 2024 • Hwanjun Song, Hang Su, Igor Shalyminov, Jason Cai, Saab Mansour
Automated evaluation is crucial for streamlining text summarization benchmarking and model development, given the costly and time-consuming nature of human evaluation.
no code implementations • 7 Mar 2024 • Yuwei Zhang, Siffi Singh, Sailik Sengupta, Igor Shalyminov, Hang Su, Hwanjun Song, Saab Mansour
The triplet task gauges the model's understanding of two semantic concepts paramount in real-world conversational systems-- negation and implicature.
1 code implementation • 6 Mar 2024 • Jianfeng He, Hang Su, Jason Cai, Igor Shalyminov, Hwanjun Song, Saab Mansour
Semi-supervised dialogue summarization (SSDS) leverages model-generated summaries to reduce reliance on human-labeled data and improve the performance of summarization models.
Abstractive Text Summarization
Natural Language Understanding
1 code implementation • 5 Mar 2024 • Hossein Aboutalebi, Hwanjun Song, Yusheng Xie, Arshit Gupta, Justin Sun, Hang Su, Igor Shalyminov, Nikolaos Pappas, Siffi Singh, Saab Mansour
Development of multimodal interactive systems is hindered by the lack of rich, multimodal (text, images) conversational data, which is needed in large quantities for LLMs.
1 code implementation • 20 Feb 2024 • Liyan Tang, Igor Shalyminov, Amy Wing-mei Wong, Jon Burnsky, Jake W. Vincent, Yu'an Yang, Siffi Singh, Song Feng, Hwanjun Song, Hang Su, Lijia Sun, Yi Zhang, Saab Mansour, Kathleen McKeown
We find that there are diverse errors and error distributions in model-generated summaries and that non-LLM based metrics can capture all error types better than LLM-based evaluators.
no code implementations • 14 Dec 2023 • Doyoung Kim, Dongmin Park, Yooju Shin, Jihwan Bang, Hwanjun Song, Jae-Gil Lee
We propose a novel framework DropTop that suppresses the shortcut bias in online continual learning (OCL) while being adaptive to the varying degree of the shortcut bias incurred by continuously changing environment.
1 code implementation • 12 Dec 2023 • Hwanjun Song, Minseok Kim, Jae-Gil Lee
Multi-label classification poses challenges due to imbalanced and noisy labels in training data.
1 code implementation • 18 Nov 2023 • Doyoung Kim, Susik Yoon, Dongmin Park, YoungJun Lee, Hwanjun Song, Jihwan Bang, Jae-Gil Lee
In real-world continual learning (CL) scenarios, tasks often exhibit intricate and unpredictable semantic shifts, posing challenges for fixed prompt management strategies which are tailored to only handle semantic shifts of uniform degree (i. e., uniformly mild or uniformly abrupt).
no code implementations • 20 Oct 2023 • Hwanjun Song, Igor Shalyminov, Hang Su, Siffi Singh, Kaisheng Yao, Saab Mansour
Our experiments show that DisCal outperforms prior methods in abstractive summarization distillation, producing highly abstractive and informative summaries.
2 code implementations • 9 Oct 2023 • Sangmin Bae, Jongwoo Ko, Hwanjun Song, Se-Young Yun
To tackle the high inference latency exhibited by autoregressive language models, previous studies have proposed an early-exiting framework that allocates adaptive computation paths for each token based on the complexity of generating the subsequent token.
no code implementations • 25 Mar 2023 • Hwanjun Song, Jihwan Bang
Prompt-OVD is an efficient and effective framework for open-vocabulary object detection that utilizes class embeddings from CLIP as prompts, guiding the Transformer decoder to detect objects in both base and novel classes.
1 code implementation • CVPR 2023 • Sangmook Kim, Sangmin Bae, Hwanjun Song, Se-Young Yun
In this work, we first demonstrate that the superiority of two selector models depends on the global and local inter-class diversity.
1 code implementation • 22 Mar 2023 • Jemin Lee, Yongin Kwon, Sihyeong Park, Misun Yu, Jeman Park, Hwanjun Song
For mobile devices, achieving optimal acceleration for ViTs necessitates the strategic integration of quantization techniques and efficient hybrid transformer structures.
1 code implementation • ICCV 2023 • Dahuin Jung, Dongyoon Han, Jihwan Bang, Hwanjun Song
However, we observe that the use of a prompt pool creates a domain scalability problem between pre-training and continual learning.
2 code implementations • 13 Oct 2022 • Dongmin Park, Yooju Shin, Jihwan Bang, YoungJun Lee, Hwanjun Song, Jae-Gil Lee
Unlabeled data examples awaiting annotations contain open-set noise inevitably.
1 code implementation • 19 Jul 2022 • Sukmin Yun, Jaehyung Kim, Dongyoon Han, Hwanjun Song, Jung-Woo Ha, Jinwoo Shin
Understanding temporal dynamics of video is an essential aspect of learning better video representations.
no code implementations • 1 Jul 2022 • Wonyoung Shin, Jonghun Park, Taekang Woo, Yongwoo Cho, Kwangjin Oh, Hwanjun Song
Understanding vision and language representations of product content is vital for search and recommendation applications in e-commerce.
2 code implementations • 30 May 2022 • Jang-Hyun Kim, Jinuk Kim, Seong Joon Oh, Sangdoo Yun, Hwanjun Song, JoonHyun Jeong, Jung-Woo Ha, Hyun Oh Song
The great success of machine learning with massive amounts of data comes at a price of huge computation costs and storage for training and tuning.
no code implementations • 11 May 2022 • Jaehoon Oh, Sungnyun Kim, Namgyu Ho, Jin-Hwa Kim, Hwanjun Song, Se-Young Yun
Cross-domain few-shot learning (CD-FSL), where there are few target samples under extreme differences between source and target domains, has recently attracted huge attention.
1 code implementation • 3 May 2022 • Sangmook Kim, Wonyoung Shin, Soohyuk Jang, Hwanjun Song, Se-Young Yun
Robustness is becoming another important challenge of federated learning in that the data collection process in each client is naturally accompanied by noisy labels.
1 code implementation • 17 Apr 2022 • Hwanjun Song, Deqing Sun, Sanghyuk Chun, Varun Jampani, Dongyoon Han, Byeongho Heo, Wonjae Kim, Ming-Hsuan Yang
Transformers have been widely used in numerous vision problems especially for visual recognition and detection.
3 code implementations • CVPR 2022 • Jihwan Bang, Hyunseo Koh, Seulki Park, Hwanjun Song, Jung-Woo Ha, Jonghyun Choi
A large body of continual learning (CL) methods, however, assumes data streams with clean labels, and online learning scenarios under noisy data streams are yet underexplored.
1 code implementation • 19 Mar 2022 • Minseok Kim, Hwanjun Song, Yooju Shin, Dongmin Park, Kijung Shin, Jae-Gil Lee
It is featured with an adaptive learning rate for each parameter-interaction pair for inducing a recommender to quickly learn users' up-to-date interest.
2 code implementations • 1 Feb 2022 • Jaehoon Oh, Sungnyun Kim, Namgyu Ho, Jin-Hwa Kim, Hwanjun Song, Se-Young Yun
This data enables self-supervised pre-training on the target domain, in addition to supervised pre-training on the source domain.
no code implementations • 13 Dec 2021 • Steven Euijong Whang, Yuji Roh, Hwanjun Song, Jae-Gil Lee
In this survey, we study the research landscape for data collection and data quality primarily for deep learning applications.
1 code implementation • NeurIPS 2021 • Dongmin Park, Hwanjun Song, Minseok Kim, Jae-Gil Lee
A deep neural network (DNN) has achieved great success in many machine learning tasks by virtue of its high expressive power.
1 code implementation • ICLR 2022 • Hwanjun Song, Deqing Sun, Sanghyuk Chun, Varun Jampani, Dongyoon Han, Byeongho Heo, Wonjae Kim, Ming-Hsuan Yang
Transformers are transforming the landscape of computer vision, especially for recognition tasks.
Ranked #16 on
Object Detection
on COCO 2017 val
no code implementations • ICLR 2022 • Yooju Shin, Susik Yoon, Sundong Kim, Hwanjun Song, Jae-Gil Lee, Byung Suk Lee
Time-series data are ubiquitous these days, but lack of the labels in time-series data is regarded as a hurdle for its broad applicability.
no code implementations • 29 Sep 2021 • Jihwan Bang, Hyunseo Koh, Seulki Park, Hwanjun Song, Jung-Woo Ha, Jonghyun Choi
Specifically, we argue the importance of both diversity and purity of examples in the episodic memory of continual learning models.
no code implementations • 14 Jun 2021 • Seulki Park, Hwanjun Song, Daeho Um, Dae Ung Jo, Sangdoo Yun, Jin Young Choi
Deep neural network can easily overfit to even noisy labels due to its high capacity, which degrades the generalization performance of a model.
no code implementations • 8 Dec 2020 • Hwanjun Song, Minseok Kim, Dongmin Park, Yooju Shin, Jae-Gil Lee
In the seeding phase, the network is updated using all the samples to collect a seed of clean samples.
1 code implementation • 16 Jul 2020 • Hwanjun Song, Minseok Kim, Dongmin Park, Yooju Shin, Jae-Gil Lee
Deep learning has achieved remarkable success in numerous domains with help from large amounts of big data.
no code implementations • 19 Nov 2019 • Hwanjun Song, Minseok Kim, Sundong Kim, Jae-Gil Lee
Compared with existing batch selection methods, the results showed that Recency Bias reduced the test error by up to 20. 97% in a fixed wall-clock training time.
no code implementations • 19 Nov 2019 • Hwanjun Song, Minseok Kim, Dongmin Park, Jae-Gil Lee
In this paper, we claim that such overfitting can be avoided by "early stopping" training a deep neural network before the noisy labels are severely memorized.
no code implementations • 23 Oct 2019 • Dongmin Park, Susik Yoon, Hwanjun Song, Jae-Gil Lee
Learning a good distance measure for distance-based classification in time series leads to significant performance improvement in many tasks.
no code implementations • 25 Sep 2019 • Hwanjun Song, Minseok Kim, Dongmin Park, Jae-Gil Lee
In this paper, we claim that such overfitting can be avoided by "early stopping" training a deep neural network before the noisy labels are severely memorized.
1 code implementation • 15 Jun 2019 • Hwanjun Song, Minseok Kim, Jae-Gil Lee
Owing to the extremely high expressive power of deep neural networks, their side effect is to totally memorize training data even when the labels are extremely noisy.
Ranked #15 on
Learning with noisy labels
on ANIMAL
no code implementations • ICLR 2019 • Hwanjun Song, Sundong Kim, Minseok Kim, Jae-Gil Lee
Neural networks can converge faster with help from a smarter batch selection strategy.