Counterfactual Attention Supervision

Choi, Seungtaek; Park, Haeju; Hwang, Seung-won

doi:10.1109/icdm.2019.00115

Cited by 5 publications

(3 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…FRETS (Jauhar et al, 2016) uses a log-linear model conditioned on alignment scores between cells in tables and individual QA pairs in the training set. NEOP (Cho et al, 2018) uses a multi-layer sequential network with attention supervision to answer queries conditioned on tables. MANYMODALQA (Hannan et al, 2020) uses a modality selection network along with pretrained state-of-the-art text-based QA, Table-based QA, and Image-based QA models to jointly answer questions over text, tables, and images.…”

Section: Table Question Answeringmentioning

confidence: 99%

cTBL: Augmenting Large Language Models for Conversational Tables

Sundar¹,

Heck²

2023

Preprint

View full text Add to dashboard Cite

An open challenge in multimodal conversational AI requires augmenting large language models with information from textual and nontextual sources for multi-turn dialogue. To address this problem, this paper introduces Conversational Tables (CTBLS), a three-step encoder-decoder architecture to retrieve tabular information and generate dialogue responses grounded on the retrieved information. CTBLS uses Transformer encoder embeddings for Dense Table Retrieval and obtains up to 5% relative improvement in Top-1 and Top-3 accuracy over sparse retrieval on the HYRBIDIALOGUE dataset. Additionally, CTBLS performs tabular knowledge retrieval using encoder and decoder models, resulting in up to 46% relative improvement in ROUGE scores and better human evaluation for response generation on HYRBIDIALOGUE.

show abstract

Section: Table Question Answeringmentioning

confidence: 99%

cTBL: Augmenting Large Language Models for Conversational Tables

Sundar¹,

Heck²

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Alternatively, machine may mine or augment attention supervision: (Tang et al, 2019) automatically mines attention supervision by masking-out highly attentive words in a progressive manner. (Choi et al, 2019) augments counterfactual observations to debias human attention supervision via instance similarity. Our work is of combining the strength of the two works: we automatically improve attention supervision via self-supervision signals, but we build it with free task-level resources.…”

Section: Attention To/from Machinementioning

confidence: 99%

Less is More: Attention Supervision with Counterfactuals for Text Classification

Choi¹,

Park²,

Yeo³

et al. 2020

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)

Self Cite

View full text Add to dashboard Cite

We aim to leverage human and machine intelligence together for attention supervision. Specifically, we show that human annotation cost can be kept reasonably low, while its quality can be enhanced by machine selfsupervision. Specifically, for this goal, we explore the advantage of counterfactual reasoning, over associative reasoning typically used in attention supervision. Our empirical results show that this machine-augmented human attention supervision is more effective than existing methods requiring a higher annotation cost, in text classification tasks, including sentiment analysis and news categorization.

show abstract

“…However, due to patent applications, these findings were not included in the paper, and only the CFD analysis results were presented at a conference. 20,21…”

Section: Introductionmentioning

confidence: 99%

Evaluating one engine inoperative conditions in subscale electric vertical take-off and landing aircraft: An in-depth tethered hover test analysis

Kang,

Cho,

Choi

et al. 2023

Proceedings of the Institution of Mechanical Engineers, Part G:

View full text Add to dashboard Cite

The direction of propeller rotation in a distributed electric-powered vertical take-off and landing (eVTOL) aircraft significantly influences control forces and induced drag during both helicopter and fixed-wing modes. This study proposes a strategy to determine the most effective rotational direction for each propeller. This approach effectively mitigates one-engine-inoperative (OEI) conditions during helicopter mode for a subscale eVTOL referred to as the optionally piloted personal air vehicle (OPPAV). Moreover, the study developed an optimal control law using the linear programming method, which minimizes the maximum power required for individual motors under OEI conditions. The lowest maximum power was achieved under OEI conditions when both the front and rear propellers in each pod rotated in the same direction. Furthermore, to validate this proposed control law, a tethered hover test was performed using the subscale OPPAV under OEI conditions. Our findings demonstrate that determining the rotational direction of propellers using the newly proposed reconfiguration control method significantly enhances the safety of eVTOL aircraft operating under OEI conditions.

show abstract

Counterfactual Attention Supervision

Cited by 5 publications

References 17 publications

cTBL: Augmenting Large Language Models for Conversational Tables

cTBL: Augmenting Large Language Models for Conversational Tables

Less is More: Attention Supervision with Counterfactuals for Text Classification

Evaluating one engine inoperative conditions in subscale electric vertical take-off and landing aircraft: An in-depth tethered hover test analysis

Contact Info

Product

Resources

About