Proposed C2-CRS employs coarse-to-fine contrastive learning to fuse multi-type external data for improved conversational recommendation.

Yuanhang Zhou, Kun Zhou, Wayne Xin Zhao et al.

Proposes AIS framework with two-stage human annotation to evaluate if NLG outputs are source-supported, applicable across multiple tasks.

Hannah Rashkin, Vitaly Nikolaev, Matthew Lamm et al.

Introduces QuALITY, a long-text (≈5000 words) multiple-choice QA dataset, with models achieving only 55.4% accuracy versus 93.5% by humans.

Richard Yuanzhe Pang, Alicia Parrish, Nitish Joshi et al.

Gopher (280B parameters) uses Transformer architecture, trained on MassiveText, outperforming SOTA on 152 tasks, especially in knowledge and comprehension.

Jack W. Rae, Sebastian Borgeaud, Trevor Cai et al.

RETRO integrates large-scale retrieval with language models, achieving GPT-3 level performance with 25x fewer parameters by retrieving from a 2 trillion token database.

Sebastian Borgeaud, Arthur Mensch, Jordan Hoffmann et al.

This paper compares prompting, imitation learning, and preference modeling across model scales, highlighting preference ranking's advantages and sample efficiency improvements via pretraining.

Amanda Askell, Yuntao Bai, Anna Chen et al.

XLS-R model leverages wav2vec 2.0 for cross-lingual speech representation learning using 128 languages.

Arun Babu, Changhan Wang, Andros Tjandra et al.

LayoutLM model using Transformer architecture enhances Document AI task accuracy.

Lei Cui, Yiheng Xu, Tengchao Lv et al.

Proposes EN and PALM algorithms for on-the-fly co-occurrence rectification, enabling scalable large-vocabulary topic inference with high robustness.

Moontae Lee, Sungjun Cho, Kun Dong et al.

Introduces BBQ, a handcrafted bias benchmark for question answering, revealing models' reliance on stereotypes especially under low-information conditions, with bias scores up to 100%.

Alicia Parrish, Angelica Chen, Nikita Nangia et al.

P-Tuning v2 employs deep layered prompts, matching fine-tuning performance with only 0.1%-3% trainable parameters across various model scales and tasks.

Xiao Liu, Kaixuan Ji, Yicheng Fu et al.

Proposes RemBERT, a distilled multilingual evaluation metric reaching 92.6% of the large model's performance with only one-third parameters.

Amy Pu, Hyung Won Chung, Ankur P. Parikh et al.

Proposes CKL framework with INVARIANTLAMA, UPDATEDLAMA, NEWLAMA datasets and FUAR metric; uses parameter expansion methods to improve knowledge retention and acquisition.

Joel Jang, Seonghyeon Ye, Sohee Yang et al.

Proposes a controllable neural dialogue summarization framework with personal entity planning, improving factual consistency and diversity; achieves ROUGE-2 of 58.7 on SAMSum.

Zhengyuan Liu, Nancy F. Chen

Proposes NTRD framework using neural templates and slot filling for recommender dialogue, outperforming SOTA with 1.806% ReR@1 and zero-shot capabilities.

Zujie Liang, Huang Hu, Can Xu et al.

RnG-KBQA combines contrastive ranking and generation, greatly improving zero-shot and generalization in KBQA.

Xi Ye, Semih Yavuz, Kazuma Hashimoto et al.

This study tracks neural language models' (NLMs) learning trajectories, revealing highly consistent stages across architectures and data, driven by an underlying inductive bias.

Leshem Choshen, Guy Hacohen, Daphna Weinshall et al.

This study analyzes Transformer attention weights in NMT, revealing biases towards uninformative tokens but demonstrating interpretability and improvement methods.

Javier Ferrando, Marta R. Costa-jussà

PICARD constrains autoregressive decoding via incremental parsing, significantly improving SQL accuracy with 75.5% exact match on Spider.

Torsten Scholak, Nathan Schucher, Dzmitry Bahdanau

Proposed a framework to address knowledge conflicts in QA, improving OOD generalization by 4%-7%.

Shayne Longpre, Kartik Perisetla, Anthony Chen et al.