Papers by Qinyuan Ye
LLM-driven Instruction Following: Progresses and Concerns (2023.emnlp-tutorial)
Copied to clipboard
| Challenge: | a tutorial on task instruction is aimed at researchers and practitioners interested in NLP generalization . labeled examples are unlikely to be available in large numbers or do not exist . |
| Approach: | This tutorial will examine the progress of natural language processing (NLP) using labeled examples. authors propose that task instructions act as a novel resource for supervision. |
| Outcome: | This tutorial aims to answer questions about instruction-driven NLP . it focuses on the use of task instructions in a low-shot scenario . |
LEAN-LIFE: A Label-Efficient Annotation Framework Towards Learning from Explanation (2020.acl-demos)
Copied to clipboard
Dong-Ho Lee, Rahul Khanna, Bill Yuchen Lin, Seyeon Lee, Qinyuan Ye, Elizabeth Boschee, Leonardo Neves, Xiang Ren
| Challenge: | Existing frameworks for sequence labeling and classification require massive human effort and labeling data is limited. |
| Approach: | They propose a web-based, Label-Efficient AnnotatioN framework that allows an annotator to provide the needed labels for a task and can capture explanations for each labeling decision. |
| Outcome: | The proposed framework surpasses baseline F1 scores by 5-10 percentage points while using 2X times fewer labeled instances. |
Estimating Large Language Model Capabilities without Labeled Test Data (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Large Language Models have shown impressive ability to perform in-context learning from only a few examples, but their accuracy varies widely from task to task. |
| Approach: | They propose a method that trains a meta-model using LLM confidence scores as features to perform ICL accuracy estimation. |
| Outcome: | The proposed method improves over baselines across 7 out of 12 settings and achieves the same accuracy as evaluating on 40 sampled examples per task. |
FiD-ICL: A Fusion-in-Decoder Approach for Efficient In-Context Learning (2023.acl-long)
Copied to clipboard
| Challenge: | Large pre-trained models are capable of few-shot in-context learning (ICL) however, concatenated demonstrations are often excessively long and require additional computation. |
| Approach: | They propose to apply fusion-in-decoder (FiD) models to perform few-shot in-context learning (ICL) they propose to use concatenation-based, early-fusion, intermediate- and late-fusion methods to improve efficiency . |
| Outcome: | The proposed methods outperform concatenation-based models on 11 held-out tasks. |
Sparse Distillation: Speeding Up Text Classification by Using Bigger Student Models (2022.naacl-main)
Copied to clipboard
| Challenge: | Existing methods to reduce inference cost by distilling transformer models into lightweight student models are limited for high-volume use cases. |
| Approach: | They propose to distill state-of-the-art transformer models into lightweight student models to reduce computation cost at inference time. |
| Outcome: | The proposed pipeline achieves up to 600x speed-up on GPUs and CPUs on six single-sentence text classification tasks and in domain generalization settings. |
Teaching Machine Comprehension with Compositional Explanations (2020.findings-emnlp)
Copied to clipboard
| Challenge: | Recent advances in machine reading comprehension rely heavily on large-scale annotated corpora, which are timeconsuming and costly to collect. |
| Approach: | They propose to use semi-structured explanations to “teach” machines reading comprehension using a small number of semi-structural explanations that explicitly inform machines why answer spans are correct. |
| Outcome: | The proposed method achieves 70.14% F1 score with supervision from 26 explanations on the SQuAD dataset, comparable to plain supervised learning using 1,100 labeled instances yielding a 12x speed up. |
Safe Inputs but Unsafe Output: Benchmarking Cross-modality Safety Alignment of Large Vision-Language Models (2025.findings-naacl)
Copied to clipboard
Siyin Wang, Xingsong Ye, Qinyuan Cheng, Junwen Duan, Shimin Li, Jinlan Fu, Xipeng Qiu, Xuanjing Huang
| Challenge: | Recent studies focus on single-modality threats, but this approach fails to address cross-modal safety alignment. |
| Approach: | They propose a safety alignment challenge to evaluate cross-modality safety alignment . they propose 'Safe Inputs but Unsafe Output' to consider safety of single modalities . |
| Outcome: | The proposed safety alignment challenge examines cases where modalities are safe independently but could lead to unsafe outputs when combined. |
Learning to Generate Task-Specific Adapters from Task Description (2021.acl-short)
Copied to clipboard
| Challenge: | Pre-trained text-to-text transformers have achieved impressive performance across a range of NLP tasks, such as question answering and commonsense reasoning. |
| Approach: | They propose a framework that improves text-to-text transformer’s generalization ability to unseen tasks by training a hypernetwork to generate task-specific adapters from task descriptions. |
| Outcome: | Experiments on ZEST and a synthetic SQuAD dataset show that Hypter improves upon fine-tuning baselines. |
Looking Beyond Label Noise: Shifted Label Distribution Matters in Distantly Supervised Relation Extraction (D19-1)
Copied to clipboard
| Challenge: | Existing studies on DS-based relation extraction (RE) methods focus on handling label noise, but other factors may have been overlooked. |
| Approach: | They propose a method to automatically adjust DS-RE models to a shifted label distribution problem . they find this problem exists in real-world DS datasets and can be overcome . |
| Outcome: | The proposed method achieves consistent performance gains on DS-trained models with an up to 23% relative F1 improvement, which verifies their assumptions. |
CrossFit: A Few-shot Learning Challenge for Cross-task Generalization in NLP (2021.emnlp-main)
Copied to clipboard
| Challenge: | We study whether and how cross-task generalization ability can be acquired . we use CrossFit to standardize seen/unseen task partitions and evaluation protocols . |
| Approach: | They propose a problem setup for studying cross-task generalization ability which standardizes seen/unseen task partitions and data access during different learning stages. |
| Outcome: | The proposed model can be used to build few-shot learners across diverse tasks. |
LLM can Achieve Self-Regulation via Hyperparameter Aware Generation (2024.findings-acl)
Copied to clipboard
Siyin Wang, Shimin Li, Tianxiang Sun, Jinlan Fu, Qinyuan Cheng, Jiasheng Ye, Junjie Ye, Xipeng Qiu, Xuanjing Huang
| Challenge: | Existing decoding strategies and hyperparameters may not be optimal for each sample. |
| Approach: | They propose a model that auto-regulates decoding strategies and hyperparameters . this approach eliminates the need for extensive manual tuning, they argue . |
| Outcome: | The proposed model eliminates the need for extensive manual tuning, offering a more autonomous, self-regulate model behavior. |
On the Influence of Masking Policies in Intermediate Pre-training (2021.emnlp-main)
Copied to clipboard
Qinyuan Ye, Belinda Z. Li, Sinong Wang, Benjamin Bolte, Hao Ma, Wen-tau Yih, Xiang Ren, Madian Khabsa
| Challenge: | Existing studies show that inserting an intermediate pre-training stage improves performance of masked language models. |
| Approach: | They propose methods to automate the discovery of optimal masking policies via direct supervision or meta-learning. |
| Outcome: | The proposed method outperforms the heuristic of masking named entities on TriviaQA and can be generalizable beyond that task. |
Cross-Task Generalization Abilities of Large Language Models (2024.naacl-srw)
Copied to clipboard
| Challenge: | a thesis proposal advocates for the crucial role of cross-task generalization in NLP systems. |
| Approach: | They propose to benchmark cross-task generalization abilities with diverse NLP tasks . they also propose to develop model architectures for improving cross- task generalization . |
| Outcome: | This paper compares cross-task generalization abilities with diverse NLP tasks . it also analyzes and predicts the generalization landscape of current state-of-the-art large language models . |
How Predictable Are Large Language Model Capabilities? A Case Study on BIG-bench (2023.findings-emnlp)
Copied to clipboard
| Challenge: | a recent study shows that large language models can be used to predict performance on new configurations. |
| Approach: | They investigate the predictability of large language model capabilities by using BIG-bench . they find a subset of BIG-Bench tasks as informative as BIG-bnch Hard . |
| Outcome: | The proposed model achieves an R2 score greater than 95% on BIG-bench . the model is 3 smaller than BIG-Bench Hard, and the model performs better on the full set. |
Prompt Engineering a Prompt Engineer (2024.findings-acl)
Copied to clipboard
| Challenge: | Recent studies indicate that large language models can be meta-prompted to perform automatic prompt engineering, but their potential is limited due to insufficient guidance for complex reasoning in the meta-prompt. |
| Approach: | They propose to infuse three key components into a meta-prompt to guide reasoning . they find prompts that outperform “let’s think step by step” by 6.3% on MultiArith and 3.1% on GSM8K . |
| Outcome: | The proposed method outperforms “let’s think step by step” by 6.3% on MultiArith and 3.1% on GSM8K and outperfies baselines on counterfactual tasks by 6.9%. |
Eliciting and Understanding Cross-task Skills with Task-level Mixture-of-Experts (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Pre-trained transformer models are capable of multitasking on diverse NLP tasks, but little is known about how multitaskability and cross-task generalization is achieved. |
| Approach: | They propose to use a transformer-based mixture-of-expert model with a router component to choose among experts dynamically and flexibly. |
| Outcome: | The proposed models improve the average performance gain (ARG) metric by 2.6% when adapting to unseen tasks, and by 5.6% in zero-shot generalization settings. |