Search Results for author: Hongyin Luo

Found 31 papers, 19 papers with code

Quantifying Generalization Complexity for Large Language Models

1 code implementation2 Oct 2024 Zhenting Qi, Hongyin Luo, Xuliang Huang, Zhuokai Zhao, Yibo Jiang, Xiangjun Fan, Himabindu Lakkaraju, James Glass

Scylla disentangles generalization from memorization via assessing model performance on both in-distribution (ID) and out-of-distribution (OOD) data through 20 tasks across 5 levels of complexity.

Memorization

Addition is All You Need for Energy-efficient Language Models

no code implementations1 Oct 2024 Hongyin Luo, Wei Sun

The new algorithm costs significantly less computation resource than 8-bit floating point multiplication but achieves higher precision.

Natural Language Understanding Question Answering

Training Task Experts through Retrieval Based Distillation

no code implementations7 Jul 2024 Jiaxin Ge, Xueying Jia, Vijay Viswanathan, Hongyin Luo, Graham Neubig

One of the most reliable ways to create deployable models for specialized tasks is to obtain an adequate amount of high-quality task-specific data.

Diversity Retrieval

Self-MoE: Towards Compositional Large Language Models with Self-Specialized Experts

no code implementations17 Jun 2024 Junmo Kang, Leonid Karlinsky, Hongyin Luo, Zhen Wang, Jacob Hansen, James Glass, David Cox, Rameswar Panda, Rogerio Feris, Alan Ritter

Our findings highlight the critical role of modularity, the applicability of Self-MoE to multiple base LLMs, and the potential of self-improvement in achieving efficient, scalable, and adaptable systems.

Math

Adaptive Query Rewriting: Aligning Rewriters through Marginal Probability of Conversational Answers

no code implementations16 Jun 2024 Tianhua Zhang, Kun Li, Hongyin Luo, Xixin Wu, James Glass, Helen Meng

A novel approach is then proposed to assess retriever's preference for these candidates by the probability of answers conditioned on the conversational query by marginalizing the Top-$K$ passages.

Conversational Question Answering Passage Retrieval +1

THREAD: Thinking Deeper with Recursive Spawning

1 code implementation27 May 2024 Philip Schroeder, Nathaniel Morgan, Hongyin Luo, James Glass

We apply THREAD in the settings of LLM task solving and question answering, where the dynamic threading allows the model to recursively decompose the given task or question into progressively simpler sub-problems that can be solved by separate child threads.

Few-Shot Learning Question Answering

RQ-RAG: Learning to Refine Queries for Retrieval Augmented Generation

no code implementations31 Mar 2024 Chi-Min Chan, Chunpu Xu, Ruibin Yuan, Hongyin Luo, Wei Xue, Yike Guo, Jie Fu

To this end, we propose learning to Refine Query for Retrieval Augmented Generation (RQ-RAG) in this paper, endeavoring to enhance the model by equipping it with capabilities for explicit rewriting, decomposition, and disambiguation.

In-Context Learning RAG +2

HALC: Object Hallucination Reduction via Adaptive Focal-Contrast Decoding

2 code implementations1 Mar 2024 Zhaorun Chen, Zhuokai Zhao, Hongyin Luo, Huaxiu Yao, Bo Li, Jiawei Zhou

While large vision-language models (LVLMs) have demonstrated impressive capabilities in interpreting multi-modal contexts, they invariably suffer from object hallucinations (OH).

Hallucination Object +2

Self-Specialization: Uncovering Latent Expertise within Large Language Models

no code implementations29 Sep 2023 Junmo Kang, Hongyin Luo, Yada Zhu, Jacob Hansen, James Glass, David Cox, Alan Ritter, Rogerio Feris, Leonid Karlinsky

Recent works have demonstrated the effectiveness of self-alignment in which a large language model is aligned to follow general instructions using instructional data generated from the model itself starting from a handful of human-written seeds.

Hallucination Instruction Following +2

Joint Audio and Speech Understanding

1 code implementation25 Sep 2023 Yuan Gong, Alexander H. Liu, Hongyin Luo, Leonid Karlinsky, James Glass

Humans are surrounded by audio signals that include both speech and non-speech sounds.

DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Models

3 code implementations7 Sep 2023 Yung-Sung Chuang, Yujia Xie, Hongyin Luo, Yoon Kim, James Glass, Pengcheng He

Despite their impressive capabilities, large language models (LLMs) are prone to hallucinations, i. e., generating content that deviates from facts seen during pretraining.

TruthfulQA

Entailment as Robust Self-Learner

1 code implementation26 May 2023 Jiaxin Ge, Hongyin Luo, Yoon Kim, James Glass

Experiments on binary and multi-class classification tasks show that SimPLE leads to more robust self-training results, indicating that the self-trained entailment models are more efficient and trustworthy than large language models on language understanding tasks.

Multi-class Classification Natural Language Understanding +1

SAIL: Search-Augmented Instruction Learning

no code implementations24 May 2023 Hongyin Luo, Yung-Sung Chuang, Yuan Gong, Tianhua Zhang, Yoon Kim, Xixin Wu, Danny Fox, Helen Meng, James Glass

Large language models (LLMs) have been significantly improved by instruction fine-tuning, but still lack transparency and the ability to utilize up-to-date knowledge and information.

Denoising Fact Checking +3

Listen, Think, and Understand

2 code implementations18 May 2023 Yuan Gong, Hongyin Luo, Alexander H. Liu, Leonid Karlinsky, James Glass

On the other hand, modern large language models (LLMs) exhibit emerging reasoning ability but they lack audio perception capabilities.

Ranked #3 on Music Question Answering on MusicQA (using extra training data)

Language Modelling Large Language Model +1

Chain of Thought Prompt Tuning in Vision Language Models

no code implementations16 Apr 2023 Jiaxin Ge, Hongyin Luo, Siyuan Qian, Yulu Gan, Jie Fu, Shanghang Zhang

Chain of Thought is a simple and effective approximation to human reasoning process and has been proven useful for natural language processing (NLP) tasks.

Domain Generalization Image Classification +4

Interpretable Unified Language Checking

1 code implementation7 Apr 2023 Tianhua Zhang, Hongyin Luo, Yung-Sung Chuang, Wei Fang, Luc Gaitskell, Thomas Hartvigsen, Xixin Wu, Danny Fox, Helen Meng, James Glass

Despite recent concerns about undesirable behaviors generated by large language models (LLMs), including non-factual, biased, and hateful language, we find LLMs are inherent multi-task language checkers based on their latent representations of natural and social knowledge.

Fact Checking Fairness +2

Logic Against Bias: Textual Entailment Mitigates Stereotypical Sentence Reasoning

1 code implementation10 Mar 2023 Hongyin Luo, James Glass

Due to their similarity-based learning objectives, pretrained sentence encoders often internalize stereotypical assumptions that reflect the social biases that exist within their training corpora.

Natural Language Inference Sentence +1

Meta-learning for downstream aware and agnostic pretraining

no code implementations6 Jun 2021 Hongyin Luo, Shuyan Dong, Yung-Sung Chuang, Shang-Wen Li

Neural network pretraining is gaining attention due to its outstanding performance in natural language processing applications.

Meta-Learning

Cooperative Self-training of Machine Reading Comprehension

1 code implementation NAACL 2022 Hongyin Luo, Shang-Wen Li, Mingye Gao, Seunghak Yu, James Glass

Pretrained language models have significantly improved the performance of downstream language understanding tasks, including extractive question answering, by providing high-quality contextualized word embeddings.

Extractive Question-Answering Machine Reading Comprehension +6

Knowledge Grounded Conversational Symptom Detection with Graph Memory Networks

2 code implementations EMNLP (ClinicalNLP) 2020 Hongyin Luo, Shang-Wen Li, James Glass

Given a set of explicit symptoms provided by the patient to initiate a dialog for diagnosing, the system is trained to collect implicit symptoms by asking questions, in order to collect more information for making an accurate diagnosis.

Goal-Oriented Dialog

Prototypical Q Networks for Automatic Conversational Diagnosis and Few-Shot New Disease Adaption

no code implementations19 May 2020 Hongyin Luo, Shang-Wen Li, James Glass

Experiments showed that the ProtoQN significantly outperformed the baseline DQN model in both supervised and few-shot learning scenarios, and achieves state-of-the-art few-shot learning performances.

Deep Reinforcement Learning Few-Shot Learning

Learning Word Representations with Cross-Sentence Dependency for End-to-End Co-reference Resolution

1 code implementation EMNLP 2018 Hongyin Luo, Jim Glass

In this work, we present a word embedding model that learns cross-sentence dependency for improving end-to-end co-reference resolution (E2E-CR).

Coreference Resolution Sentence

Adaptive Bidirectional Backpropagation: Towards Biologically Plausible Error Signal Transmission in Neural Networks

2 code implementations23 Feb 2017 Hongyin Luo, Jie Fu, James Glass

However, it has been argued that this is not biologically plausible because back-propagating error signals with the exact incoming weights are not considered possible in biological neural systems.

DrMAD: Distilling Reverse-Mode Automatic Differentiation for Optimizing Hyperparameters of Deep Neural Networks

1 code implementation5 Jan 2016 Jie Fu, Hongyin Luo, Jiashi Feng, Kian Hsiang Low, Tat-Seng Chua

The performance of deep neural networks is well-known to be sensitive to the setting of their hyperparameters.

Cannot find the paper you are looking for? You can Submit a new open access paper.