Papers by Shou-De Lin

11 papers
Neuron-Level Differentiation of Memorization and Generalization in Large Language Models (2025.emnlp-main)

Copied to clipboard

Challenge: Existing models exhibit memorization and generalization behaviors in ways that are not easily interpretable or controllable.
Approach: They propose to use a GPT-2 and LLaMA-3.2 model to identify distinct neuron subsets responsible for each behavior to steer the model toward memorization or generalization.
Outcome: The proposed models show that inference-time interventions on these neurons can steer the model’s behavior toward memorization or generalization.
Benchmarking Uncertainty Metrics for LLM Target-Aware Search (2025.findings-emnlp)

Copied to clipboard

Challenge: Existing uncertainty metrics for LLM search methods do not capture the diverse types of uncertainty needed to guide different optimization goals.
Approach: They propose a framework for uncertainty benchmarking that captures four different uncertainty types . the uncertainty types Answer, Correctness, Aleatoric, and Epistemic serve different optimization goals .
Outcome: The proposed framework identifies four different uncertainty types . the uncertainty types serve different optimization goals in LLM search .
Role-Sensitive Neurons: A Neuron-Level Gain Control Mechanism for Confidence Steering (2026.findings-acl)

Copied to clipboard

Challenge: Large language models (LLMs) exhibit striking behavioral flexibility.
Approach: They propose to identify a sparse sub-network of Role-Sensitive Neurons (RSNs) that governs the transition from hesitation to action.
Outcome: The proposed framework allows precise regulation of abstention behavior by intervention on this subspace.
Beyond Facts- Benchmarking Distributional Reading Comprehension in Large Language Models (2026.findings-acl)

Copied to clipboard

Challenge: Existing reading comprehension benchmarks focus on factual information, but many real-world tasks require distributional knowledge expressed across text.
Approach: They propose a reading comprehension benchmark for LLMs to evaluate their ability to infer distributional knowledge from natural language.
Outcome: Experiments with multiple LLMs show that the model outperforms baselines, but performance varies widely across distribution types and characteristics.
LiveCLKTBench: Towards Reliable Evaluation of Cross-Lingual Knowledge Transfer in Multilingual LLMs (2026.acl-long)

Copied to clipboard

Challenge: Evaluating cross-lingual knowledge transfer in large language models is challenging, as correct answers in a target language may arise either from genuine transfer or from prior exposure during pre-training.
Approach: They propose a pipeline to isolate and measure cross-lingual knowledge transfer by identifying self-contained, time-sensitive knowledge entities from real-world domains and generating factual questions.
Outcome: The proposed pipeline analyzes multiple LLMs across five languages and shows that cross-lingual transfer is strongly influenced by linguistic distance and often asymmetric across language directions.
Self-Discriminative Learning for Unsupervised Document Embedding (N19-1)

Copied to clipboard

Challenge: Existing methods for document embedding learning do not consider inter-document relationships.
Approach: They propose to exploit the inter-document information and directly model the relations of documents in embedding space with a discriminative network and a novel objective.
Outcome: The proposed method has errors that are 5 to 13% lower than state-of-the-art models and is even more pronounced in scarce label setting.
Transferable Embedding Inversion Attack: Uncovering Privacy Risks in Text Embeddings without Model Queries (2024.acl-long)

Copied to clipboard

Challenge: Recent advances in text embedding models have significantly streamlined the process of generating embeddables.
Approach: They develop a transfer attack method that uses a surrogate model to mimic the victim model's behavior and infers sensitive information from embeddings without direct access.
Outcome: The proposed method outperforms existing methods and reveals potential privacy vulnerabilities in embedding technologies.
Multiple Text Style Transfer by using Word-level Conditional Generative Adversarial Network with Two-Phase Training (D19-1)

Copied to clipboard

Challenge: Generative adversarial network (GAN) is a popular model for text style transfer . but, training GAN often suffers from mode collapse problem, which causes that the transferred text is little related to the original text.
Approach: They propose a non-parallel text style transfer model with a word-level conditional architecture and a two-phase training procedure to maintain style-unrelated words while changing others.
Outcome: The proposed model outperforms state-of-the-art models on three real-world datasets in transfer accuracy and fluency.
Text-centric Alignment for Bridging Test-time Unseen Modality (2025.findings-emnlp)

Copied to clipboard

Challenge: a text-centric alignment method is used to handle unseen modalities and dynamic modality combinations at test time.
Approach: They propose a text-centric alignment method that unifies different input modalities into a single semantic text representation by leveraging in-context learning with Large Language Models and uni-modal foundation models.
Outcome: The proposed method unifies input modalities into a single semantic representation . it significantly improves the ability to manage unseen, diverse, and unpredictable modality combinations .
Word Relation Autoencoder for Unseen Hypernym Extraction Using Word Embeddings (D18-1)

Copied to clipboard

Challenge: Lexicon relation extraction given distributional representation of words is an important topic in NLP.
Approach: They propose to use a word relation autoencoder to extract hypernyms from vocabularies . they propose to analyze the pollution and construct an indicator to measure it .
Outcome: The proposed model outperforms the competitors on several hypernym-like lexicon datasets.
Controlling Sequence-to-Sequence Models - A Demonstration on Neural-based Acrostic Generator (D19-3)

Copied to clipboard

Challenge: acrostic is a form of writing that the first token of each line forms a meaningful sequence.
Approach: They propose a generalized acrostic generation system that can hide certain messages in a flexible pattern specified by the users.
Outcome: The proposed system can hide certain messages in a flexible pattern specified by the users.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations