Papers by Yuntian Deng
Sequence-to-Lattice Models for Fast Translation (2021.findings-emnlp)
Copied to clipboard
| Challenge: | Existing non-autoregressive machine translation models have decoders that are difficult to port to NAT models. |
| Approach: | They propose a sequence-to-lattice model that replaces the decoder with a search lattice. |
| Outcome: | The proposed model is faster than past non-autoregressive generation approaches and more accurate than reducing the number of decoder layers. |
WildVis: Open Source Visualizer for Million-Scale Chat Logs in the Wild (2024.emnlp-demo)
Copied to clipboard
| Challenge: | Currently, the volume and complexity of chat logs makes it difficult to analyze individual conversations. |
| Approach: | They propose a tool that enables fast, versatile, and large-scale conversation analysis by combining search and visualization capabilities with a list of criteria. |
| Outcome: | The proposed tool can be extended to handle millions of chat logs and other datasets. |
Rationales for Sequential Predictions (2021.emnlp-main)
Copied to clipboard
| Challenge: | Sequence models produce accurate predictions, but their decision making processes are hard to explain. |
| Approach: | They propose an efficient algorithm to approximate sequential objective by identifying the most faithful rationales. |
| Outcome: | The proposed algorithm is best at optimizing the sequential objective and provides the most faithful rationales. |
Interactive Training: Feedback-Driven Neural Network Optimization (2025.emnlp-demos)
Copied to clipboard
| Challenge: | In traditional neural network training, static optimization methods lack flexibility and responsiveness . authors demonstrate that Interactive Training provides superior training stability and reduced sensitivity to initial hyperparameters . |
| Approach: | They propose an open-source framework that enables real-time feedback-driven optimization of neural networks by human experts or automated AI agents. |
| Outcome: | The proposed framework achieves superior training stability, reduced sensitivity to initial hyperparameters, and improved adaptability to evolving user needs. |
Neural Linguistic Steganography (D19-1)
Copied to clipboard
| Challenge: | linguistic steganography encrypts a secret message into a cover signal . language is a pragmatic cover signal due to its benign occurrence and independence from any one medium. |
| Approach: | They propose a technique that encrypts a secret message into a cover signal . language is a particularly pragmatic cover signal due to its benign occurrence . |
| Outcome: | The proposed technique generates realistic looking cover sentences as evaluated by humans while preserving security by matching the cover message distribution with the language model distribution. |
Tree Prompting: Efficient Task Adaptation without Fine-Tuning (2023.emnlp-main)
Copied to clipboard
| Challenge: | Pretrained language models (LMs) are the main interface for applying them to new tasks, but their large size makes them difficult to fine-tune with gradients for specific downstream tasks. |
| Approach: | They propose to use training data to form a decision tree based on prompt-LM calls, with each prompt determined by the outcomes of previous calls. |
| Outcome: | The proposed method improves accuracy over competing methods and is competitive with fine-tuning. |
TokDrift: When LLM Speaks in Subwords but Code Speaks in Grammar (2026.acl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) for code rely on subword tokenizers learned from mixed natural language text and programming language code but driven by statistics rather than grammar. |
| Approach: | They propose a framework that applies semantic-preserving rewrite rules to create code variants differing only in tokenization. |
| Outcome: | The proposed framework can create code variants differing only in tokenization . the findings highlight the need for grammar-aware tokenization for future code LLMs. |
From Chat Logs to Collective Insights: Aggregative Question Answering (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing approaches to analyzing large-scale conversation logs treat interactions as independent, missing critical insights. |
| Approach: | They propose a task that requires models to reason explicitly over thousands of user-chatbot interactions to answer aggregational queries. |
| Outcome: | The proposed task requires models to reason over thousands of user-chatbot interactions to answer aggregational queries such as identifying emerging concerns among demographics. |
Bottom-Up Abstractive Summarization (D18-1)
Copied to clipboard
| Challenge: | Existing approaches to summarize text using end-to-end content selectors have had mixed success in content selection, for example copying full sentences from the source document. |
| Approach: | They propose to use content selectors to over-determine phrases in a source document that should be part of the summary. |
| Outcome: | The proposed model over-determines phrases in a source document that should be part of the summary while generating fluent summaries. |
Model Criticism for Long-Form Text Generation (2022.emnlp-main)
Copied to clipboard
| Challenge: | Language models generate fluent text, but it remains unclear whether output retains coherent high-level structure. |
| Approach: | They propose to use a statistical tool to evaluate high-level structure of text . they compare distributions between real and generated data in latent space . |
| Outcome: | The proposed model criticism compares distributions between real and generated data in a latent space . different generative processes identify specific failure modes of the underlying model . |