Papers by Minghan Li

15 papers
S2G-RAG: Structured Sufficiency and Gap Judging for Iterative Retrieval-Augmented QA (2026.acl-long)

Copied to clipboard

Challenge: Retrieval-augmented generation grounds language models in external evidence, but multi-hop question answering remains difficult . iterative pipelines must control what to retrieve next and when evidence is adequate.
Approach: They propose an iterative framework with an explicit controller, S2G-Judge . they map structured gap items into the next retrieval query to produce stable retrieval trajectories .
Outcome: Experiments on TriviaQA, HotpotQA, and 2WikiMultiHopQA show that S2G-RAG improves multi-hop QA performance and robustness under multi-turn retrieval.
Benchmarking Vision-Language Models on Chinese Ancient Documents: From OCR to Knowledge Reasoning (2026.findings-acl)

Copied to clipboard

Challenge: Existing document benchmarks focus on English printed texts or simplified Chinese . current vision-language models struggle with visual complexity and poor adaptability .
Approach: They propose a benchmark to evaluate Chinese ancient documents' visual/linguistic complexity . ancient documents are valuable cultural heritage, but they face challenges in digitization and understanding .
Outcome: the first benchmark for Chinese ancient documents evaluates VLMs from OCR to knowledge reasoning . ancient documents carry thousands of years of Chinese history and culture . traditional methods only scan images, while current models struggle with visual complexity .
CELI: Simple yet Effective Approach to Enhance Out-of-Domain Generalization of Cross-Encoders. (2024.naacl-short)

Copied to clipboard

Challenge: Existing cross-encoders do not capture all information into the [CLS] token . Xiong et al., 2021) find that the out-of-domain approach is less robust.
Approach: They introduce a cross-encoder with late interaction that incorporates a late interaction layer into existing models.
Outcome: The proposed method improves BEIR by 5% without compromising in-domain effectiveness or search latency.
Loki: An Open-Source Tool for Fact Verification (2025.coling-demos)

Copied to clipboard

Challenge: Loki is an open-source fact-checking tool designed to address the growing problem of misinformation.
Approach: They propose a tool that breaks down the fact-checking task into five steps . they propose LOKI, which offers a semiautomated, human-in-the-loop approach .
Outcome: a new open-source tool is designed to address the growing problem of misinformation . the tool breaks down the fact-checking task into five steps to assist human judgment .
INarIG: Iterative Non-autoregressive Instruct Generation Model For Word-Level Auto Completion (2023.findings-emnlp)

Copied to clipboard

Challenge: Existing models for word-level autocompletion (WLAC) only use human typed sequences as prefixes in decoding module.
Approach: They propose a novel iterative nonautoregressive instruct generation model for WLAC task . it uses human typed sequences and iterating decoding with subwords to fully utilize input information.
Outcome: The proposed model is more competent in dealing with low-frequency words, and achieves state-of-the-art results on the WMT22 and benchmark datasets.
CITADEL: Conditional Token Interaction via Dynamic Lexical Routing for Efficient and Effective Multi-Vector Retrieval (2023.acl-long)

Copied to clipboard

Challenge: Existing multi-vector retrieval methods are slower and require more space to store indices compared to their single-vektor counterparts.
Approach: They propose a multi-vector retrieval method that uses dynamic lexical routing to route different token vectors to the predicted lexicals.
Outcome: The proposed method achieves state-of-the-art performance on several benchmark datasets while being nearly 40 times faster than the current state-out-of the-art method.
Aggretriever: A Simple Approach to Aggregate Textual Representations for Robust Dense Passage Retrieval (2023.tacl-1)

Copied to clipboard

Challenge: Pre-trained language models have been successful in knowledge-intensive tasks, but recent research calls into question the robustness of these singlevector models.
Approach: They propose to exploit knowledge in a pre-trained language model for dense passage retrieval by aggregating contextualized token embeddings into a dense vector.
Outcome: The proposed model significantly improves the effectiveness of dense retrieval models on in-domain and zero-shot evaluations without introducing substantial training overhead.
Unifying Multimodal Retrieval via Document Screenshot Embedding (2024.emnlp-main)

Copied to clipboard

Challenge: Existing document retrieval pipelines require document parsing and content extraction to prepare input for indexing.
Approach: They propose a retrieval paradigm that regards document screenshots as a unified input format . they leverage a large vision-language model to directly encode document screenshot into dense representations .
Outcome: The proposed method outperforms existing retrieval pipelines in a text-intensive context.
Multi-Task Dense Retrieval via Model Uncertainty Fusion for Open-Domain Question Answering (2021.findings-emnlp)

Copied to clipboard

Challenge: Existing approaches to multitask dense retrieval are not effective due to corpus inconsistency.
Approach: They propose to train individual dense passage retrievers for different open-domain question-answering tasks and aggregate their predictions during test time.
Outcome: The proposed method achieves state-of-the-art performance on 5 benchmark QA datasets, with up to 10% improvement in top-100 accuracy compared to a joint-training multi-task DPR on SQuAD.
Text Style Transfer Back-Translation (2023.acl-long)

Copied to clipboard

Challenge: Current methods require large amount of bilingual training data, which is challenging and sometimes impossible task.
Approach: They propose a method to modify the style of inputs by modifying the source side of BT data.
Outcome: The proposed method significantly improves translation quality against popular BT benchmarks on high-resource and low-resourced language pairs.
GLIER: Generative Legal Inference and Evidence Ranking for Legal Case Retrieval (2026.acl-long)

Copied to clipboard

Challenge: Existing dense retrieval methods neglect the explicit legal logic that underpins legal relevance.
Approach: They propose a framework that reformulates retrieval as an inference process over latent legal variables.
Outcome: GLIER outperforms strong baselines like SAILER and KELLER in a legal case-based retrieval task . the framework exhibits exceptional data efficiency even when trained with only 10% of the data .
Domain Adaptation for Dense Retrieval and Conversational Dense Retrieval through Self-Supervision by Meticulous Pseudo-Relevance Labeling (2024.lrec-main)

Copied to clipboard

Challenge: Recent studies have shown that dense retrieval models generalize less well than interaction-based models on out-of-distribution data sets.
Approach: They propose to combine query-generation approach with self-supervision approach in which pseudo-relevance labels are automatically generated on the target domain.
Outcome: The proposed approach is based on a T5-3B model for pseudo-positive labeling and hard negatives on conversational dense retrieval models.
Simple and Effective Unsupervised Redundancy Elimination to Compress Dense Vectors for Passage Retrieval (2021.emnlp-main)

Copied to clipboard

Challenge: Dense passage retrieval improves ranking accuracy in open-domain question answering but at the cost of large space and memory requirements.
Approach: They propose a simple unsupervised pipeline that includes principal component analysis (PCA), product quantization, and hybrid search to improve space efficiency.
Outcome: The proposed pipeline achieves good accuracy–space trade-offs, for example, 48 compression with less than 3% drop in top-100 retrieval accuracy on average or 96 compression without drop in space requirements.
Certified Error Control of Candidate Set Pruning for Two-Stage Relevance Ranking (2022.emnlp-main)

Copied to clipboard

Challenge: In information retrieval, candidate set pruning is used to speed up two-stage relevance ranking but lacks accurate error control and empirical guarantees.
Approach: They propose a method that guarantees the test error after pruning is controlled under a user-specified threshold with high probability.
Outcome: The proposed method reduces the average set size from 1000 to 27, increasing reranking speed by about 37 times while keeping MRR@10 greater than a pre-specified value of 0.38 with about 90% empirical coverage.
How to Train Your Dragon: Diverse Augmentation Towards Generalizable Dense Retrieval (2023.findings-emnlp)

Copied to clipboard

Challenge: Existing techniques to improve dense retrieval suffer from effectiveness tradeoffs between supervised and zero-shot retrieval, some argue due to the limited model capacity.
Approach: They propose to use diverse queries and sources of supervision to train a generalizable DR to achieve high accuracy in both supervised and zero-shot retrieval.
Outcome: The proposed DR can achieve state-of-the-art in supervised and zero-shot evaluations without increasing model size.

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