Papers by Wu Ning

31 papers
Knowledgeable Prompt-tuning: Incorporating Knowledge into Prompt Verbalizer for Text Classification (2022.acl-long)

Copied to clipboard

Challenge: Recent studies suggest that pre-trained language models have gained rich knowledge during pre-training.
Approach: They propose to tune pre-trained language models with task-specific prompts to improve and stabilize prompttuning.
Outcome: Extensive experiments on zero and few-shot text classification tasks show that prompt-tuning improves and stabilizes prompttun-ing.
From Adversarial Arms Race to Model-centric Evaluation: Motivating a Unified Automatic Robustness Evaluation Framework (2023.findings-acl)

Copied to clipboard

Challenge: Existing models of robustness evaluation are incomprehensive, impractical, and invalid .
Approach: They propose a framework for automatic robustness evaluation that shifts towards model-centric evaluation to further exploit the advantages of adversarial attacks.
Outcome: The proposed framework is based on a model-centric evaluation protocol and a robustness evaluation protocol.
Interpretable Safety Alignment via SAE-Constructed Low-Rank Subspace Adaptation (2026.acl-long)

Copied to clipboard

Challenge: Prior work has shown that safety behaviors are governed by low-rank structures . Low-Rank Adaptation (LoRA) consistently underperforms full fine-tuning and reinforcement learning on safety benchmarks .
Approach: They propose a safety alignment system that disentangles safety-relevant directions into monosemantic features and constructs an interpretable safety subspace from SAE directions.
Outcome: Empirically, the proposed model achieves 99.6% safety rates across multiple model families and scales . low-rank Adaptation consistently underperforms full fine-tuning and reinforcement learning on safety benchmarks compared with previous methods .
Selected Languages are All You Need for Cross-lingual Truthfulness Transfer (2025.coling-main)

Copied to clipboard

Challenge: Existing methods for truthfulness enhancement in English are limited to multilingual scenarios.
Approach: They propose a method for cross-lingual truthfulness transfer that uses language bias and transfer contributions to select an optimal subset of all tested languages and employ translation instruction tuning for cross language truthfulness transfers.
Outcome: The proposed method reduces multilingual representation disparity and boosts cross-lingual truthfulness transfer of LLMs.
Empowering Dual-Encoder with Query Generator for Cross-Lingual Dense Retrieval (2022.emnlp-main)

Copied to clipboard

Challenge: Existing methods to distill knowledge from cross-encoder re-ranker to dual-encoding retriever are lacking in the cross-lingual setting.
Approach: They propose to use a query generator as the teacher in the cross-lingual setting to distill knowledge to a dual-encoder retrieval model.
Outcome: The proposed method outperforms state-of-the-art methods on two benchmark datasets.
Step Guided Reasoning: Improving Mathematical Reasoning using Guidance Generation and Step Reasoning (2025.emnlp-main)

Copied to clipboard

Challenge: Existing approaches to improve mathematical reasoning require extensive datasets for training or depend on few-shot methods that compromise computational accuracy.
Approach: They propose a training-free adaptation framework that efficiently equips general-purpose pre-trained language models with enhanced mathematical reasoning capabilities.
Outcome: The proposed framework outperforms Qwen2.5-72B-Math-Instruct on MMLU-STEM with a score of 90.9%, compared to 87.3%.
Quantifying and Improving the Robustness of Retrieval-Augmented Language Models Against Spurious Features in Grounding Data (2026.acl-long)

Copied to clipboard

Challenge: Existing studies on robustness to explicit noise (e.g., document semantics) but overlook implicit noise (spurious features).
Approach: They propose a framework to quantify the robustness of RAGs against spurious features by integrating a data synthesis pipeline and a taxonomy.
Outcome: The proposed framework quantifies the robustness of RALMs against spurious features.
SEAD: Self-Evolving Agent for Multi-Turn Service Dialogue (2026.findings-acl)

Copied to clipboard

Challenge: Large Language Models have demonstrated remarkable capabilities in open-domain dialogues, but their performance in service dialogues remains suboptimal.
Approach: They propose a framework that enables agents to learn effective strategies without large-scale human annotations.
Outcome: The proposed framework decouples user modeling into two components that provide adaptive training scenarios rather than acting as an unfair adversary.
Improving Temporal Relation Extraction with a Globally Acquired Statistical Resource (N18-1)

Copied to clipboard

Challenge: Existing temporal extraction systems that extract temporal relations can be improved by using a resource that provides prior knowledge of the temporal order that events usually follow.
Approach: They propose to use a probabilistic knowledge base acquired in the news domain to extract temporal relations between events from the New York Times articles over a 20-year span.
Outcome: The proposed system and resource are both publicly available.
Lexicon-Enhanced Self-Supervised Training for Multilingual Dense Retrieval (2022.findings-emnlp)

Copied to clipboard

Challenge: Recent multilingual pre-trained models perform poorly on multilingual retrieval tasks due to lack of multilingual training data.
Approach: They propose to mine and generate self-supervised training data based on large-scale unlabeled corpus and introduce query generator to generate more queries in target languages for unlabed passages.
Outcome: The proposed method performs better than baselines on a Mr. TYDI dataset and an industrial dataset from a commercial search engine.
Beyond Rejection Sampling: Trajectory Fusion for Scaling Mathematical Reasoning (2026.findings-acl)

Copied to clipboard

Challenge: Large language models (LLMs) fine-tuned using rejection sampling retain only correct reasoning trajectories . however, this paradigm treats supervision as a binary filter that systematically excludes teacher-generated errors, leaving a gap in how reasoning failures are modeled during training.
Approach: They propose a fine-tuning strategy that reframes rejection sampling as a structured supervision construction process.
Outcome: The proposed approach outperforms RFT on multiple math benchmarks while retaining only correct reasoning trajectories.
MoSE: Modality Split and Ensemble for Multimodal Knowledge Graph Completion (2022.emnlp-main)

Copied to clipboard

Challenge: Existing methods to predict missing entities share relation representation across modalities, which results in mutual interference between modality.
Approach: They propose a framework for multimodal knowledge graph completion that learns modality-split relation embeddings for each modality instead of a single modality shared one.
Outcome: The proposed framework outperforms state-of-the-art methods on three KG datasets.
Easy, Reproducible and Quality-Controlled Data Collection with CROWDAQ (2020.emnlp-demos)

Copied to clipboard

Challenge: Efficient data collection is important for advancing research and building time-sensitive applications.
Approach: They propose an open-source platform that standardizes the data collection pipeline . it includes customizable user interface components, automated annotator qualification, and saved pipelines .
Outcome: The proposed platform simplifies data annotation significantly on diverse datasets . it can be used by researchers and engineers to improve reproducibility and minimize overhead .
GraphDx: A Cost-Aware Knowledge-Enhanced Multi-Agent Framework for Sequential Diagnosis (2026.findings-acl)

Copied to clipboard

Challenge: Existing Large Language Models struggle to reason systematically under cost constraints . Existing approaches lack the knowledge-reasoning capability to reason under cost .
Approach: They propose a knowledge-enhanced framework that leverages large language models to construct MDKGs . they propose three collaborative agents that handle language understanding and generation .
Outcome: GraphDx improves diagnostic success rates from 50–68% to 79–93% while reducing test costs by 20–54%.
E2E-GMNER: End-to-End Generative Grounded Multimodal Named Entity Recognition (2026.findings-acl)

Copied to clipboard

Challenge: Existing approaches decouple textual entity recognition and visual grounding, leading to error accumulation and suboptimal joint optimization.
Approach: They propose a fully end-to-end generative framework that unifies recognition, semantic typing, visual grounding and implicit knowledge reasoning within a single multimodal large language model.
Outcome: The proposed framework achieves highly competitive performance compared with state-of-the-art methods.
V-GameGym: Visual Game Generation for Code Large Language Models (2026.findings-acl)

Copied to clipboard

Challenge: Existing code-related benchmarks focus on single modality rather than visual game development.
Approach: They propose a multimodal benchmark for evaluating code large language models in visual game generation that integrates a clustering-based curation methodology and a pipeline for visual code synthesis.
Outcome: The proposed framework assesses code generation and visual game generation using a sandbox environment.
From Behavioral Performance to Internal Competence: Interpreting Vision-Language Models with VLM-Lens (2025.emnlp-demos)

Copied to clipboard

Challenge: Existing vision-language models are based on exactmatch based accuracy and its derivations to evaluate performance.
Approach: They propose a toolkit that supports systematic benchmarking, analysis, and interpretation of vision-language models by extracting intermediate outputs from any layer during the forward pass of open-source VLMs.
Outcome: The proposed toolkit supports 16 state-of-the-art base VLMs and their over 30 variants and is extensible to accommodate new models without changing the core logic.
Self-adaptive Context and Modal-interaction Modeling For Multimodal Emotion Recognition (2023.findings-acl)

Copied to clipboard

Challenge: Existing methods to predict emotion label for a given utterance lack modeling of diverse dependency ranges and inconsistent treatment of contribution for various modalities.
Approach: They propose a multimodal emotion recognition in conversation task that uses context and multiple modalities to predict emotion label for a given utterance.
Outcome: The proposed method outperforms the state-of-the-art methods on three multimodal datasets.
Enhancing Learning-Based Binary Code Similarity Detection Model through Adversarial Training with Multiple Function Variants (2024.findings-emnlp)

Copied to clipboard

Challenge: Existing Learning-Based Binary Code Similarity Detection (LB-BCSD) methods exhibit lower accuracy in recognizing functions with the same functionality but different implementations.
Approach: They propose a gradient-guided adversarial attack method based on critical code called FuncFooler which perturbs critical code to generate multiple variants of the same function.
Outcome: The proposed method increases the accuracy of the current Learning-Based Binary Code Similarity Detection (LB-BCSD) model by 5%-7%.
Challenges to Open-Domain Constituency Parsing (2022.findings-acl)

Copied to clipboard

Challenge: Existing findings on cross-domain constituency parsing are only made on a limited number of domains.
Approach: They manually annotate a high-quality constituency treebank containing five domains and analyze challenges to open-domain constituency parsing using a set of linguistic features.
Outcome: The proposed model significantly improves the performance of the proposed model on the domain-variant features.
Overcoming Language Priors in Visual Question Answering via Distinguishing Superficially Similar Instances (2022.coling-1)

Copied to clipboard

Challenge: Existing VQA models rely on the superficial correlation between question type and frequent answers to make predictions, without really understanding the input.
Approach: They propose a training framework that explicitly encourages the VQA model to distinguish between superficially similar instances.
Outcome: The proposed framework achieves state-of-the-art performance on VQA-CP v2 . it explicitly encourages the model to distinguish between the superficially similar instances .
TEMP: Taxonomy Expansion with Dynamic Margin Loss through Taxonomy-Paths (2021.emnlp-main)

Copied to clipboard

Challenge: Existing taxonomies are unable to maintain coverage due to the rising of new concepts . TEMP uses pre-trained contextual encoders to predict the position of new ideas .
Approach: They propose a self-supervised taxonomy expansion method that ranks taxonomies by ranking them . they use pre-trained contextual encoders to train the model with dynamic margin loss .
Outcome: The proposed method outperforms state-of-the-art taxonomy expansion methods by 14.3% and 15.8% on public benchmarks.
XGLUE: A New Benchmark Dataset for Cross-lingual Pre-training, Understanding and Generation (2020.emnlp-main)

Copied to clipboard

Challenge: XGLUE provides a benchmark dataset to train large-scale cross-lingual pre-trained models . XCLUE provides 11 diversified tasks that cover both understanding and generation scenarios .
Approach: They introduce a new benchmark dataset to train large-scale cross-lingual pre-trained models using multilingual and bilingual corpora.
Outcome: The proposed dataset is labeled in English and includes only natural language understanding tasks.
TORQUE: A Reading Comprehension Dataset of Temporal Ordering Questions (2020.emnlp-main)

Copied to clipboard

Challenge: Current machine reading comprehension benchmarks have no questions that test temporal phenomena . a new study studies reading comprehension for temporal relations .
Approach: They propose a reading comprehension benchmark built on news snippets and 21k human-generated questions querying temporal relationships.
Outcome: The new reading comprehension benchmark TORQUE achieves an exact-match score of 51% on the test set . the benchmark is built on 3.2k news snippets with 21k human-generated questions .
A Meta-framework for Spatiotemporal Quantity Extraction from Text (2022.acl-long)

Copied to clipboard

Challenge: a meta-framework for news events that extracts quantities from text is proposed . a previous work on news events focused on extracting event mentions, attributes, and relationships .
Approach: They propose a meta-framework for solving the NLP problem of spatiotemporal quantity extraction . they demonstrate the framework is general and extensible, and shareable crowdsourcing pipeline and baseline models are used .
Outcome: The proposed framework is general and extensible, the authors say . it can extract quantity from news streams, quickly respond to emergencies, investigate incidents .
ControlMath: Controllable Data Generation Promotes Math Generalist Models (2024.emnlp-main)

Copied to clipboard

Challenge: Currently, mathematical reasoning is one of the most challenging areas for closed-source LLMs.
Approach: They propose an iterative method involving an equation-generator module and two LLM-based agents that generate diverse equations and transform them into math word problems.
Outcome: The proposed method enables the generation of diverse math problems, not limited to specific domains or distributions.
Breaking Language Barriers in Multilingual Mathematical Reasoning: Insights and Observations (2024.findings-emnlp)

Copied to clipboard

Challenge: Existing research focuses on developing powerful large language models for mathematical reasoning within monolingual languages.
Approach: They propose to use translation to build powerful multilingual math reasoning models . they propose different training strategies to build xMR LLMs that outperform open-source LLM .
Outcome: The proposed model outperforms open-source LLMs and surpasses ChatGPT in few-shot scenarios.
A Multi-Axis Annotation Scheme for Event Temporal Relations (P18-1)

Copied to clipboard

Challenge: Existing temporal relation (TempRel) annotation schemes have low inter-annotator agreements even between experts, suggesting that the current annotation task needs a better definition.
Approach: They propose to annotate temporal relation (TempRel) annotation schemes based on event start-points instead of a conventional 60’s-80’s model.
Outcome: The proposed model improves IAA from the conventional 60’s to 80’s and can be used by crowdsourcing to alleviate labor intensity.
iTool: Reinforced Fine-Tuning with Dynamic Deficiency Calibration for Advanced Tool Use (2025.emnlp-main)

Copied to clipboard

Challenge: Synthesizing tool-use data through real-world simulations is effective for enhancing large language models (LLMs) however, training gains decay as synthetic data increases, and the model struggles to benefit from more synthetic data.
Approach: They propose an iterative reinforced fine-tuning strategy to improve LLMs with external tools to augment their capabilities.
Outcome: The proposed method achieves 13.11% better performance than the same-size base model and outperforms larger open-source and closed-source models.
Joint Reasoning for Temporal and Causal Relations (P18-1)

Copied to clipboard

Challenge: a cause must occur earlier than its effect, temporal and causal relations are closely related . a joint inference framework is developed for studying temporal, causal relations .
Approach: They propose a joint inference framework for temporal and causal relations . they use constraints inherent in time and causality to enforce constraints .
Outcome: The proposed framework improves extraction of temporal and causal relations from text.
TinyJudge: Unverifiable Constraint Alignment via Lightweight Specialist Ensembles (2026.acl-long)

Copied to clipboard

Challenge: Instruction Following (IF) is a core capability of LLMs, requiring strict adherence to diverse constraints.
Approach: They propose a framework that uses tiny language models to evaluate instruction following . they propose to use a set of specialized tiny language model to provide rewards for soft constraints.
Outcome: The proposed framework outperforms baseline models by 12% and speeds up training time by 3.

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