Papers by Mengdi Zhang

11 papers
Do Large Language Models excel in Complex Logical Reasoning with Formal Language? (2025.emnlp-main)

Copied to clipboard

Challenge: Existing studies on LLMs have focused on formal language, but evaluations of their performance are limited.
Approach: They propose to use a formal language to evaluate LLMs across logical reasoning problems using formal languages.
Outcome: The proposed model outperforms Instruct models in three dimensions, taxonomy of tasks, and format of trajectories, and achieves the best generalization performance across other languages.
SCoder: Progressive Self-Distillation for Bootstrapping Small-Scale Data Synthesizers to Empower Code LLMs (2025.findings-emnlp)

Copied to clipboard

Challenge: Existing code large language models rely on large-scale instruction data distilled from proprietary LLMs for fine-tuning, which typically incurs high costs.
Approach: They propose an iterative self-distillation approach to bootstrap small-scale LLMs . they use large-scale instruction data distilled from proprietary LLM for fine-tuning .
Outcome: The proposed method reduces reliance on proprietary LLMs and minimizes costs.
The Role of Visual Modality in Multimodal Mathematical Reasoning: Challenges and Insights (2025.acl-long)

Copied to clipboard

Challenge: Existing models that leverage visual information do not improve math reasoning performance . authors suggest that visual information is important for multimodal reasoning .
Approach: They propose a dataset to require image reliance for problem-solving and challenge models with similar, yet distinct, images that change the correct answer.
Outcome: The proposed model performance is unaffected by changes to or removal of images in the dataset.
LogicPro: Improving Complex Logical Reasoning via Program-Guided Learning (2025.acl-long)

Copied to clipboard

Challenge: LogicPro is a data synthesis method that uses LeetCode-style algorithm problems and their corresponding Program solutions to generate complex logic data.
Approach: They propose a new method which leverages LeetCode-style algorithm Problems and their corresponding Program solutions to synthesize complex logic data in text format.
Outcome: The proposed method outperforms existing models for BBH27, LogicBench, DROP, AR-LSAT, and GSM8K, and a wide range of reasoning datasets.
How Do Your Code LLMs perform? Empowering Code Instruction Tuning with Really Good Data (2024.emnlp-main)

Copied to clipboard

Challenge: Recent research has shown that code pre-trained models improve coding capabilities.
Approach: They propose a code data pruning strategy to identify which datasets are high-quality code instruction data.
Outcome: The proposed model achieves state-of-the-art performance using fewer training data.
DolphCoder: Echo-Locating Code Large Language Models with Diverse and Multi-Objective Instruction Tuning (2024.acl-long)

Copied to clipboard

Challenge: Numerous code large language models (LLMs) have been proposed to enhance code generation performance.
Approach: They propose a diverse instruction model DolphCoder with self-evaluating for code generation that learns diverse instruction targets and combines a code evaluation objective to enhance its code generation ability.
Outcome: The proposed model achieves superior performance on the HumanEval and MBPP benchmarks, demonstrating new insights for future code instruction tuning work.
Revisit Self-Debugging with Self-Generated Tests for Code Generation (2025.acl-long)

Copied to clipboard

Challenge: Large language models (LLMs) have made significant advances in code generation, but they still face challenges when tackling complex programming tasks beyond their basic capabilities.
Approach: They propose to integrate self-generated tests into the code generation process . they propose to use post-execution and in-exection self-debugging to mitigate test bias .
Outcome: The proposed method improves the performance of large language models in code generation tasks by leveraging execution feedback from tests.
Incorporating Dynamic Semantics into Pre-Trained Language Model for Aspect-based Sentiment Analysis (2022.findings-acl)

Copied to clipboard

Challenge: Aspect-based sentiment analysis (ABSA) predicts sentiment polarity towards a specific aspect in a sentence.
Approach: They propose to use a dynamic aspect-oriented semantics-based method to learn ABSA.
Outcome: The proposed method can learn dynamic aspect-oriented semantics for ABSA on three benchmark datasets.
SWE-Mutation: Can LLMs Generate Reliable Test Suites in Software Engineering? (2026.findings-acl)

Copied to clipboard

Challenge: Evaluating software engineering capabilities is a core component of large language models (LLMs).
Approach: They propose a benchmark to evaluate LLM-generated test suites that introduces mutated solutions that attempt to "fool" them.
Outcome: The proposed test suites are based on 2,636 mutated variants derived from 800 original instances and include a multilingual subset spanning nine programming languages.
Prejudge-Before-Think: Enhancing Large Language Models at Test-Time by Process Prejudge Reasoning (2025.findings-emnlp)

Copied to clipboard

Challenge: Experimental results from competition-level complex reasoning demonstrate that bootstrapping with process prejudge can significantly enhance the reasoning ability of LLMs.
Approach: They propose a new process prejudge strategy for LLM reasoning that bootstraps with process prejudgment .
Outcome: The proposed method can be bootstrapped with process prejudge in LLM reasoning . it allows the model to anticipate errors rather than relying on trial and error.
Contrastive Multi-document Question Generation (2021.eacl-main)

Copied to clipboard

Challenge: Multi-document question generation focuses on generating a question that covers the common aspect of multiple documents, but a naive model trained only using the targeted document set may generate too generic questions that cover a larger scope than delineated by the document set.
Approach: They propose a contrastive learning strategy where given ‘positive’ and ‘negative’ sets of documents, generate a question that is closely related to the ‘positive' set but far away from the ‘negative' set.
Outcome: The proposed model significantly outperforms several strong baselines, as measured by automatic metrics and human evaluation.

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