Papers by Yunxiang Zhang
MultiHiertt: Numerical Reasoning over Multi Hierarchical Tabular and Textual Data (2022.acl-long)
Copied to clipboard
| Challenge: | Existing benchmarks for numerical reasoning over hybrid data only include a single flat table in each document . |
| Approach: | They propose a new benchmark with QA pairs over multi hierarchical tabular and textual data. |
| Outcome: | The proposed model is more complex and challenging than existing benchmarks and is available on github . it uses facts retrieving to extract relevant facts from both tables and text and symbolic reasoning over retrieved facts. |
Mind Reader: Latent User Demand-Guided Content Optimization for Generative Search Engine (2026.acl-long)
Copied to clipboard
Tong Chen, JiaWei Guo, Yuxi Li, Baiming Chen, Houxing Ren, Zhang Zhiwei, Yunxiang Zhang, Hanyang Xia, Kun Liang, Zhaoran Fan
| Challenge: | Generative Search Engines (GSEs) have reshaped information retrieval and Generating Engine Optimization (GEO) emerges to improve the content visibility in GSEs’ responses. |
| Approach: | They propose a method to optimize content to cover latent semantic information of GSEs by decomposing query into diverse perspectives and capturing underlying semantic information. |
| Outcome: | The proposed method outperforms baselines and effectively improves content visibility (with up to 2.44x objective metrics and 1.23x subjective metrics on average). |
Generate, Discriminate, Evolve: Enhancing Context Faithfulness via Fine-Grained Sentence-Level Self-Evolution (2025.findings-acl)
Copied to clipboard
Kun Li, Tianhua Zhang, Yunxiang Li, Hongyin Luo, Abdalla Mohamed Salama Sayed Moustafa, Xixin Wu, James R. Glass, Helen M. Meng
| Challenge: | Existing methods to improve context faithfulness in large language models are either inadequate or overlook the potential for self-improvement. |
| Approach: | They propose a framework that enhances context faithfulness through fine-grained sentence-level optimization. |
| Outcome: | Experiments on ASQA and ConFiQA datasets show that GenDiE surpasses baselines in faithfulness and correctness and exhibits robust performance for domain adaptation. |
RAG-Zeval: Enhancing RAG Responses Evaluator through End-to-End Reasoning and Ranking-Based Reinforcement Learning (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing evaluation frameworks rely on direct prompting of resource-intensive models with complex multi-stage prompts, introducing significant computational cost and underutilizing models’ reasoning capabilities. |
| Approach: | They propose a framework that trains evaluators with reinforcement learning to generate comprehensive and sound assessments with detailed explanation in one-pass. |
| Outcome: | The proposed framework outperforms baseline evaluation frameworks that rely on LLMs with 10-100 more parameters and achieves the strongest correlation with human judgments. |
Learning to Ideate for Machine Learning Engineering Agents (2026.eacl-short)
Copied to clipboard
Yunxiang Zhang, Kang Zhou, Zhichao Xu, Kiran Ramnath, Yun Zhou, Sangmin Woo, Haibo Ding, Lin Lee Cheong
| Challenge: | Existing machine learning engineering (MLE) agents struggle to iteratively optimize their implemented algorithms for effectiveness. |
| Approach: | They propose a framework that separates ideation from implementation that allows an implementation agent to request strategic help from a dedicated Ideator. |
| Outcome: | The proposed framework outperforms implementation-only agent baselines on MLE-Bench and can be trained with reinforcement learning to generate more effective ideas. |
Logit Arithmetic Elicits Long Reasoning Capabilities Without Training (2026.findings-acl)
Copied to clipboard
Yunxiang Zhang, Muhammad Khalifa, Lechen Zhang, Xin Liu, Ayoung Lee, Xinliang Frederick Zhang, Farima Fatahi Bayat, Lu Wang
| Challenge: | Large reasoning models exhibit long chain-of-thought reasoning with complex strategies such as backtracking and self-verification, yet, these capabilities typically require resource-intensive post-training. |
| Approach: | They propose a decoding-time approach which transfers long chain-of-thought reasoning capabilities from a substantially smaller reasoning guider to a large non-reasoning target. |
| Outcome: | The proposed method improves performance over a model 21x smaller than the target model by 21.5% and 24.2% over the model. |
Interpreting the Robustness of Neural NLP Models to Textual Perturbations (2022.findings-acl)
Copied to clipboard
| Challenge: | Modern Natural Language Processing models are sensitive to input perturbations and their performance can decrease when applied to noisy data. |
| Approach: | They propose to explain the extent to which a model is affected by an unseen textual perturbation by the learnability of the perturbation. |
| Outcome: | The proposed model is better at identifying a perturbation (higher learnability) but worse at ignoring it (lower robustness). |
Skill-Aware Data Selection and Fine-Tuning for Data-Efficient Reasoning Distillation (2026.acl-short)
Copied to clipboard
| Challenge: | Large reasoning models such as DeepSeek-R1 and their distilled variants achieve impressive performance on complex reasoning tasks, yet their costs remain substantial. |
| Approach: | They propose a skill-centric distillation framework that efficiently transfers reasoning ability to weaker models with two components: (1) Skill-based data selection, which prioritizes examples targeting the student model’s weaker skills, and (2) Skillaware fine-tuning, which encourages explicit skill decomposition during problem solving. |
| Outcome: | The proposed framework surpasses baselines on Qwen3-4B and Qwend3-8B and focuses on skills emphasized during training. |
MOVER: Mask, Over-generate and Rank for Hyperbole Generation (2022.naacl-main)
Copied to clipboard
| Challenge: | despite being a common figure of speech, hyperbole is under-researched in Figurative Language Processing . we use an unsupervised method to generate hyperbolic paraphrases from literal sentences . |
| Approach: | They propose an unsupervised method for hyperbole generation that does not require parallel literal-hyperbole pairs. |
| Outcome: | The proposed method outperforms baseline systems and is based on a large-scale English hyperbole corpus. |
Merging Generated and Retrieved Knowledge for Open-Domain QA (2023.emnlp-main)
Copied to clipboard
| Challenge: | Open-domain question answering systems often have retrieval modules but retrieving passages from external knowledge sources is known to suffer from insufficient knowledge coverage. |
| Approach: | They propose a Compatibility-Oriented knowledge Merging framework to leverage both sources of information by matching LLM-generated passages with retrieved counterparts into compatible pairs. |
| Outcome: | The proposed framework outperforms baselines on three out of four tested open-domain QA benchmarks. |
Small Language Models Need Strong Verifiers to Self-Correct Reasoning (2024.findings-acl)
Copied to clipboard
Yunxiang Zhang, Muhammad Khalifa, Lajanugen Logeswaran, Jaekyeom Kim, Moontae Lee, Honglak Lee, Lu Wang
| Challenge: | Existing studies show that large language models can self-correct their outputs by generating a critique and revising it based on the critique. |
| Approach: | They propose a pipeline that prompts small language models to collect self-correction data that supports the training of self-refinement abilities. |
| Outcome: | The proposed pipeline improves the self-correction abilities of two models on five datasets spanning math and commonsense reasoning. |