Papers by Feng Xia
CLaMP 3: Universal Music Information Retrieval Across Unaligned Modalities and Unseen Languages (2025.findings-acl)
Copied to clipboard
Shangda Wu, Guo Zhancheng, Ruibin Yuan, Junyan Jiang, SeungHeon Doh, Gus Xia, Juhan Nam, Xiaobing Li, Feng Yu, Maosong Sun
| Challenge: | Music information retrieval (MIR) is a field that aims at developing computational tools for processing, organizing, and accessing music data. |
| Approach: | They propose a framework that aligns music modalities with multilingual text in a shared representation space. |
| Outcome: | Experiments show CLaMP 3 performs state-of-the-art on multiple MIR tasks . it surpasses baselines and shows excellent generalization in multimodal and multilingual contexts . |
Beyond Linguistic Cues: Fine-grained Conversational Emotion Recognition via Belief-Desire Modelling (2024.lrec-main)
Copied to clipboard
| Challenge: | Emotion recognition in conversation (ERC) is essential for dialogue systems to identify the emotions expressed by speakers. |
| Approach: | They propose a method that incorporates both belief and desire to accurately identify emotions by extracting emotion-eliciting events from utterances and construct graphs that represent beliefs and desires in conversations. |
| Outcome: | The proposed model outperforms existing models on four popular ERC datasets and validates its performance with multiple state-of-the-art models. |
Pruning Pre-trained Language Models Without Fine-Tuning (2023.acl-long)
Copied to clipboard
| Challenge: | Existing methods to prune Pre-trained Language Models (PLMs) are overparameterized and require fine-tuning. |
| Approach: | They propose a pruning method that uses first-order pruning to prune PLMs while fine-tuning the remaining weights. |
| Outcome: | The proposed method outperforms first-order pruning and zero-order methods at sparsity levels. |
Teaching Large Language Models Number-Focused Headline Generation With Key Element Rationales (2025.findings-naacl)
Copied to clipboard
| Challenge: | Existing studies focus only on textual quality and numerical accuracy for headline generation. |
| Approach: | They propose a framework for using rationales of key elements of Topic, Entities, and Numerical reasoning in news articles to enhance LLMs' ability to generate topic-aligned texts with precise numerical accuracy. |
| Outcome: | The proposed framework improves the ability of large language models to generate high-quality texts with precise numerical accuracy. |
SPD-Faith Bench: Diagnosing and Improving Faithfulness in Chain-of-Thought for Multimodal Large Language Models (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing studies on multimodal faithfulness have focused on perceptual hallucinations, raising concerns about the validity of reasoning traces. |
| Approach: | They propose a diagnostic benchmark that enforces explicit visual comparison to assess faithfulness of reasoning traces. |
| Outcome: | The proposed framework improves visual routing and aligns reasoning with perception. |
UniCOQE: Unified Comparative Opinion Quintuple Extraction As A Set (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing methods decompose the COQE task into multiple subtasks and solve them in a pipeline manner, but ignore the intrinsic connection between subtask and the error propagation among stages. |
| Approach: | They propose a unified generative model that solves COQE in one shot by concatenating all the comparative tuples into a target output sequence. |
| Outcome: | The proposed model significantly outperforms the SOTA method on multiple benchmarks and ablation experiments. |
Telling the Whole Story: A Manually Annotated Chinese Dataset for the Analysis of Humor in Jokes (D19-1)
Copied to clipboard
| Challenge: | Humor plays important role in human communication, which makes it important problem for natural language processing. |
| Approach: | They propose a novel annotation scheme to give scenarios of how humor arises in text . they report reasonable agreement between annotators and analyze the dataset . |
| Outcome: | The proposed scheme gives scenarios of how humor arises in text . it contains key words that trigger humor, character relationship, scene, and humor categories . |
Automatic Document Selection for Efficient Encoder Pretraining (2022.emnlp-main)
Copied to clipboard
| Challenge: | Pretraining language models is expensive and data-intensive, but can it be improved? Several studies have found that directly pretraining on task data is more effective . |
| Approach: | They propose to automatically identify smaller yet domain-representative subsets by pretraining a model on a target domain. |
| Outcome: | The proposed method outperforms random selection on perplexity and downstream tasks with 20x less data and 3x fewer training iterations and 2x less estimated cloud compute cost. |
Please refuse to answer me! Mitigating Over-Refusal in Large Language Models via Adaptive Contrastive Decoding (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for mitigating over-refusal can't maintain low refusal ratio for harmless queries while keeping high for malicious queries. |
| Approach: | They propose a model-agnostic approach to mitigate over-refusal in large language models . they propose an adaptive contrastive decoding strategy that incorporates or removes the refusal token distribution . |
| Outcome: | The proposed approach reduces the refusal ratio for over-refusal queries by 10.35% while increasing the refusal rate for malicious queries by 0.13%. |
FOFO: A Benchmark to Evaluate LLMs’ Format-Following Capability (2024.acl-long)
Copied to clipboard
| Challenge: | Existing benchmarks fail to assess large language models’ format-following proficiency adequately. |
| Approach: | They propose a benchmark to evaluate large language models' ability to follow complex, domain-specific formats. |
| Outcome: | The proposed framework evaluates large language models' ability to follow complex, domain-specific formats across open-source and closed-source models. |
SURVEYFORGE : On the Outline Heuristics, Memory-Driven Generation, and Multi-dimensional Evaluation for Automated Survey Writing (2025.acl-long)
Copied to clipboard
| Challenge: | SURVEYFORGE automates survey paper writing, but quality gap between LLM-generated and human-written surveys remains significant. |
| Approach: | They propose a survey tool that automatically generates and refines human-written surveys. |
| Outcome: | Experiments show that SURVEYFORGE outperforms previous work such as AutoSurvey in outline quality and content quality. |
Premise-based Multimodal Reasoning: Conditional Inference on Joint Textual and Visual Clues (2022.acl-long)
Copied to clipboard
Qingxiu Dong, Ziwei Qin, Heming Xia, Tian Feng, Shoujie Tong, Haoran Meng, Lin Xu, Zhongyu Wei, Weidong Zhan, Baobao Chang, Sujian Li, Tianyu Liu, Zhifang Sui
| Challenge: | Existing work in vision language cross-modal reasoning uses binary or multi-choice classification based on source image and textual query. |
| Approach: | They propose a task where a textual premise is the background presumption on each source image. |
| Outcome: | The proposed task is based on a dataset of 15,360 movie screenshots and human-curated premise templates from 6 pre-defined categories. |
SciCustom: A Framework for Custom Evaluation of Scientific Capabilities in Large Language Models (2026.acl-long)
Copied to clipboard
Yiyang Gu, Junwei Yang, Junyu Luo, Ye Yuan, Bin Feng, Yingce Xia, Shufang Xie, Kaili Liu, Bohan Wu, Qi Shi, Haoran Li, Beier Xiao, Zhiping Xiao, Xiao Luo, Weizhi Zhang, Philip S. Yu, Zequn Liu, Ming Zhang
| Challenge: | Existing evaluations of large language models fail to reflect fine-grained capabilities . existing benchmarks are manually curated or domain-generic, limiting scalability and alignment with real use cases. |
| Approach: | They propose a framework that allows custom construction of benchmarks from large-scale scientific data to evaluate application-specific scientific capabilities in LLMs. |
| Outcome: | The proposed framework reveals fine-grained differences in scientific capabilities that standard benchmarks overlook . it allows custom construction of benchmarks from large-scale scientific data to evaluate application-specific capabilities in LLMs. |
BFS-Prover: Scalable Best-First Tree Search for LLM-based Automatic Theorem Proving (2025.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to theorem proving in large language models rely on value functions and/or Monte Carlo Tree Search (MCTS), but the potential of simpler methods like Best-First Tree Search remains underexplored. |
| Approach: | They propose a scalable expert iteration framework that implements strategic data filtering at each expert iteration round, excluding problems solvable via beam search node expansion to focus on harder cases. |
| Outcome: | The proposed framework achieves a state-of-the-art score of 72.95 on the MiniF2F test set and challenges the perceived necessity of complex tree search methods. |
Ask Again, Then Fail: Large Language Models’ Vacillations in Judgment (2024.acl-long)
Copied to clipboard
| Challenge: | Existing large language models often waver in their judgments when faced with follow-up questions . this is a challenge for generating reliable responses and building user trust . |
| Approach: | They propose a Follow-up Questioning Mechanism and two metrics to quantify this inconsistency . they also develop a framework that teaches large language models to maintain original correct judgments . |
| Outcome: | The proposed framework improves the general capabilities of large language models by allowing them to maintain original correct judgments. |
Cultural Bias Matters: A Cross-Cultural Benchmark Dataset and Sentiment-Enriched Model for Understanding Multimodal Metaphors (2025.acl-long)
Copied to clipboard
| Challenge: | Metaphors are pervasive in communication, making them crucial for natural language processing. |
| Approach: | They propose a multicultural multimodal metaphor dataset designed for cross-cultural studies of metaphor in Chinese and English. |
| Outcome: | The proposed model improves metaphor comprehension across cultural backgrounds and cultural domains. |
Debiasing Large Language Models via Adaptive Causal Prompting with Sketch-of-Thought (2026.findings-eacl)
Copied to clipboard
| Challenge: | Existing prompting methods for Large Language Models (LLMs) suffer from excessive token usage and limited generalisability across diverse reasoning tasks. |
| Approach: | They propose an Adaptive Causal Prompting with Sketch-of-Thought framework that leverages structural causal models to infer the causal effect of a query on its answer. |
| Outcome: | The proposed framework outperforms existing prompting baselines in terms of accuracy, robustness, and computational efficiency. |
Incremental Few-shot Text Classification with Multi-round New Classes: Formulation, Dataset and System (2021.naacl-main)
Copied to clipboard
| Challenge: | Text classification is usually studied by labeling texts with relevant categories from a predefined set. |
| Approach: | They propose a task where a system incrementally handles multiple rounds of new classes . they propose two entailment approaches, ENTAILMENT and HYBRID, which show promise . |
| Outcome: | The proposed task is based on a few-shot text classification task in the NLP domain. |
AutoL2S: Auto Long-Short Reasoning for Efficient Large Language Models (2026.findings-acl)
Copied to clipboard
Feng Luo, Yu-Neng Chuang, Guanchu Wang, Hoang Anh Duy Le, Shaochen Zhong, Hongyi Liu, Jiayi Yuan, Yang Sui, Vladimir Braverman, Vipin Chaudhary, Xia Hu
| Challenge: | Existing approaches to distilling large language models (LLMs) are inefficient and generate excessively long chain-of-thought reasoning even for inputs that admit concise solutions. |
| Approach: | They propose a distillation framework that empowers non-reasoning LLMs to think only when necessary. |
| Outcome: | The proposed framework reduces reasoning length up to 71% with minimal accuracy loss while preserving accuracy. |