Papers with fine-tuning models
FlashBack: Efficient Retrieval-Augmented Language Modeling for Fast Inference (2025.findings-acl)
Copied to clipboard
| Challenge: | Retrieval-Augmented Language Modeling (RALM) is a popular approach for large language models. |
| Approach: | They propose a modular RALM that integrates large language models with documents from an external corpus to improve inference efficiency. |
| Outcome: | The proposed method improves inference efficiency with appending context pattern while maintaining decent performance after fine-tuning by Low-Rank Adaption. |
Elastic weight consolidation for better bias inoculation (2021.eacl-main)
Copied to clipboard
| Challenge: | Recent studies have shown that the lack of suitable inductive biases in sentence-pair classification models can cause misclassifications on training datasets. |
| Approach: | They propose to use elastic weight consolidation (EWC) to fine-tune models to mitigate biases while being less susceptible to catastrophic forgetting. |
| Outcome: | The proposed model improves on fact verification and stress tests while maintaining the original task accuracy. |
LegalLens: Leveraging LLMs for Legal Violation Identification in Unstructured Text (2024.eacl-long)
Copied to clipboard
Dor Bernsohn, Gil Semo, Yaron Vazana, Gila Hayat, Ben Hagag, Joel Niklaus, Rohit Saha, Kyryl Truskovskyi
| Challenge: | a recent study focused on detecting legal violations within unstructured textual data . a similar study focused only on associating violations with potentially affected individuals . |
| Approach: | They constructed two datasets using Large Language Models (LLMs) they publicize the results to advance legal natural language processing research . |
| Outcome: | The proposed datasets and the code used for the experiments have been released to advance legal natural language processing (NLP) |
Exploring the Universal Vulnerability of Prompt-based Learning Paradigm (2022.findings-naacl)
Copied to clipboard
| Challenge: | Prompt-based learning inherits the vulnerability from pre-training, where model predictions can be misled by inserting triggers into the text. |
| Approach: | They propose a potential solution to mitigate this vulnerability by injecting triggers into pre-trained language models using only plain text. |
| Outcome: | The proposed learning paradigm inherits the vulnerability from the pre-training stage . it can totally control or severely decrease the performance of prompt-based models . |
Advancing Precise Outline-Conditioned Text Generation with Task Duality and Explicit Outline Control (2024.eacl-long)
Copied to clipboard
| Challenge: | Existing studies on outline-conditioned text generation focus on generating text using provided outlines as rough sketches, but lack of clarity and rationality of the rough outlines hampers quality of the generated text. |
| Approach: | They propose a novel task that requires generating stories based on specific, sentence-level outlines. |
| Outcome: | The proposed framework improves the quality of precise outline-conditioned text generation. |
Domaino1s: Guiding LLM Reasoning for Explainable Answers in High-Stakes Domains (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing Large Language Models (LLMs) generate brief answers without reasoning processes and explanations. |
| Approach: | They propose supervised fine-tuning and tree search to enhance LLMs’ reasoning capabilities on domain tasks. |
| Outcome: | The proposed model improves on stock investment recommendation and legal reasoning QA tasks. |
SLIM: Subtrajectory-Level Elimination for More Effective Reasoning (2025.findings-emnlp)
Copied to clipboard
Xifeng Yao, Chengyuan Ma, Dongyu Lang, Yinhao Ni, Zhiwei Xu, Huarui Xie, Zihao Chen, Guang Shen, Dandan Tu, Yi Bai, Changzheng Zhang
| Challenge: | Notable examples include OpenAI’s o1/o3/o4 series and DeepSeek-R1 . |
| Approach: | They develop a framework to identify suboptimal subtrajectories based on human-established criteria . they also use a sampling algorithm to select data whose reasoning process is free from suboptimally subtravertories to the highest degree . |
| Outcome: | The proposed method reduces the number of suboptimal subtrajectories by 25.9% during the inference process. |
ODD: A Benchmark Dataset for the Natural Language Processing Based Opioid Related Aberrant Behavior Detection (2024.naacl-long)
Copied to clipboard
Sunjae Kwon, Xun Wang, Weisong Liu, Emily Druhl, Minhee Sung, Joel Reisman, Wenjun Li, Robert Kerns, William Becker, Hong Yu
| Challenge: | Opioid related aberrant behaviors (ORABs) present novel risk factors for opioid overdose. |
| Approach: | They propose to use a biomedical natural language processing benchmark dataset to classify ORABs from patients’ EHR notes into nine categories: confirmed aberrant behavior, suggested aberrant behaviors, Opioids, indication, diagnosed opioid dependency, Benzodiazepines, medication changes, and Central Nervous System-related. |
| Outcome: | The proposed dataset outperforms two state-of-the-art models in most categories and the gains are especially higher among uncommon classes. |
FroM: Frobenius Norm-Based Data-Free Adaptive Model Merging (2025.findings-emnlp)
Copied to clipboard
| Challenge: | a new adaptive merging method is proposed to improve fine-tuning performance . traditional methods often encounter task interference when merging full fine-uning models . |
| Approach: | They propose an adaptive merging method that directly measures model parameters using the Frobenius norm . |
| Outcome: | The proposed method outperforms baseline methods in various fine-tuning scenarios. |
On the Feasibility of In-Context Probing for Data Attribution (2025.findings-naacl)
Copied to clipboard
| Challenge: | In-context probing (ICP) can be used to identify training data that contributes to model outputs, but many data attribution methods, such as influence functions, use model gradients and are computationally expensive. |
| Approach: | They propose to use in-context probing (ICP) to proxy for gradient-based data attribution for data selection under conditions contingent on data similarity. |
| Outcome: | The proposed method can be used to identify training data that contribute to model outputs and fine tune models on training data. |
Synthetic Eggs in Many Baskets: The Impact of Synthetic Data Diversity on LLM Fine-Tuning (2026.findings-acl)
Copied to clipboard
| Challenge: | Increasing demand for training data is causing language models to be trained on synthetic data, a new study finds . fine-tuning models on synthetic datasets reduces self-preference bias . |
| Approach: | They investigate the impact of diversity of synthetic data on fine-tuned large language models. |
| Outcome: | The proposed model can mitigate distribution collapse, maintain diversity of output distribution, and reduce self-preference bias. |
MathFusion: Enhancing Mathematical Problem-solving of LLM through Instruction Fusion (2025.acl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have shown impressive progress in mathematical problem-solving . current approaches to enhance mathematical reasoning focus on instance-level modifications . |
| Approach: | They propose a framework that enhances mathematical reasoning through cross-problem instruction synthesis. |
| Outcome: | The proposed framework boosts mathematical reasoning by 18.0 points while maintaining high data efficiency. |
A Generative Adaptive Replay Continual Learning Model for Temporal Knowledge Graph Reasoning (2025.acl-long)
Copied to clipboard
| Challenge: | Existing Continual Learning (CL)-based Temporal Knowledge Graph Reasoning methods are incomplete and reorganize historical facts without preserving historical knowledge. |
| Approach: | They propose a method which generates and adaptively replays historical entity distributions from the whole historical context. |
| Outcome: | The proposed method outperforms baselines in reasoning and mitigating forgetting. |
Modeling Legal Reasoning: LM Annotation at the Edge of Human Agreement (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing research examines simple classification tasks, but ability of LMs to classify on complex tasks is less well understood. |
| Approach: | They analyze a Supreme Court opinion annotated by a team of domain experts . they find generative models perform poorly when given instructions equal to human annotators . |
| Outcome: | The proposed model performs poorly when given instructions equal to instructions given to human annotations . strongest results derive from fine-tuning models on the annotated dataset . |
Multitask-Bench: Unveiling and Mitigating Safety Gaps in LLMs Fine-tuning (2025.coling-main)
Copied to clipboard
| Challenge: | Recent advances in Large Language Models (LLMs) have led to their adoption across a wide range of tasks, ranging from code generation to machine translation and sentiment analysis. |
| Approach: | They propose to fine-tune LLMs on benign (non-harmful) data to ensure safe outputs. |
| Outcome: | The proposed model reduces attack success rates across a range of tasks without compromising its usefulness. |
CITB: A Benchmark for Continual Instruction Tuning (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for instruction tuning do not leverage the rich natural language instructions. |
| Approach: | They propose to use a benchmark to study how instruction tuning works in CL tasks. |
| Outcome: | The proposed method can achieve similar or better results than existing CL methods. |
MONETA: Multimodal Industry Classification through Geographic Information with Multi Agent Systems (2026.acl-long)
Copied to clipboard
| Challenge: | MONETA is the first multimodal industry classification benchmark with text and geospatial sources. |
| Approach: | They propose a multimodal industry classification benchmark using text and geospatial sources. |
| Outcome: | The proposed model increases the accuracy of the existing models by 22.80%. |
Automated Fact-Checking in Dialogue: Are Specialized Models Needed? (2023.emnlp-main)
Copied to clipboard
| Challenge: | Prior work has shown that typical fact-checking models struggle with claims made in conversation. |
| Approach: | They propose to fine-tune models for dialogue on conversational data to improve performance on typical fact-checking. |
| Outcome: | The proposed models perform better on stand-alone claims than state-of-the-art models for dialogue while maintaining their performance on standalone claim. |
SAFENUDGE: Safeguarding Large Language Models in Real-time with Tunable Safety-Performance Trade-offs (2025.emnlp-main)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) are susceptible to jailbreak attacks, or adversarial prompts eliciting high-risk behavior. |
| Approach: | They propose a safeguard that combines Controlled Text Generation and "nudging" it adds minimal latency to inference and reduces successful jailbreak attempts by up to 37.3% . |
| Outcome: | The proposed safeguard reduces successful jailbreak attempts by between 28.1% and 37.3% by guiding the LLM towards a safe response. |
DeKeyNLU: Enhancing Natural Language to SQL Generation through Task Decomposition and Keyword Extraction (2025.findings-emnlp)
Copied to clipboard
Jian Chen, Zhenyan Chen, Xuming Hu, Peilin Zhou, Yining Hua, Han Fang, Cissy Hing Yee Choy, Xinmei Ke, Jingfeng Luo, Zixuan Yuan
| Challenge: | NL2SQL provides a model-centric paradigm that simplifies database access for non-technical users . challenges such as inaccurate task decomposition and keyword extraction remain major bottlenecks . |
| Approach: | They propose a RAG-based NL2SQL pipeline that employs three modules for query understanding, entity retrieval, and generation to improve SQL generation accuracy. |
| Outcome: | The proposed pipeline improves the accuracy of query generation on BIRD and Spider datasets. |
Training Language Models to Use Prolog as a Tool (2026.findings-acl)
Copied to clipboard
| Challenge: | Language models often produce plausible but incorrect reasoning traces that are difficult to verify. |
| Approach: | They train language models to use Prolog as an external symbolic reasoning tool . they find an accuracy–auditability trade-off between tuning for correctness alone and using Prolog only for the final computation . |
| Outcome: | The proposed model outperforms supervised fine-tuning on a clean version of GSM8K. |
UPDESH: Synthesizing Grounded Instruction Tuning Data for 13 Indic Languages (2026.acl-long)
Copied to clipboard
Pranjal A Chitale, Varun Gumma, Sanchit Ahuja, Prashant Kodali, Manan Uppadhyay, Deepthi Sudharsan, Sunayana Sitaram
| Challenge: | Developing culturally grounded multilingual AI systems is challenging for low-resource languages . synthetic data is underexplored, but its effectiveness in multilingual and multicultural contexts is understudied . |
| Approach: | They propose a top-up synthetic data generation framework grounded in Wikipedia content . they use 9.5M data points across 13 Indian languages and English to generate a high-quality dataset . |
| Outcome: | The proposed model improves on NLG tasks and narrows performance gaps with high-resource languages. |