Papers with gradient-based optimization
Parameter-efficient Tuning for Large Language Model without Calculating Its Gradients (2023.emnlp-main)
Copied to clipboard
| Challenge: | Recent parameter-efficient tuning methods can only save 30% of training memory . gradient computation and backpropagation are still necessary for these methods . |
| Approach: | They propose a parameter-efficient tuning method that can be used to fine-tune large language models without calculating gradients. |
| Outcome: | The proposed method saves 30% of training memory and improves performance on large language models. |
Latent-Variable Generative Models for Data-Efficient Text Classification (D19-1)
Copied to clipboard
| Challenge: | Generative classifiers offer potential advantages over discriminative classifications, including data efficiency and zero-shot learning. |
| Approach: | They introduce discrete latent variables into generative story to improve classifiers' performance . they empirically characterize performance of their models on six text classification datasets . |
| Outcome: | The proposed model outperforms discriminative and generative classifiers on six text classification datasets. |
SEQˆ3: Differentiable Sequence-to-Sequence-to-Sequence Autoencoder for Unsupervised Abstractive Sentence Compression (N19-1)
Copied to clipboard
| Challenge: | Neural sequence-to-sequence models are currently the dominant approach in natural language processing tasks, but require massive parallel corpora. |
| Approach: | They propose a sequence-to-sequence-tosequnce autoencoder with words as latent variables . they apply the model to unsupervised abstractive sentence compression . |
| Outcome: | The proposed model achieves promising results in unsupervised sentence compression on benchmark datasets. |
Iterative Refinement in the Continuous Space for Non-Autoregressive Neural Machine Translation (2020.emnlp-main)
Copied to clipboard
| Challenge: | Existing non-autoregressive inference procedures that refine in token space often require computational overhead. |
| Approach: | They propose an efficient inference procedure that iteratively refines translation purely in the continuous space using a latent variable instead of the latent variables. |
| Outcome: | The proposed procedure is twice as efficient and more effective than the existing EM-like inference procedure. |
It’s Not Just Size That Matters: Small Language Models Are Also Few-Shot Learners (2021.naacl-main)
Copied to clipboard
| Challenge: | Pretraining ever-larger language models on massive corpora requires enormous amounts of compute. |
| Approach: | They propose to convert textual inputs into cloze questions that contain a task description . they also exploit unlabeled data to improve their performance . |
| Outcome: | The proposed model outperforms GPT-3 with PET/iPET with cloze questions and unlabeled data. |
DiLM: Distilling Dataset into Language Model for Text-level Dataset Distillation (2024.findings-naacl)
Copied to clipboard
| Challenge: | Existing methods to extract word embeddings from training datasets are not efficient for training other models. |
| Approach: | They propose a method to distill a training dataset into a textual model by combining a small number of informative synthetic samples. |
| Outcome: | The proposed method outperforms existing methods on training datasets and language models. |
Zer0-Jack: A memory-efficient gradient-based jailbreaking method for black box Multi-modal Large Language Models (2026.eacl-long)
Copied to clipboard
| Challenge: | Multi-modal large language models are highly vulnerable to jailbreak attacks due to their additional modality. |
| Approach: | They propose a black-box jailbreak framework based on zeroth-order optimization . they propose generating malicious images and patch-wise block coordinate descent . |
| Outcome: | The proposed framework achieves 98.2% success on MiniGPT-4 and 95% on the Harmful Behaviors Multi-modal dataset while jailbreaking commercial models such as GPT-4o. |
Logical Neural Networks for Knowledge Base Completion with Embeddings & Rules (2022.emnlp-main)
Copied to clipboard
Prithviraj Sen, Breno William Carvalho, Ibrahim Abdelaziz, Pavan Kapanipathi, Salim Roukos, Alexander Gray
| Challenge: | Knowledge base completion (KBC) is a human-interpretable dialect . rule-based KBC has a high quality but low accuracy . |
| Approach: | They propose to use logical neural networks to learn both kinds of rules in a common framework using gradient-based optimization. |
| Outcome: | The proposed method improves by 10% relative to SotA rule-based methods and by combining it with knowledge graph embeddings it achieves an additional 7.5% relative improvement. |
Investigating Robustness and Interpretability of Link Prediction via Adversarial Modifications (N19-1)
Copied to clipboard
| Challenge: | Existing approaches focus on improving accuracy and overlook other aspects such as robustness and interpretability. |
| Approach: | They propose adversarial modifications for link prediction models that identify influential facts and evaluate their sensitivity to addition of fake facts. |
| Outcome: | The proposed model evaluates the robustness of the model to the addition of fake facts and the interpretability of the models. |
Enhancing the Transferability of Jailbreak Attacks on Large Language Models via Exploiting Reparameterization Invariance (2026.acl-long)
Copied to clipboard
| Challenge: | Existing token-level attacks have shown efficacy on open-source models but suffer from poor cross-model transferability. |
| Approach: | They propose a framework to improve cross-model transferability by modifying model parameters and generating update directions according to differences in output distributions rather than parameter-space distances. |
| Outcome: | The proposed framework improves cross-model transferability and success rates on open-source models. |
Reasoning as Gradient: Scaling MLE Agents Beyond Tree Search (2026.findings-acl)
Copied to clipboard
Yifei Zhang, Xu Yang, Xiao Yang, Bowen Xian, Qizheng Li, Shikai Fang, Jingyuan Li, Jian Wang, Minrui Xu, Yuge Zhang, Weiqing Liu, Jiang Bian
| Challenge: | LLM-based agents for machine learning engineering rely on tree search to rank candidates. |
| Approach: | They propose an LLM-based agent that operationalizes gradient-based optimization. |
| Outcome: | The proposed agent achieves a state-of-the-art 35.1% any-medal rate on MLE-Bench with a limited budget on a single GPU. |
Gradient-based Adversarial Attacks against Text Transformers (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for obtaining adversarial examples are difficult with text data. |
| Approach: | They propose a gradient-based adversarial attack against transformer models that searches for a distribution of adversarials parameterized by a continuous-valued matrix. |
| Outcome: | The proposed attack outperforms existing methods on a variety of natural language tasks with matching imperceptibility. |
Tell Me How to Ask Again: Question Data Augmentation with Controllable Rewriting in Continuous Space (2020.emnlp-main)
Copied to clipboard
| Challenge: | Existing data augmentation techniques for natural language processing tasks are difficult to design. |
| Approach: | They propose a controllable rewriting based question data augmentation method for machine reading comprehension, question generation and question-answering natural language inference tasks. |
| Outcome: | The proposed method generates high-quality, high-quality question data samples on machine reading comprehension, question generation, and question-answering natural language inference tasks. |
Activation Scaling for Steering and Interpreting Language Models (2024.findings-emnlp)
Copied to clipboard
| Challenge: | a successful intervention should flip the correct with the wrong token, while remaining sparse. |
| Approach: | They propose to use activation scaling to flip the correct with the wrong token . they use gradient-based optimization to learn and evaluate a specific kind of efficient intervention . |
| Outcome: | The proposed method performs comparable with steering vectors but is much less minimal. |
Latent Factor Models Meets Instructions: Goal-conditioned Latent Factor Discovery without Task Supervision (2025.naacl-long)
Copied to clipboard
Zhouhang Xie, Tushar Khot, Bhavana Dalvi Mishra, Harshit Surana, Julian McAuley, Peter Clark, Bodhisattwa Prasad Majumder
| Challenge: | Instruction-following LLMs have recently allowed systems to discover hidden concepts from a collection of unstructured documents based on a natural language description of the purpose of the discovery (i.e., goal). |
| Approach: | They propose a goal-oriented latent factor discovery system that integrates LLM’s instruction-following ability with statistical models to handle large, noisy datasets where LLM reasoning alone falls short. |
| Outcome: | The proposed system improves task performance by 5-52% over baselines and 1.8 times as often as the best alternative, on average, in human evaluation. |
Decoder Tuning: Efficient Language Understanding as Decoding (2023.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to adapt pre-trained models with parameters frozen are based on input-side adaptation, which requires thousands of API queries. |
| Approach: | They propose to train a model-as-a-service (MaaS) setting to provide only the inference APIs for users . they argue that input-side adaptation could be arduous due to the lack of gradient signals . |
| Outcome: | The proposed model outperforms state-of-the-art algorithms with a 200x speed-up. |
Model Interpretability and Rationale Extraction by Input Mask Optimization (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for creating explanations for black-box models struggle with deriving easily interpretable explanations. |
| Approach: | They propose a model-agnostic method to generate extractive explanations for neural network predictions using masking parts of the input that the model does not consider indicative of the respective class. |
| Outcome: | The proposed method achieves state-of-the-art results in a paragraph-level rationale extraction task, showing that this task can be performed without training a specialized model. |
DRIV-EX: Counterfactual Explanations for Driving LLMs (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) are increasingly used as reasoning engines in autonomous driving, yet their decision-making remains opaque. |
| Approach: | They propose to use gradient-based optimization on continuous embeddings to identify the input shifts required to flip a model’s decision. |
| Outcome: | The proposed method exposes latent biases and provides concrete insights to improve the robustness of LLM-based driving agents. |
OSCR-Attack: One-Shot Character Level Attacks through Self-Optimizing Continuous Relaxation (2026.findings-acl)
Copied to clipboard
Lingyi Kong, Zhuo Liu, Zhanghao Hu, Qilong Qiu, Yutao Yang, Jingjing Xue, Zheng Wang, Lin Gui, Feiping Nie
| Challenge: | Character-level adversarial attacks preserve semantics but are costly and inefficient . generative LLMs are gaining popularity due to their uncertainty and vulnerability to textual adversarials . |
| Approach: | They propose an end-to-end framework that transforms discrete choices into continuous representations and a conflict resolution strategy that maps them back into discrete insertion operations. |
| Outcome: | The proposed framework improves ASR by 21.45% points and accelerates the attack by 3.66 times compared to baselines. |
Explaining Differences Between Model Pairs in Natural Language through Sample Learning (2025.emnlp-main)
Copied to clipboard
| Challenge: | a framework that generates faithful natural language explanations of when and how two ML models converge or diverge in their predictions requires access to training data. |
| Approach: | They propose a framework that generates faithful natural language explanations of when and how two ML models converge or diverge in their predictions. |
| Outcome: | The proposed framework generates faithful natural language explanations of when and how two models diverge in their predictions. |