David H. Yang, Yuxuan Zhu, Mohammad Mohammadi Amiri, Keerthiram Murugesan, Tejaswini Pedapati, Subhajit Chaudhury, Pin-Yu Chen. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

We present EvalAssist, a framework that simplifies the LLM- as-a-judge workflow. The system provides an online criteria development environment, where users can interactively build, test, and share custom evaluation criteria in a structured and portable format. A library of LLM based evaluators is made available that incorporates various algorithmic innovations such as token-probability based judgement, positional bias checking, and certainty estimation that help to engender trust in the evaluation process. We have computed extensive benchmarks and also deployed the system internally in our organization with several hundreds of users.

Pretrained Language Models (PLMs) have become the de facto starting point for fine-tuning on downstream tasks. However, as model sizes continue to increase, traditional fine-tuning of all parameters becomes challenging. To address this, parameter-efficient fine-tuning (PEFT) methods have gained popularity as a means to adapt PLMs effectively. In parallel, recent studies have revealed the presence of activation sparsity within the intermediate outputs of the multilayer perceptron (MLP) blocks in transformers. Low activation density enables efficient model inference on sparsity-aware hardware. Building upon this insight, in this work, we propose a novel density loss that encourages higher activation sparsity (equivalently, lower activation density) in the pre-trained models. We demonstrate the effectiveness of our approach by utilizing mainstream PEFT techniques, including QLoRA, LoRA, Adapter, and Prompt/Prefix Tuning, to facilitate efficient model adaptation across diverse downstream tasks. Experiments show that our proposed method, DEFT (Density-Efficient Fine-Tuning), can consistently reduce activation density by up to 44.94% on RoBERTa (Large) and by 53.19 (encoder density) and 90.60% (decoder density) on Flan-T5-XXL (11B) compared to PEFT, using GLUE and QA (SQuAD) benchmarks respectively while maintaining competitive performance on downstream tasks. We also introduce ADA-DEFT, an adaptive variant of our DEFT approach, which achieves significant memory and runtime savings during inference for large models. For instance, ADA-DEFT reduces runtime by 8.75% and memory usage by 16.78% in Flan-T5-XL and by 2.79% and 2.54%, respectively, in Flan-T5- XXL. Additionally, we showcase that DEFT works complementarily with quantized and pruned models.

Recent advances in Large Language Models (LLMs) have yielded impressive successes on many language tasks. However, efficient processing of long contexts using LLMs remains a significant challenge. We introduce EpMAN – a method for processing long contexts in an episodic memory module while holistically attending to semantically-relevant context chunks. Output from episodic attention is then used to reweigh the decoder’s self-attention to the stored KV cache of the context during training and generation. When an LLM decoder is trained using EpMAN, its performance on multiple challenging single-hop long-context recall and question-answering benchmarks is found to be stronger and more robust across the range from 16k to 256k tokens than baseline decoders trained with self-attention, and popular retrieval-augmented generation frameworks.

Pre-trained vision-language models are able to interpret visual concepts and language semantics. Prompt learning, a method of constructing prompts for text encoders or image encoders, elicits the potentials of pre-trained models and readily adapts them to new scenarios. Compared to fine-tuning, prompt learning enables the model to achieve comparable or better performance using fewer trainable parameters. Besides, prompt learning freezes the pre-trained model and avoids the catastrophic forgetting issue in the fine-tuning. Continuous prompts inserted into the input of every transformer layer (i.e. deep prompts) can improve the performances of pre-trained models on downstream tasks. For i-th transformer layer, the inserted prompts replace previously inserted prompts in the (i − 1)-th layer. Although the self-attention mechanism contextualizes newly inserted prompts for the current layer and embeddings from the previous layer’s output, removing all inserted prompts from the previous layer inevitably loses information contained in the continuous prompts. In this work, we propose Modular Prompt Learning (MPL) that is designed to promote the preservation of information contained in the inserted prompts. We evaluate the proposed method on base-to-new generalization and cross-dataset tasks. On average of 11 datasets, our method achieves 0.7% performance gain on the base-to-new generalization task compared to the state-of-the-art method. The largest improvement on the individual dataset is 10.7% (EuroSAT dataset). Our code is available at https://github.com/Zhenhan-Huang/Modular-Prompt-Learning.

Enterprises have a growing need to identify relevant tables in data lakes; e.g. tables that are unionable, joinable, or subsets of each other. Tabular neural models can be help-ful for such data discovery tasks. In this paper, we present TabSketchFM, a neural tabular model for data discovery over data lakes. First, we propose novel pre-training: a sketch-based approach to enhance the effectiveness of data discovery in neural tabular models. Second, we finetune the pretrained model for identifying unionable, joinable, and subset table pairs and show significant improvement over previous tabular neural models. Third, we present a detailed ablation study to highlight which sketches are crucial for which tasks. Fourth, we use these finetuned models to perform table search; i.e., given a query table, find other tables in a corpus that are unionable, joinable, or that are subsets of the query. Our results demonstrate significant improvements in F1 scores for search compared to state-of-the-art techniques. Finally, we show significant transfer across datasets and tasks establishing that our model can generalize across different tasks and over different data lakes.

Reducing the likelihood of generating harmful and toxic output is an essential task when aligning large language models (LLMs). Existing methods mainly rely on training an external reward model (i.e., another language model) or fine-tuning the LLM using self-generated data to influence the outcome. In this paper, we show that LLMs have the capability of self-detoxification without external reward model learning or retraining of the LM. We propose \textit{Self-disciplined Autoregressive Sampling (SASA)}, a lightweight controlled decoding algorithm for toxicity reduction of LLMs. SASA leverages the contextual representations from an LLM to learn linear subspaces from labeled data characterizing toxic v.s. non-toxic output in analytical forms. When auto-completing a response token-by-token, SASA dynamically tracks the margin of the current output to steer the generation away from the toxic subspace, by adjusting the autoregressive sampling strategy. Evaluated on LLMs of different scale and nature, namely Llama-3.1-Instruct (8B), Llama-2 (7B), and GPT2-L models with the RealToxicityPrompts, BOLD, and AttaQ benchmarks, SASA markedly enhances the quality of the generated sentences relative to the original models and attains comparable performance to state-of-the-art detoxification techniques, significantly reducing the toxicity level by only using the LLM's internal representations.

Prompt learning is an effective way to exploit the potential of large-scale pre-trained foundational models. Continuous prompts parameterize context tokens in prompts by turning them into differentiable vectors. Deep continuous prompts insert prompts not only in the input but also in the intermediate hidden representations. Manually designed deep continuous prompts exhibit a remarkable improvement compared to the zero-shot pre-trained model on downstream tasks. How to automate the continuous prompt design is an underexplored area, and a fundamental question arises, is manually designed deep prompt strategy optimal? To answer this question, we propose a method dubbed differentiable prompt learning (DPL). The DPL method is formulated as an optimization problem to automatically determine the optimal context length of the prompt to be added to each layer, where the objective is to maximize the performance. We test the DPL method on the pre-trained CLIP. We empirically find that by using only limited data, our DPL method can find deep continuous prompt configuration with high confidence. The performance on the downstream tasks exhibits the superiority of the automatic design: our method boosts the average test accuracy by 2.60% on 11 datasets compared to baseline methods. Besides, our method focuses only on the prompt configuration (i.e. context length for each layer), which means that our method is compatible with the baseline methods that have sophisticated designs to boost the performance. We release our code in https://github.com/Zhenhan-Huang/Differentiable-Prompt-Learn.

User flow diagram for EvalAssist in the direct assessment evaluation, illustrating criteria definition, test data input, annotation, AI evaluator selection, result review, iterative adjustments, and criteria export for dataset-wide evaluation via SDK.

Building models for natural language processing (NLP) tasks remains a daunting task for many, requiring significant technical expertise, efforts, and resources. In this demonstration, we present AutoText, an end-to-end AutoAI framework for text, to lower the barrier of entry in building NLP models. AutoText combines state-of-the-art AutoAI optimization techniques and learning algorithms for NLP tasks into a single extensible framework. Through its simple, yet powerful UI, non-AI experts (e.g., domain experts) can quickly generate performant NLP models with support to both control (e.g., via specifying constraints) and understand learned models.

In recent years there has been a considerable amount of research on local post hoc explanations for neural networks. However, work on building interpretable neural architectures has been relatively sparse. In this paper, we present a novel neural architecture, CoFrNet, inspired by the form of continued fractions which are known to have many attractive properties in number theory, such as fast convergence of approximations to real numbers. We show that CoFrNets can be efficiently trained as well as interpreted leveraging their particular functional form. Moreover, we prove that such architectures are universal approximators based on a proof strategy that is different than the typical strategy used to prove universal approximation results for neural networks based on infinite width (or depth), which is likely to be of independent interest. We experiment on nonlinear synthetic functions and are able to accurately model as well as estimate feature attributions and even higher order terms in some cases, which is a testament to the representational power as well as interpretability of such architectures. To further showcase the power of CoFrNets, we experiment on seven real datasets spanning tabular, text and image modalities, and show that they are either comparable or significantly better than other interpretable models and multilayer perceptrons, sometimes approaching the accuracies of state-of-the-art models.

Many automated machine learning methods, such as those for hyperparameter and neural architecture optimization, are computationally expensive because they involve training many different model configurations. In this work, we present a new method that saves computational budget by terminating poor configurations early on in the training. In contrast to existing methods, we consider this task as a ranking and transfer learning problem. We qualitatively show that by optimizing a pairwise ranking loss and leveraging learning curves from other data sets, our model is able to effectively rank learning curves without having to observe many or very long learning curves. We further demonstrate that our method can be used to accelerate a neural architecture search by a factor of up to 100 without a significant performance degradation of the discovered architecture. In further experiments we analyze the quality of ranking, the influence of different model components as well as the predictive behavior of the model.

Data science is labor-intensive and human experts are scarce but heavily involved in every aspect of it. This makes data science time consuming and restricted to experts with the resulting quality heavily dependent on their experience and skills. To make data science more accessible and scalable, we need its democratization. Automated Data Science (AutoDS) is aimed towards that goal and is emerging as an important research and business topic. We introduce and define the AutoDS challenge, followed by a proposal of a general AutoDS framework that covers existing approaches but also provides guidance for the development of new methods. We categorize and review the existing literature from multiple aspects of the problem setup and employed techniques. Then we provide several views on how AI could succeed in automating end-to-end AutoDS. We hope this survey can serve as insightful guideline for the AutoDS field and provide inspiration for future research.

The growing interest in both the automation of machine learning and deep learning has inevitably led to the development of a wide variety of methods to automate deep learning. The choice of network architecture has proven critical, and many improvements in deep learning are due to new structuring of it. However, deep learning techniques are computationally intensive and their use requires a high level of domain knowledge. Even a partial automation of this process therefore helps to make deep learning more accessible for everyone. In this tutorial we present a uniform formalism that enables different methods to be categorized and compare the different approaches in terms of their performance. We achieve this through a comprehensive discussion of the commonly used architecture search spaces and architecture optimization algorithms based on reinforcement learning and evolutionary algorithms as well as approaches that include surrogate and one-shot models. In addition, we discuss approaches to accelerate the search for neural architectures based on early termination and transfer learning and address the new research directions, which include constrained and multi-objective architecture search as well as the automated search for data augmentation, optimizers, and activation functions.

There is a rich and growing literature on producing local contrastive/counterfactual explanations for black-box models (e.g. neural networks). In these methods, for an input, an explanation is in the form of a contrast point differing in very few features from the original input and lying in a different class. Other works try to build globally interpretable models like decision trees and rule lists based on the data using actual labels or based on the black-box models predictions. Although these interpretable global models can be useful, they may not be consistent with local explanations from a specific black-box of choice. In this work, we explore the question: Can we produce a transparent global model that is simultaneously accurate and consistent with the local (contrastive) explanations of the black-box model? We introduce a local consistency metric that quantifies if the local explanations for the black-box model are also applicable to the proxy/surrogate globally transparent model. Based on a key insight we propose a novel method where we create custom boolean features from local contrastive explanations of the black-box model and then train a globally transparent model that has higher local consistency compared with other known strategies in addition to being accurate.

We present DATASET EVOLVER, an interactive Jupyter notebook-based tool to support data scientists perform feature engineering for classification tasks. It provides users with suggestions on new features to construct, based on automated feature engineering algorithms. Users can navigate the given choices in different ways, validate the impact, and selectively accept the suggestions. DATASET EVOLVER is a pluggable feature engineering framework where several exploration strategies could be added. It currently includes meta-learning based exploration and reinforcement learning based exploration. The suggested features are constructed using well-defined mathematical functions and are easily interpretable. Our system provides a mixed-initiative system of a user being assisted by an automated agent to efficiently and effectively solve the complex problem of feature engineering. It reduces the effort of a data scientist from hours to minutes.

Current tools for exploratory data analysis (EDA) require users to manually select data attributes, statistical computations and visual encodings. This can be daunting for large-scale, complex data. We introduce Foresight, a system that helps the user rapidly discover visual insights from large high-dimensional datasets. Formally, an "insight" is a strong manifestation of a statistical property of the data, e.g., high correlation between two attributes, high skewness or concentration about the mean of a single attribute, a strong clustering of values, and so on. For each insight type, Foresight initially presents visualizations of the top k instances in the data, based on an appropriate ranking metric. The user can then look at "nearby" insights by issuing "insight queries" containing constraints on insight strengths and data attributes. Thus the user can directly explore the space of insights, rather than the space of data dimensions and visual encodings as in other visual recommender systems. Foresight also provides "global" views of insight space to help orient the user and ensure a thorough exploration process. Furthermore, Foresight facilitates interactive exploration of large datasets through fast, approximate sketching.