United States

Large Language Models (LLMs) possess vast amounts of knowledge within their parameters, prompting research into methods for locating and editing this knowledge. Previous work has largely focused on locating entity-related (often _single-token_) facts in smaller models. However, several key questions remain unanswered: (1) _How can we effectively locate query-relevant neurons in contemporary autoregressive LLMs, such as Llama and Mistral?_ (2) _How can we address the challenge of long-form text generation?_ (3) _Are there localized knowledge regions in LLMs?_ In this study, we introduce Query-Relevant Neuron Cluster Attribution (QRNCA), a novel architecture-agnostic framework capable of identifying query-relevant neurons in LLMs. QRNCA allows for the examination of long-form answers beyond triplet facts by employing the proxy task of multi-choice question answering. To evaluate the effectiveness of our detected neurons, we build two multi-choice QA datasets spanning diverse domains and languages. Empirical evaluations demonstrate that our method outperforms baseline methods significantly. Further, analysis of neuron distributions reveals the presence of visible localized regions, particularly within different domains. Finally, we show potential applications of our detected neurons in knowledge editing and neuron-based prediction.

AAAI 2025

Identifying Query-Relevant Neurons in Large Language Models for Long-Form Texts

interpretability analysis and evaluation of nlp models

snlp

## Welcome to EMNLP 2024! 
We are excited to welcome you to one of the most prominent conferences in the field of Natural Language Processing. This year, EMNLP 2024 is being held in a hybrid format,
offering both virtual and in-person participation in beautiful Miami. Due to a record-breaking number of submissions, we've expanded the total number of accepted papers to accommodate more cutting-edge research from around the globe.
### [Conference Handbook](https://drive.google.com/file/d/1WPROgxjLAC96AJL7Ugy0tEnYm7dkrbHt/view?usp=sharing)

You are required to register for this event. **Please register [here](https://2024.emnlp.org/registration/).** The EMNLP 2024 event page on Underline will be open to public one week prior to the event.

Please register!

EMNLP 2024

EMNLP 2024 will take place in Miami, Florida from Nov 12th to Nov 16th, 2024, at the Hyatt Regency Miami Hote and on Underline for remote participants.

technical paper

We are pleased to announce the Thirty-Ninth AAAI Conference on Artificial Intelligence (AAAI-25), which will be held in Philadelphia, Pennsylvania at the Pennsylvania Convention Center from February 25 to March 4, 2025.

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. AAAI-25 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and a range of other activities to be announced.

### [Invited Speakers](https://aaai.org/conference/aaai/aaai-25/aaai-25-invited-speakers/)

Register [here](https://aaai.org/conference/aaai/aaai-25/registration/)

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. AAAI-25 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and a range of other activities to be announced.



Multi-Agent Path Finding (MAPF) is the problem of planning collision-free paths for multiple agents. Unfortunately, during the execution of a MAPF plan, agents may suffer from unexpected delays. Given such delays, Switchable Temporal Plan Graph (STPG) provides a framework to search for an acyclic Temporal Plan Graph (TPG) with the minimum execution cost, which guarantees deadlock- and collision-free execution. However, existing optimal algorithms like Mixed-Integer Linear Programming (MILP) or Graph-Based Switchable Edge Search (GSES) are too slow. This paper speeds up GSES with four techniques: stronger admissible heuristics, edge grouping, prioritized branching, and incremental implementation. Experimenting on 4 different types of maps with varying numbers of agents, our method, Improved GSES (IGSES), consistently achieves more than double the success rates of GSES and demonstrates up to a 30-fold speedup on instances where both methods find solutions.

Speedup Techniques for Switchable Temporal Plan Graph Optimization

Cross-domain recommendation (CDR) mitigates data sparsity and cold-start issues in recommendation systems. While recent CDR approaches using graph neural networks (GNNs) capture complex user-item interactions, they rely on manually designed architectures that are often suboptimal and labor-intensive. Additionally, extracting valuable behavioral information from source domains to improve target domain recommendations remains challenging. To address these challenges, we propose Behavior importance-aware Graph Neural Architecture Search (BiGNAS), a framework that jointly optimizes GNN architecture and data importance for CDR. BiGNAS introduces two key components: a Cross-Domain Customized Supernetwork and a Graph-Based Behavior Importance Perceptron. The supernetwork, as a one-shot, retrain-free module, automatically searches the optimal GNN architecture for each domain without the need for retraining. The perceptron uses auxiliary learning to dynamically assess the importance of source domain behaviors, thereby improving target domain recommendations. Extensive experiments on benchmark CDR datasets and a large-scale industry advertising dataset demonstrate that BiGNAS consistently outperforms state-of-the-art baselines. To the best of our knowledge, this is the first work to jointly optimize GNN architecture and behavior data importance for cross-domain recommendation.

Behavior Importance-Aware Graph Neural Architecture Search for Cross-Domain Recommendation

In this paper, we introduce a new family of argument-ranking semantics which can be seen as a refinement of the classification of arguments into skeptically accepted, credulously accepted and rejected. To this end we use so-called social ranking functions which have been developed recently to rank individuals based on their performance in groups. We provide necessary and sufficient conditions for a social ranking function to give rise to an argument-ranking semantics satisfying the desired refinement property.

An Extension-Based Argument-Ranking Semantics: Social Rankings in Abstract Argumentation

Physics-informed neural networks solve partial differential equations by training neural networks. Since this method approximates infinite-dimensional PDE solutions with finite collocation points, minimizing discretization errors by selecting suitable points is essential for accelerating the learning process. Inspired by number theoretic methods for numerical analysis, we introduce good lattice training and periodization tricks, which ensure the conditions required by the theory. Our experiments demonstrate that GLT requires 2--7 times fewer collocation points, resulting in lower computational cost, while achieving competitive performance compared to typical sampling methods.

Number Theoretic Accelerated Learning of Physics-Informed Neural Networks

Recent advancements in AI alignment techniques have significantly improved the alignment of large language models (LLMs) with static human preferences. However, the dynamic nature of human preferences can render some prior training data outdated or even erroneous, ultimately causing LLMs to deviate from contemporary human preferences and societal norms. Existing methodologies, whether they involve the curation of new data for continual alignment or the manual correction of outdated data for re-alignment, demand costly human resources.
To address this challenge, we propose a novel approach, Large Language Model Behavior Correction with Influence Function Recall and Post-Training (LANCET), which requires no human involvement. LANCET consists of two phases: (1) using influence functions to identify the training data that significantly impact undesirable model outputs, and (2) applying an Influence function-driven Bregman Optimization (IBO) technique to adjust the model’s behavior based on these influence distributions. Our experiments demonstrate that LANCET effectively and efficiently correct inappropriate behaviors of LLMs. Furthermore, LANCET can outperform methods that rely on collecting human preferences, and it enhances the interpretability of learning human preferences within LLMs.

Correcting Large Language Model Behavior via Influence Function

Human experts typically integrate numerical and textual multimodal information to analyze time series. However, most traditional deep learning predictors rely solely on unimodal numerical data, using a fixed-length window for training and prediction on a single dataset, and cannot adapt to different scenarios. The powered pre-trained large language model has introduced new opportunities for time series analysis. Yet, existing methods are either inefficient in training, incapable of handling textual information, or lack zero-shot forecasting capability. In this paper, we innovatively model time series as a foreign language and construct ChatTime, a unified framework for time series and text processing. As an out-of-the-box multimodal time series foundation model, ChatTime provides zero-shot forecasting capability and supports bimodal input/output for both time series and text. We design a series of experiments to verify the superior performance of ChatTime across multiple tasks and scenarios, and create four multimodal datasets to address data gaps. The experimental results demonstrate the potential and utility of ChatTime. Specifically, in the unimodal zero-shot forecasting task, ChatTime achieves 99.9\% of the previous state-of-the-art using only 4\% of the pre-training data. In the multimodal tasks of context-guided forecasting and time series question answering, ChatTime outperforms the previous state-of-the-art by 3.7\% and 36.6\%, respectively.

ChatTime: A Unified Multimodal Time Series Foundation Model Bridging Numerical and Textual Data

Equilibrium problems in Bayesian auction games can be described as systems of differential equations. Depending on the model assumptions, these equations might be such that we do not have a rigorous mathematical solution theory. The lack of analytical or numerical techniques with guaranteed convergence for the equilibrium problem has plagued the field and limited equilibrium analysis to rather simple auction models such as single-object auctions. 
Recent advances in equilibrium learning led to algorithms that find equilibrium under a wide variety of model assumptions. Monotonicity and the Minty condition are the known sufficient conditions for learning algorithms to converge to an equilibrium in games. 
Not much is known about convergence of learning algorithms beyond these conditions. 
We analyze first- and second-price auctions where simple learning algorithms consistently converge to an equilibrium.
The analysis is challenging, because these properties need to be shown in infinite dimensions. 
Interestingly, we show that neither monotonicity nor pseudo- or quasi-monotonicity holds for the respective variational inequalities (VIs). 
The second-price auction's equilibrium is a Minty-type solution, but the first-price auction is not. 
However, the analysis via infinite-dimensional VIs allows us to get ex-post guarantees for gradient-based algorithms. 
We show that the Bayes--Nash equilibrium is the unique solution to the VI within the class of uniformly increasing bid functions, which ensures that gradient-based algorithms attain the equilibrium in case of convergence, as also observed in numerical experiments.

Beyond Monotonicity: On the Convergence of Learning Algorithms in Standard Auction Games

Initial discussion of AI literacy assessment has focused on competency frameworks and learning standards rather than materials for classroom use. Responsible AI for Computational Action (RAICA), a constructionist AI curriculum for middle and high school students, includes assessment materials to support teachers with the evaluation of student AI fluency competencies in their classrooms. These materials include exit tickets used as formative assessments at the end of each lesson and both teacher and student-facing rubrics. After beta-testing a module of the curriculum with nine teachers and 282 students, we reviewed teacher usage data and feedback as well as student responses. The review process surfaced a number of improvements to the materials to better align them with classroom teaching practice. These included clarifying language and adding visual scaffolds. We present the assessment materials and iterative design process used to bridge the gap between the theoretical AI literacy competencies and their practical implementation in classrooms.

Supporting AI Fluency Teaching Through the Development of Assessments for Classroom Use

The rapid and nearly pervasive impact of artificial intelligence on fields as diverse as medicine, law, banking, and the arts has made many students who would never enroll in a computer science class become interested in understanding elements of artificial intelligence. Fueled by questions about how this technology would change their own fields, these students are not seeking to become experts in building AI systems but instead are searching for a sufficient understanding to be safe, effective, and informed users. In this paper, we describe a first-of-its-kind course offering, "Artificial Intelligence for Future Presidents" designed and taught during the spring of 2024. We share rationale on the design and structure of the course, consider how best to convey complex technical information to students without the background in programming or mathematics, and consider methods for supporting an understanding of the limits of this technology.

EAAI-127 - Artificial Intelligence for Future Presidents: Teaching AI Literacy to Everyone

Spatial reasoning is a core component of an agent’s ability to operate in or reason about the physical world. LLMs are widely promoted as having abilities to reason about a wide variety of domains, including commonsense. In this talk I will discuss the ability of state-of-the-art LLMs to perform commonsense reasoning, particularly with regard to spatial information. Across a wide range of LLMs, although they show abilities rather better than chance, they still struggle with many questions and tasks, for example when reasoning about directions, or topological relations.

Downloads

Next from AAAI 2025

Speedup Techniques for Switchable Temporal Plan Graph Optimization

Similar lecture

Neuron-Level Knowledge Attribution in Large Language Models

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from AAAI 2025

Speedup Techniques for Switchable Temporal Plan Graph Optimization

Similar lecture

Neuron-Level Knowledge Attribution in Large Language Models

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Downloads