United States

Prompt instruction tuning is a popular approach to better adjust monolithic, pretrained LLMs for downstream tasks. Yet, how to extend this approach to handle multiple tasks and data distributions is an open question.  To address this gap, we propose using \emph{Mixture of Prompts} (MoPs) with smart gating functionality. The latter identifies relevant skills embedded in different groups of prompts and dynamically weighs experts (i.e., collection of prompts) based on the target task. Experiments show that MoPs are resilient to model compression, data source, and task composition, making them highly versatile and applicable in various contexts. 
In practice, MoPs can simultaneously mitigate prompt training ``interference&#39;&#39; in multi-task, multi-source scenarios (e.g., task and data heterogeneity across sources) and possible implications from model approximations. Empirically, MoPs can reduce final perplexity from 9\% up to 70\% in federated cases and from 3\% up to 30\% in centralized cases, compared to baselines.

AAAI 2025

Sweeping Heterogeneity with Smart MoPs: Mixture of Prompts for LLM Task Adaptation

Prompt instruction tuning is a popular approach to better adjust monolithic, pretrained LLMs for downstream tasks. Yet, how to extend this approach to handle multiple tasks and data distributions is an open question.  To address this gap, we propose using \emph{Mixture of Prompts} (MoPs) with smart gating functionality. The latter identifies relevant skills embedded in different groups of prompts and dynamically weighs experts (i.e., collection of prompts) based on the target task. Experiments show that MoPs are resilient to model compression, data source, and task composition, making them highly versatile and applicable in various contexts. 
In practice, MoPs can simultaneously mitigate prompt training ``interference'' in multi-task, multi-source scenarios (e.g., task and data heterogeneity across sources) and possible implications from model approximations. Empirically, MoPs can reduce final perplexity from 9\% up to 70\% in federated cases and from 3\% up to 30\% in centralized cases, compared to baselines.

poster

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.



We consider the Coalition Structure Learning (CSL) problem in multi-agent systems, motivated by the existence of coalitions in many real-world systems, e.g., trading platforms and auction systems. In this problem, there is a hidden coalition structure within a set of $n$ agents, which affects the behavior of the agents in games. Our goal is to actively design a sequence of games for the agents to play, such that observations in these games can be used to learn the hidden coalition structure. In particular, we consider the setting where in each round, we design and present a game together with a strategy profile to the agents, and receive a multiple-bit observation -- for each agent, we observe whether or not they would like to deviate from the specified strategy. We show that we can learn the coalition structure in $O(\log n)$ rounds if we are allowed to design any normal-form game, matching the information-theoretical lower bound. For practicality, we extend the result to settings where we can only choose games of a specific format, and design algorithms to learn the coalition structure in these settings. For most settings, our complexity matches the theoretical lower bound up to a constant factor.

Deviate or Not: Learning Coalition Structures with Multiple-bit Observations in Games

The purpose of partial multi-label feature selection is to select the most representative feature subset, where the data comes from partial multi-label datasets that have label ambiguity issues. For label disambiguation, previous methods mainly focus on utilizing the information inside the labels and the relationship between the labels and features. However, the information existing in the feature space is rarely considered, especially in partial multi-label scenarios where the noises is considered to be concentrated in the label space while the feature information is correct. This paper proposes a method based on  latent space alignment, which uses the information mined in feature space to disambiguate in latent space through the structural consistency between labels and features. In addition, previous methods overestimate the consistency of features and labels in the latent space after convergence. We comprehensively consider the similarity of latent space projections to feature space and label space, and propose new feature selection term. This method also significantly improves the positive label identification ability of the selected features. Comprehensive experiments demonstrate the superiority of the proposed method.

Reconsidering Feature Structure Information and Latent Space Alignment in Partial Multi-label Feature Selection

Density Functional Theory (DFT) stands as a widely used and efficient approach for addressing the many-electron Schrödinger equation across various domains such as physics, chemistry, and biology. However, a core challenge that persists over the long term pertains to refining the exchange-correlation (XC) approximation. This approximation significantly influences the triumphs and shortcomings observed in DFT applications. Nonetheless, a prevalent issue among XC approximations is the presence of systematic errors, stemming from deviations from the mathematical properties of the exact XC functional. For example, although both B3LYP and DM21 (DeepMind 21) exhibit improvements over previous benchmarks, there is still potential for further refinement. In this paper, we propose a strategy for enhancing XC approximations by estimating the neural uncertainty of the XC functional, named Residual XC-Uncertain Functional. Specifically, our approach involves training a neural network to predict both the mean and variance of the XC functional, treating it as a Gaussian distribution. To ensure stability in each sampling point, we construct the mean by combining traditional XC approximations with our neural predictions, mitigating the risk of divergence or vanishing values. It is crucial to highlight that our methodology excels particularly in cases where systematic errors are pronounced. Empirical outcomes from three benchmark tests substantiate the superiority of our approach over existing state-of-the-art methods. Our approach not only surpasses related techniques but also significantly outperforms both the popular B3LYP and the recent DM21 methods, achieving average RMSE improvements of 62\% and 37\%, respectively, across the three benchmarks: W4-17, G21EA, and G21IP.

Learning Generalized Residual Exchange-Correlation-Uncertain Functional for Density Functional Theory

Abstraction is key to scaling up reinforcement learning (RL). However, autonomously learning abstract representations and temporally extended actions, like options, to improve transfer and generalization remains a challenging open problem. This paper presents a novel approach for inventing, representing, and utilizing options in continual RL settings to address streams of stochastic problems with long-horizons, sparse rewards, and unknown transition and reward functions. 

Our approach continually learns and maintains an interpretable state abstraction, and uses it to invent high-level options with abstract symbolic representations. These options meet three key desiderata: (1) reusability across problem instances, (2) mutual independence for minimizing interference, and (3) composability for solving new tasks effectively. Our main contributions are approaches for continually learning transferable, generalizable options with symbolic representations, and for integrating search techniques with RL to efficiently plan over these learned options to solve new problems. Empirical results demonstrate that the resulting approach effectively learns and transfers abstract knowledge across problem instances, achieving superior sample efficiency compared to state-of-the-art methods.

Autonomous Option Invention for Continual Hierarchical Reinforcement Learning and Planning

We address an advanced challenge of predicting pedestrian occupancy as an extension of multi-view pedestrian detection in urban traffic.
To support this, we have created a new synthetic dataset called MVP-Occ, designed for dense pedestrian scenarios in large-scale scenes.
Our dataset provides detailed representations of pedestrians using voxel structures, accompanied by rich semantic scene understanding labels, facilitating visual navigation and insights into pedestrian spatial information.
Furthermore, we present a robust baseline model, termed OmniOcc, capable of predicting both the voxel occupancy state and panoptic labels for the entire scene from multi-view images.
Through in-depth analysis, we identify and evaluate the key elements of our proposed model, highlighting their specific contributions and importance.

Multi-View Pedestrian Occupancy Prediction with a Novel Synthetic Dataset

Despite the rapid development of Chinese vision-language models (VLMs), most existing Chinese vision-language (VL) datasets are constructed on Western-centric images from existing English VL datasets. The cultural bias in the images makes these datasets unsuitable for evaluating VLMs in Chinese culture. To remedy this issue, we present a new Chinese Vision-Language Understanding Evaluation (CVLUE) benchmark dataset, where the selection of object categories and images is entirely driven by Chinese native speakers, ensuring that the source images are representative of Chinese culture. The benchmark contains four distinct VL tasks ranging from image-text retrieval to visual question answering, visual grounding and visual dialogue. We present a detailed statistical analysis of CVLUE and provide a baseline performance analysis with several open-source multilingual VLMs on CVLUE and its English counterparts to reveal their performance gap between English and Chinese. Our in-depth category-level analysis reveals a lack of Chinese cultural knowledge in existing VLMs. We also find that fine-tuning on Chinese culture-related VL datasets effectively enhances VLMs' understanding of Chinese culture.

CVLUE: A New Benchmark Dataset for Chinese Vision-Language Understanding Evaluation

Long-term series forecasting aims to predict future data over long horizons based on historical information. However, existing methods struggle to effectively utilize long lookback windows due to overfitting, computational resource constraints, or information extraction challenges, thereby limiting them to using limited lookback windows for predicting long-term future series. To address these issues, this paper introduces the Input Refinement and Prediction Auxiliary (IRPA) framework, a lightweight model consisting of four linear layers designed to extract key information from ultra-long lookback windows to enhance limited lookback windows and assist prediction processes. IRPA comprises an Input Refinement Module (IRM) and a Prediction Auxiliary Module (PAM), each constructed from two linear layer sub-modules. The IRM performs effective decomposition and patching of ultra-long sequences, refining seasonal and trend features to increase the information density in limited lookback windows and mitigate overfitting and parameter inflation. The PAM extracts historical similarities and seasonal patterns from ultra-long lookback windows to significantly improve prediction accuracy. IRPA substantially extends the utilization of lookback windows, offering a lightweight and efficient solution with broad applicability. Experimental results on eight datasets show IRPA reduces the Mean Squared Error (MSE) by an average of 16.1% for various models.

Efficiently Enhancing Long-term Series Forecasting via Ultra-long Lookback Windows

Transformer-based Super-Resolution (SR) methods have surpassed CNN-based SR approaches due to their ability to capture long-range dependencies through self-attention layers. However, transformers' high computational complexity has increased the demand for lightweight approaches to enhance their practicality. To address this, we introduce the Attention-Sharing Information Distillation (ASID) network, a lightweight SR network that employs attention-sharing and an information distillation structure tailored for transformer-based SR methods. We adopt the information distillation scheme, originally designed for efficient CNN operations, to reduce the computational load of stacked self-attention layers, thereby alleviating the efficiency bottleneck. Furthermore, to further reduce the computational demands primarily imposed by self-attention operations, we implement attention-sharing across blocks, significantly lowering overall computational needs. By integrating these strategies, ASID delivers competitive performance compared to existing SR methods while requiring only about 300K parameters—substantially fewer than existing CNN-based and transformer-based SR models.

Efficient Attention-Sharing Information Distillation Transformer for Lightweight Single Image Super-Resolution

Neural Radiance Fields (NeRF) have achieved huge success in effectively capturing and representing 3D objects and scenes. However, to establish a ubiquitous presence in everyday media formats, such as images and videos, we need to fulfill three key objectives: 1. fast encoding and decoding time, 2. compact model sizes, and 3. high-quality renderings. Despite recent advancements, a comprehensive algorithm that adequately addresses all objectives has yet to be fully realized. In this work, we present CodecNeRF, a neural codec for NeRF representations, consisting of an encoder and decoder architecture that can generate a NeRF representation in a single forward pass. Furthermore, inspired by the recent parameter-efficient finetuning approaches, we propose a finetuning method to efficiently adapt the generated NeRF representations to a new test instance, leading to high-quality image renderings and compact code sizes. The proposed CodecNeRF, a newly suggested encoding-decoding-finetuning pipeline for NeRF, achieved unprecedented compression performance of more than 100x and remarkable reduction in encoding time while maintaining (or improving) the image quality on widely used 3D object datasets.

CodecNeRF: Toward Fast Encoding and Decoding, Compact, and High-quality Novel-view Synthesis

The labelling difficulty has been a longstanding problem in deep image matting. To escape from fine labels, this work explores using rough annotations such as trimaps coarsely indicating the foreground/background as supervision. We present that the cooperation between learned semantics from indicated known regions and proper assumed matting rules can help infer alpha values at transition areas. Inspired by the nonlocal principle in traditional image matting, we build a directional distance consistency loss (DDC loss) at each pixel neighborhood to constrain the alpha values conditioned on the input image. DDC loss forces the distance of similar pairs on the alpha matte and on its corresponding image to be consistent. In this way, the alpha values can be propagated from learned known regions to unknown transition areas. With only images and trimaps, a matting model can be trained under the supervision of a known loss and the proposed DDC loss. Experiments on AM-2K and P3M-10K dataset show that our paradigm achieves comparable performance with the fine-label-supervised baseline, while sometimes offers even more satisfying results than human-labelled ground truth. Code will be available.

Sweeping Heterogeneity with Smart MoPs: Mixture of Prompts for LLM Task Adaptation

Downloads

Next from AAAI 2025

Deviate or Not: Learning Coalition Structures with Multiple-bit Observations in Games