United States

Offsite-tuning is a privacy-preserving method for tuning large language models (LLMs) by sharing a lossy compressed emulator from the LLM owners with data owners for downstream task tuning. This approach protects the privacy of both the model and data owners. However, current offsite tuning methods often suffer from adaptation degradation, high computational costs, and limited protection strength due to uniformly dropping LLM layers or relying on expensive knowledge distillation. To address these issues, we propose ScaleOT, a novel privacy-utility-scalable offsite-tuning framework that effectively balances privacy and utility. ScaleOT introduces a novel layerwise lossy compression algorithm that uses reinforcement learning to obtain the importance of each layer. It employs lightweight networks, termed harmonizers, to replace the raw LLM layers. By combining important original LLM layers and harmonizers in different ratios, ScaleOT generates emulators tailored for optimal performance with various model scales for enhanced privacy protection. Additionally, we present a rank reduction method to further compress the original LLM layers, significantly enhancing privacy with negligible impact on utility. Comprehensive experiments show that ScaleOT can achieve nearly lossless offsite tuning performance compared with full fine-tuning while obtaining better model privacy.

AAAI 2025

ScaleOT: Privacy-utility-scalable Offsite-tuning with Dynamic LayerReplace and Selective Rank Compression

snlp

transparency

ethics

fairness

bias

privacy

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.



Integrating metric time into Task And Motion Planning (TAMP) is challenging, especially when multiple objects move simultaneously. Current literature focuses on classical and numeric TAMP, not considering deadlines, motions overlapping in time, and other temporal constraints.
In this paper, we fill this gap by formalizing the problem of Temporal Task and Motion Planning (TTAMP) for multiple object navigation. We propose a novel interleaved planning technique for this problem, which leverages incremental Satisfiability Modulo Theory to ensure efficient reasoning on deadlines and action duration coupled with a motion planner supporting simultaneous object motion. Geometric data on encountered obstacles prunes unreachable symbolic regions, while temporal bounds limit the geometric search space. For multiple moving objects, our algorithm contextualizes the conflicts learned from the motion planner on overlapping actions so that entire classes of temporal plans are pruned from the search space of the task planner, ensuring the eventual termination of the interplay. We provide a comprehensive benchmark suite and demonstrate the effectiveness of our solver in these scenarios.

Temporal Task and Motion Planning with Metric Time for Multiple Object Navigation

In this paper, we consider the problem of fair division of indivisible goods, where the allocation of goods impacts society. Specifically, we introduce a second valuation function for each agent, which determines the social impact of allocating a good to the agent. Such impact is considered desirable for the society -- the higher, the better. Our goal is to understand how to allocate goods fairly from the agents' perspective while maintaining society as happy as possible. To this end, we measure the impact on society using the utilitarian social welfare, and provide both possibility and impossibility results.
Our findings reveal that achieving good approximations, better than linear in the number of agents, is not possible while ensuring fairness to the agents. These impossibility results can be attributed to the fact that agents are completely unconscious of their social impact. 
Consequently, we explore scenarios where agents are socially aware, by introducing related fairness notions, and demonstrate that an appropriate definition of fairness is compatible with the social objective.

Fair Division with Social Impact

This study investigates scheduling strategies for the stochastic resource-constrained project scheduling problem with maximal time lags (SRCPSP/max). Recent advances in Constraint Programming (CP) and Temporal Networks have re-invoked interest in evaluating the advantages and drawbacks of various proactive and reactive scheduling methods. First, we present a new, CP-based fully proactive method. Second, we show how a reactive approach can be constructed using an online rescheduling procedure. A third contribution is based on partial order schedules and uses Simple Temporal Networks with Uncertainty (STNUs). Our statistical analysis shows that the STNU-based algorithm performs best in terms of solution quality, while also showing good relative offline and online computation time

Proactive and Reactive Constraint Programming for Stochastic Project Scheduling with Maximal Time-Lags

Two prominent objectives in social choice are utilitarian - maximizing the sum of agents' utilities, and leximin - maximizing the smallest agent's utility, then the second-smallest, etc. Utilitarianism is typically computationally easier to attain but is generally viewed as less fair.  This paper presents a general reduction scheme that, given a utilitarian solver, produces a distribution over outcomes that is leximin in expectation. 
Importantly, the scheme is robust in the sense that, given an approximate utilitarian solver, it produces an outcome that is approximately-leximin (in expectation) - with the same approximation factor.  We apply our scheme to several social choice problems: stochastic allocations of indivisible goods, giveaway lotteries, and fair lotteries for participatory budgeting.

Reducing Leximin Fairness to Utilitarian Optimization

Dataset condensation has significantly improved model training efficiency, but its application on devices with different computing power brings new requirements for different data sizes. For sparse graph data with non-Euclidean structures, repeated condensation of each scale may lead to significant computational costs. Thus, condensing multiple scale graphs simultaneously is the core of achieving efficient training in different on-device scenarios. Existing efficient works for multi-scale graph dataset condensation mainly perform efficient approximate computation in scale order (large-to-small or small-to-large scales). However, these two commonly used paradigms for multi-scale graph dataset condensation have serious *scaling down degradation* and *scaling up collapse*  problems of a graph. The main bottleneck of the above paradigms is whether the effective information of the original graph is fully preserved when consenting to the primary sub-scale (the first of multiple scales), which determines the condensation effect and consistency of all scales. In this paper, we proposed a novel GNN-centric **Bi**-directional **M**ulti-**S**cale **G**raph Dataset **C**ondensation (**BiMSGC**) framework, to explore unifying paradigms by operating on both large-to-small and small-to-large for multi-scale graph condensation. Based on the mutual information theory, we estimate an optimal ''*meso-scale*'' to obtain the minimum necessary dense graph preserving the maximum utility information of the original graph, and then we achieve stable and consistent ''*bi-directional*'' condensation learning by optimizing graph eigenbasis matching with information bottleneck on other scales. Encouraging empirical results on several datasets demonstrates the significant superiority of the proposed framework in graph condensation at different scales.

Bi-Directional Multi-Scale Graph Dataset Condensation via Information Bottleneck

We present an option discovery algorithm that accelerates planning by minimizing the shortest distance between any two states in the MDP. The proposed algorithm produces options that approximately minimize planning time in the multi-goal setting: it is shown to be a worst case $(4\alpha,2)$-approximation of the optimal option set, where $\alpha$ is the approximation ratio of the $k$-medians with penalties subroutine. We then present a variation, \textit{Fast Average Options}, with improved run-time and describe a general means of producing similar algorithms based on selection of a $k$-medians subroutine. We empirically evaluate our method on four discrete and two continuous control planning domains and show that it outperforms other leading option discovery algorithms.

Discovering Options That Minimize Average Planning Time

Procedural Content Generation via Machine Learning (PCGML) has enhanced game content creation, yet challenges in controllability and limited training data persist. This study addresses these issues by distilling a constructive PCG algorithm into a controllable PCGML model. We first generate a large amount of content with a constructive algorithm and label it using a Large Language Model (LLM). We use these synthetic labels to condition two PCGML models for content-specific generation, a diffusion model and the five-dollar model. This neural network distillation process ensures that the generation aligns with the original algorithm while introducing controllability through plain text. We define this text-conditioned PCGML as a Text-to-game-Map (T2M) task, offering an alternative to prevalent text-to-image multi-modal tasks. We compare our distilled models with the baseline constructive algorithm. Our analysis of the variety, accuracy, and quality of our generation demonstrates the efficacy of distilling constructive methods into controllable text-conditioned PCGML models.

Moonshine: Distilling Game Content Generators into Steerable Generative Models

Learning agents that excel at sequential decision-making tasks must continuously resolve the problem of exploration and exploitation for optimal learning. However, such interactions with the environment online might be prohibitively expensive and may involve some constraints, such as a limited budget for agent-environment interactions and restricted exploration in certain regions of the state space. Examples include selecting candidates for medical trials and training agents in complex navigation environments. This problem necessitates the study of active reinforcement learning strategies that collect minimal additional experience trajectories by reusing existing offline data previously collected by some unknown behavior policy.  In this work, we propose a representation-aware uncertainty-based active trajectory collection method that intelligently decides interaction strategies that consider the distribution of the existing offline data. With extensive experimentation, we demonstrate that our proposed method reduces additional online interaction with the environment by up to 75% over competitive baselines across various continuous control environments.

Active Reinforcement Learning Strategies for Offline Policy Improvement

Inductive Knowledge Graph Completion (KGC) aims to infer missing facts between newly emerged entities within knowledge graphs (KGs), posing a significant challenge. While recent studies have shown promising results in inferring such entities through knowledge subgraph reasoning, they suffer from (i) the semantic inconsistencies of similar relations, and (ii) noisy interactions inherent in KGs due to the presence of unconvincing knowledge for emerging entities. To address these challenges, we propose a Semantic Structure-aware Denoising Network (S$^2$DN) for inductive KGC. Our goal is to learn adaptable general semantics and reliable structures to distill consistent semantic knowledge while preserving reliable interactions within KGs. Specifically, we introduce a semantic smoothing module over the enclosing subgraphs to retain the universal semantic knowledge of relations. We incorporate a structure refining module to filter out unreliable interactions and offer additional knowledge, retaining robust structure surrounding target links. Extensive experiments conducted on three benchmark KGs demonstrate that S$^2$DN surpasses the performance of state-of-the-art models. These results demonstrate the effectiveness of S$^2$DN in preserving semantic consistency and enhancing the robustness of filtering out unreliable interactions in contaminated KGs. The code is available at https://anonymous.4open.science/r/SDN.

S$^2$DN: Learning to Denoise Unconvincing Knowledge for Inductive Knowledge Graph Completion

This paper considers the challenging problem of 3D Human Pose Estimation (HPE) from a sparse set of Inertial Measurement Units (IMUs). Existing efforts typically reconstruct a pose sequence by either directly tackling whole-body motions or focusing on distinctive spatio-temporal features of local body parts. Unfortunately, these methods ignore existing interdependent motor synergies amongst body parts, which may lead to pose estimation with ambiguous local parts. This observation motivates us to propose a hierarchical learning-based approach, HiPoser, which utilizes a hierarchical shared structure using Mamba blocks as the backbone to focus on the following estimation tasks, involving: 1) torso pose, 2) lower limbs pose, 3) upper limbs pose, and finally 4) global translation. These tasks selectively incorporate body motion states and are to be carried out sequentially in reconstructing part-based poses, which are amalgamated to estimate the final full-body pose with the global translation that satisfies inter-part consistencies. Our hierarchical structure allows HiPoser the flexibility in prioritizing different aspects of pose estimation, to emphasize more on detail or stability. Empirical evaluations over three benchmark datasets demonstrate the superiority of HiPoser over existing state-of-the-art models, suggesting that analyzing the synergistic movement of body parts is indeed important for advancing IMU-based 3D HPE.

ScaleOT: Privacy-utility-scalable Offsite-tuning with Dynamic LayerReplace and Selective Rank Compression

Downloads

Next from AAAI 2025

Temporal Task and Motion Planning with Metric Time for Multiple Object Navigation