NeurIPS 2024 Papers

In this paper, we propose a novel policy optimization framework that maximizes Return on Investment (ROI) of a policy using a fixed dataset within a Markov Decision Process (MDP) equipped with a cost function. ROI, defined as the ratio between the return and the accumulated cost of a policy, serves as a measure of efficiency of the policy. Despite the importance of maximizing ROI in various applications, it remains a challenging problem due to its nature as a ratio of two long-term values: return and accumulated cost. To address this, we formulate the ROI maximizing reinforcement learning problem as a linear fractional programming. We then incorporate the stationary distribution correction (DICE) framework to develop a practical offline ROI maximization algorithm. Our proposed algorithm, ROIDICE, yields an efficient policy that offers a superior trade-off between return and accumulated cost compared to policies trained using existing frameworks.

We study fair allocation of constrained resources, where a market designer optimizes overall welfare while maintaining group fairness. In many large-scale settings, utilities are not known in advance, but are instead observed after realizing the allocation. We therefore estimate agent utilities using machine learning. Optimizing over estimates requires trading-off between mean utilities and their predictive variances. We discuss these trade-offs under two paradigms for preference modeling – in the stochastic optimization regime, the market designer has access to a probability distribution over utilities, and in the robust optimization regime they have access to an uncertainty set containing the true utilities with high probability. We discuss utilitarian and egalitarian welfare objectives, and we explore how to optimize for them under stochastic and robust paradigms. We demonstrate the efficacy of our approaches on three publicly available conference reviewer assignment datasets. The approaches presented enable scalable constrained resource allocation under uncertainty for many combinations of objectives and preference models.

Goal-conditioned reinforcement learning is a powerful way to control an AI agent's behavior at runtime. That said, popular goal representations, e.g., target states or natural language, are either limited to Markovian tasks or rely on ambiguous task semantics. We propose representing temporal goals using compositions of deterministic finite automata (cDFAs) and use cDFAs to guide RL agents. cDFAs balance the need for formal temporal semantics with ease of interpretation: if one can understand a flow chart, one can understand a cDFA. On the other hand, cDFAs form a countably infinite concept class with Boolean semantics, and subtle changes to the automaton can result in very different tasks, making them difficult to condition agent behavior on. To address this, we observe that all paths through a DFA correspond to a series of reach-avoid tasks and propose pre-training graph neural network embeddings on "reach-avoid derived" DFAs. Through empirical evaluation, we demonstrate that the proposed pre-training method enables zero-shot generalization to various cDFA task classes and accelerated policy specialization without the myopic suboptimality of hierarchical methods.

Diffusion models represent a promising avenue for image generation, having demonstrated competitive performance in pose-guided person image generation. However, existing methods are limited to generating target images from a source image and a target pose, overlooking two critical user scenarios: generating multiple target images with different poses simultaneously and generating target images from multi-view source images. To overcome these limitations, we propose IMAGPose, a unified conditional framework for pose-guided image generation, which incorporates three pivotal modules: a feature-level conditioning (FLC) module, an image-level conditioning (ILC) module, and a cross-view attention (CVA) module. Firstly, the FLC module combines the low-level texture feature from the VAE encoder with the high-level semantic feature from the image encoder, addressing the issue of missing detail information due to the absence of a dedicated person image feature extractor. Then, the ILC module achieves an alignment of images and poses to adapt to flexible and diverse user scenarios by injecting a variable number of source image conditions and introducing a masking strategy. Finally, the CVA module introduces decomposing global and local cross-attention, ensuring local fidelity and global consistency of the person image when multiple source image prompts. The three modules of IMAGPose work together to unify the task of person image generation under various user scenarios. Extensive experiment results demonstrate the consistency and photorealism of our proposed IMAGPose under challenging user scenarios. The code and model will be available at https://github.com/muzishen/IMAGPose.

Transformer architectures have been widely adopted in foundation models. Due to their high inference costs, there is renewed interest in exploring the potential of efficient recurrent architectures (RNNs). In this paper, we analyze the differences in the representational capabilities of Transformers and RNNs across several tasks of practical relevance, including index lookup, nearest neighbor, recognizing bounded Dyck languages, and string equality. For the tasks considered, our results show separations based on the size of the model required for different architectures. For example, we show that a one-layer Transformer of logarithmic width can perform index lookup, whereas an RNN requires a hidden state of linear size. Conversely, while constant-size RNNs can recognize bounded Dyck languages, we show that one-layer Transformers require a linear size for this task. Furthermore, we show that two-layer Transformers of logarithmic size can perform decision tasks such as string equality or disjointness, whereas both one-layer Transformers and recurrent models require linear size for these tasks. We also show that a log-size two-layer Transformer can implement the nearest neighbor algorithm in its forward pass; on the other hand recurrent models require linear size. Our constructions are based on the existence of $N$ nearly orthogonal vectors in $O(\log N)$ dimensional space and our lower bounds are based on reductions from communication complexity problems. We supplement our theoretical results with experiments that highlight the differences in the performance of these architectures on practical-size sequences.

Neural algorithmic reasoning is an emerging area of machine learning that focuses on building neural networks capable of solving complex algorithmic tasks. Recent advancements predominantly follow the standard supervised learning paradigm -- feeding an individual problem instance into the network each time and training it to approximate the execution steps of a classical algorithm. We challenge this mode and propose a novel open-book learning framework. In this framework, whether during training or testing, the network can access and utilize all instances in the training dataset when reasoning for a given instance. Empirical evaluation is conducted on the challenging CLRS Algorithmic Reasoning Benchmark, which consists of 30 diverse algorithmic tasks. Our open-book learning framework exhibits a significant enhancement in neural reasoning capabilities. Further, we notice that there is recent literature suggesting that multi-task training on CLRS can improve the reasoning accuracy of certain tasks, implying intrinsic connections between different algorithmic tasks. We delve into this direction via the open-book framework. When the network reasons for a specific task, we enable it to aggregate information from training instances of other tasks in an attention-based manner. We show that this open-book attention mechanism offers insights into the inherent relationships among various tasks in the benchmark and provides a robust tool for interpretable multi-task training.

This paper tackles the challenge of active perception for robotic grasp detection in cluttered environments. Incomplete 3D geometry information can negatively affect the performance of learning-based grasp detection methods, and scanning the scene from multiple views introduces significant time costs. To achieve reliable grasping performance with efficient camera movement, we propose an active grasp detection framework based on the Neural Graspness Field (NGF), which models the scene incrementally and facilitates next-best-view planning. Constructed in real-time as the camera moves, the NGF effectively models the grasp distribution in 3D space by rendering graspness predictions from each view. For next-best-view planning, we aim to reduce the uncertainty of the NGF through a graspness inconsistency-guided policy, selecting views based on discrepancies between NGF outputs and a pre-trained graspness network. Additionally, we present a neural graspness sampling method that decodes graspness values from the NGF to improve grasp pose detection results. Extensive experiments on the GraspNet-1Billion benchmark demonstrate significant performance improvements compared to previous works. Real-world experiments show that our method achieves a superior trade-off between grasping performance and time costs.

Dynamic scene reconstruction is a long-term challenge in the field of 3D vision. Recently, the emergence of 3D Gaussian Splatting has provided new insights into this problem. Although subsequent efforts rapidly extend static 3D Gaussian to dynamic scenes, they often lack explicit constraints on object motion, leading to optimization difficulties and performance degradation. To address the above issues, we propose a novel deformable 3D Gaussian splatting framework called MotionGS, which explores explicit motion priors to guide the deformation of 3D Gaussians. Specifically, we first introduce an optical flow decoupling module that decouples optical flow into camera flow and motion flow, corresponding to camera movement and object motion respectively. Then the motion flow can effectively constrain the deformation of 3D Gaussians, thus simulating the motion of dynamic objects. Additionally, a camera pose refinement module is proposed to alternately optimize 3D Gaussians and camera poses, mitigating the impact of inaccurate camera poses. Extensive experiments in the monocular dynamic scenes validate that MotionGS surpasses state-of-the-art methods and exhibits significant superiority in both qualitative and quantitative results. Project page: https://ruijiezhu94.github.io/MotionGS_page.

This paper studies first-order policy optimization for robust average cost Markov decision processes (MDPs). Specifically, we focus on ergodic Markov chains. For robust average cost MDPs, the goal is to optimize the worst-case average cost over an uncertainty set of transition kernels. We first develop a sub-gradient of the robust average cost. Based on the sub-gradient, a robust policy mirror descent approach is further proposed. To characterize its iteration complexity, we develop a lower bound on the difference of robust average cost between two policies and further show that the robust average cost satisfies the PL-condition. We then show that with increasing step size, our robust policy mirror descent achieves a linear convergence rate in the optimality gap, and with constant step size, our algorithm converges to an $\epsilon$-optimal policy with an iteration complexity of $\mathcal{O}(1/\epsilon)$. The convergence rate of our algorithm matches with the best convergence rate of policy-based algorithms for robust MDPs. Moreover, our algorithm is the first algorithm that converges to the global optimum with general uncertainty sets for robust average cost MDPs. We provide simulation results to demonstrate the performance of our algorithm.

Unobserved discrete data are ubiquitous in many scientific disciplines, and how to learn the causal structure of these latent variables is crucial for uncovering data patterns. Most studies focus on the linear latent variable model or impose strict constraints on latent structures, which fail to address cases in discrete data involving non-linear relationships or complex latent structures. To achieve this, we explore a tensor rank condition on contingency tables for an observed variable set $\mathbf{X}_p$, showing that the rank is determined by the minimum support of a specific conditional set (not necessary in $\mathbf{X}_p$) that d-separates all variables in $\mathbf{X}_p$. By this, one can locate the latent variable through probing the rank on different observed variables set, and further identify the latent causal structure under some structure assumptions. We present the corresponding identification algorithm and conduct simulated experiments to verify the effectiveness of our method. In general, our results elegantly extend the identification boundary for causal discovery with discrete latent variables and expand the application scope of causal discovery with latent variables.

The collected data in recommender systems generally suffers selection bias. Considerable works are proposed to address selection bias induced by observed user and item features, but they fail when hidden features (e.g., user age or salary) that affect both user selection mechanism and feedback exist, which is called hidden confounding. To tackle this issue, methods based on sensitivity analysis and leveraging a few randomized controlled trial (RCT) data for model calibration are proposed. However, the former relies on strong assumptions of hidden confounding strength, whereas the latter relies on the expensive RCT data, thereby limiting their applicability in real-world scenarios. In this paper, we propose to employ heterogeneous observational data to address hidden confounding, wherein some data is subject to hidden confounding while the remaining is not. We argue that such setup is more aligned with practical scenarios, especially when some users do not have complete personal information (thus assumed with hidden confounding), while others do have (thus assumed without hidden confounding). To achieve unbiased learning, we propose a novel meta-learning based debiasing method called MetaDebias. This method explicitly models oracle error imputation and hidden confounding bias, and utilizes bi-level optimization for model training. Extensive experiments on three public datasets validate our method achieves state-of-the-art performance in the presence of hidden confounding, regardless of RCT data availability.

We study convex resource allocation problems with $m$ hard constraints under $(\varepsilon,\delta)$-joint differential privacy (Joint-DP or JDP) in an offline setting. To approximately solve the problem, we propose a generic algorithm called Noisy Dual Mirror Descent. The algorithm applies noisy Mirror Descent to a dual problem from relaxing the hard constraints for private shadow prices, and then uses the shadow prices to coordinate allocations in the primal problem. Leveraging weak duality theory, we show that the optimality gap is upper bounded by $\mathcal{O}(\frac{\sqrt{m\ln(1/\delta)}}{\varepsilon})$, and constraint violation is no more than $\mathcal{O}(\frac{\sqrt{m\ln(1/\delta)}}{\varepsilon})$ per constraint. When strong duality holds, both preceding results can be improved to $\widetilde{\mathcal{O}}(\frac{\sqrt{\ln(1/\delta)}}{\varepsilon})$ by better utilizing the geometric structure of the dual space, which is neglected by existing works. To complement our results under strong duality, we derive a minimax lower bound $\Omega(\frac{m}{\varepsilon})$ for any JDP algorithm outputting feasible allocations. The lower bound matches our upper bounds up to some logarithmic factors for $\varepsilon\geq \max(1, 1/(n\gamma))$, where $n\gamma$ is the available resource level. Numerical studies further confirm the effectiveness of our algorithm.

Despite spiking neural networks (SNNs) have demonstrated notable energy efficiency across various fields, the limited firing patterns of spiking neurons within fixed time steps restrict the expression of information, which impedes further improvement of SNN performance. In addition, current implementations of SNNs typically consider the firing rate or average membrane potential of the last layer as the output, lacking exploration of other possibilities. In this paper, we identify that the limited spike patterns of spiking neurons stem from the initial membrane potential (IMP), which is set to 0. By adjusting the IMP, the spiking neurons can generate additional firing patterns and pattern mappings. Furthermore, we find that in static tasks, the accuracy of SNNs at each time step increases as the membrane potential evolves from zero. This observation inspires us to propose a learnable IMP, which can accelerate the evolution of membrane potential and enables higher performance within a limited number of time steps. Additionally, we introduce the last time step (LTS) approach to accelerate convergence in static tasks, and we propose a label smooth temporal efficient training (TET) loss to mitigate the conflicts between optimization objective and regularization term in the vanilla TET. Our methods improve the accuracy by 4.05\% on ImageNet compared to baseline and achieve state-of-the-art performance of 87.80\% on CIFAR10-DVS and 87.86\% on N-Caltech101.

Evaluating large language models (LLMs) is challenging. Traditional ground-truth- based benchmarks fail to capture the comprehensiveness and nuance of real-world queries, while LLM-as-judge benchmarks suffer from grading biases and limited query quantity. Both of them may also become contaminated over time. User- facing evaluation, such as Chatbot Arena, provides reliable signals but is costly and slow. In this work, we propose MixEval, a new paradigm for establishing efficient, gold-standard LLM evaluation by strategically mixing off-the-shelf bench- marks. It bridges (1) comprehensive and well-distributed real-world user queries and (2) efficient and fairly-graded ground-truth-based benchmarks, by matching queries mined from the web with similar queries from existing benchmarks. Based on MixEval, we further build MixEval-Hard, which offers more room for model improvement. Our benchmarks’ advantages lie in (1) a 0.96 model ranking correlation with Chatbot Arena arising from the highly impartial query distribution and grading mechanism, (2) fast, cheap, and reproducible execution (6% of the time and cost of MMLU), and (3) dynamic evaluation enabled by the rapid and stable data update pipeline. We provide extensive meta-evaluation and analysis for our and existing LLM benchmarks to deepen the community’s understanding of LLM evaluation and guide future research directions.

Pre-trained language models have been proven to possess strong base capabilities, which not only excel in in-distribution language modeling but also show powerful abilities in out-of-distribution language modeling, transfer learning and few-shot learning. Unlike existing work focusing on the influence of scale on base capabilities, our work examines the influence of architecture on those. Specifically, our concern is: How does architecture influence the base capabilities of pre-trained language models? In this work, we attempt to explain and reverse the decline in base capabilities caused by the architecture of FFN-Wider Transformers, seeking to provide some insights. Through analysis, we found the contribution ratio of Multi-Head Attention (a combination function) to pre-trained language modeling is a key factor affecting base capabilities. FFN-Wider Transformers reduce the contribution ratio of this combination function, leading to a decline in base capabilities. We confirmed this by experiments and proposed Combination Enhanced Architecture (CEA) to address the decline in base capabilities of such models. Significantly, we extended our explanation and CEA to Mixture of Experts (MoE) Transformers. We successfully achieved significant improvements in base capabilities on a 14B parameter MoE model, demonstrating the practical application value of our work. This also indicates that our analysis has a certain guiding significance for architecture analysis, architecture improvement and architecture design.

Recent research has made significant strides in aligning large language models (LLMs) with helpfulness and harmlessness. In this paper, we argue for the importance of alignment for \emph{honesty}, ensuring that LLMs proactively refuse to answer questions when they lack knowledge, while still not being overly conservative. However, a pivotal aspect of alignment for honesty involves discerning an LLM's knowledge boundaries, which demands comprehensive solutions in terms of metric development, benchmark creation, and training methodologies. We address these challenges by first establishing a precise problem definition and defining ``honesty'' inspired by the Analects of Confucius. This serves as a cornerstone for developing metrics that effectively measure an LLM's honesty by quantifying its progress post-alignment. Furthermore, we introduce a flexible training framework which is further instantiated by several efficient fine-tuning techniques that emphasize honesty without sacrificing performance on other tasks. Our extensive experiments reveal that these aligned models show a marked increase in honesty, as indicated by our proposed metrics. We open-source all relevant resources to facilitate future research at \url{https://github.com/GAIR-NLP/alignment-for-honesty}.

Accurately identifying and organizing textual content is crucial for the automation of document processing in the field of form understanding. Existing datasets, such as FUNSD and XFUND, support entity classification and relationship prediction tasks but are typically limited to local and entity-level annotations. This limitation overlooks the hierarchically structured representation of documents, constraining comprehensive understanding of complex forms. To address this issue, we present the SRFUND, a hierarchically structured multi-task form understanding benchmark. SRFUND provides refined annotations on top of the original FUNSD and XFUND datasets, encompassing five tasks: (1) word to text-line merging, (2) text-line to entity merging, (3) entity category classification, (4) item table localization, and (5) entity-based full-document hierarchical structure recovery. We meticulously supplemented the original dataset with missing annotations at various levels of granularity and added detailed annotations for multi-item table regions within the forms. Additionally, we introduce global hierarchical structure dependencies for entity relation prediction tasks, surpassing traditional local key-value associations. The SRFUND dataset includes eight languages including English, Chinese, Japanese, German, French, Spanish, Italian, and Portuguese, making it a powerful tool for cross-lingual form understanding. Extensive experimental results demonstrate that the SRFUND dataset presents new challenges and significant opportunities in handling diverse layouts and global hierarchical structures of forms, thus providing deep insights into the field of form understanding. The original dataset and implementations of baseline methods are available at https://sprateam-ustc.github.io/SRFUND.

General-purpose artificial intelligence (AI) systems are built on massive swathes of public web data, assembled into corpora such as C4, RefinedWeb, and Dolma. To our knowledge, we conduct the first, large-scale, longitudinal audit of the consent protocols for the web domains underlying AI training corpora. Our audit of 14,000 web domains provides an expansive view of crawlable web data and how codified data use preferences are changing over time. We observe a proliferation of AI-specific clauses to limit use, acute differences in restrictions on AI developers, as well as general inconsistencies between websites' expressed intentions in their Terms of Service and their robots.txt. We diagnose these as symptoms of ineffective web protocols, not designed to cope with the widespread re-purposing of the internet for AI. Our longitudinal analyses show that in a single year (2023-2024) there has been a rapid crescendo of data restrictions from web sources, rendering ~5\%+ of all tokens in C4, or 28%+ of the most actively maintained, critical sources in C4, fully restricted from use. For Terms of Service crawling restrictions, a full 45% of C4 is now restricted. If respected or enforced, these restrictions are rapidly biasing the diversity, freshness, and scaling laws for general-purpose AI systems. We hope to illustrate the emerging crises in data consent, for both developers and creators. The foreclosure of much of the open web will impact not only commercial AI, but also non-commercial AI and academic research.

The recent large-scale text-to-image generative models have attained unprecedented performance, while people established *adaptor* modules like LoRA and DreamBooth to extend this performance to even more unseen concept tokens. However, we empirically find that this workflow often fails to accurately depict the *out-of-distribution* concepts. This failure is highly related to the low quality of training data. To resolve this, we present a framework called Controllable Adaptor Towards Out-of-Distribution Concepts (CATOD). Our framework follows the active learning paradigm which includes high-quality data accumulation and adaptor training, enabling a finer-grained enhancement of generative results. The *aesthetics* score and *concept-matching* score are two major factors that impact the quality of synthetic results. One key component of CATOD is the weighted scoring system that automatically balances between these two scores and we also offer comprehensive theoretical analysis for this point. Then, it determines how to select data and schedule the adaptor training based on this scoring system. The extensive results show that CATOD significantly outperforms the prior approaches with an 11.10 boost on the CLIP score and a 33.08% decrease on the CMMD metric.