ICLR 2025 论文列表

To use generative question-and-answering (QA) systems for decision-making and in any critical application, these systems need to provide well-calibrated confidence scores that reflect the correctness of their answers. Existing calibration methods aim to ensure that the confidence score is, *on average*, indicative of the likelihood that the answer is correct. We argue, however, that this standard (average-case) notion of calibration is difficult to interpret for decision-making in generative QA. To address this, we generalize the standard notion of average calibration and introduce QA-calibration, which ensures calibration holds across different question-and-answer groups. We then propose discretized posthoc calibration schemes for achieving QA-calibration. We establish distribution-free guarantees on the performance of this method and validate our method on confidence scores returned by elicitation prompts across multiple QA benchmarks and large language models (LLMs).

Missing values are prevalent across various fields, posing challenges for training and deploying predictive models. In this context, imputation is a common practice, driven by the hope that accurate imputations will enhance predictions. However, recent theoretical and empirical studies indicate that simple constant imputation can be consistent and competitive. This empirical study aims at clarifying *if* and *when* investing in advanced imputation methods yields significantly better predictions. Relating imputation and predictive accuracies across combinations of imputation and predictive models on 19 datasets, we show that imputation accuracy matters less i) when using expressive models, ii) when incorporating missingness indicators as complementary inputs, iii) matters much more for generated linear outcomes than for real-data outcomes. Interestingly, we also show that the use of the missingness indicator is beneficial to the prediction performance, even in MCAR scenarios. Overall, on real-data with powerful models, imputation quality has only a minor effect on prediction performance. Thus, investing in better imputations for improved predictions often offers limited benefits.

The goal of the AlgoPerf: Training Algorithms competition is to evaluate practical speed-ups in neural network training achieved solely by improving the underlying training algorithms. In the external tuning ruleset, submissions must provide workload-agnostic hyperparameter search spaces, while in the self-tuning ruleset they must be completely hyperparameter-free. In both rulesets, submissions are compared on time-to-result across multiple deep learning workloads, training on fixed hardware. This paper presents the inaugural AlgoPerf competition's results, which drew 18 diverse submissions from 10 teams. Our investigation reveals several key findings: (1) The winning submission in the external tuning ruleset, using Distributed Shampoo, demonstrates the effectiveness of non-diagonal preconditioning over popular methods like Adam, even when compared on wall-clock runtime. (2) The winning submission in the self-tuning ruleset, based on the Schedule Free AdamW algorithm, demonstrates a new level of effectiveness for completely hyperparameter-free training algorithms. (3) The top-scoring submissions were surprisingly robust to workload changes. We also discuss the engineering challenges encountered in ensuring a fair comparison between different training algorithms. These results highlight both the significant progress so far, and the considerable room for further improvements.

Benchmarking the capabilities and limitations of large language models (LLMs) in graph-related tasks is becoming an increasingly popular and crucial area of research. Recent studies have shown that LLMs exhibit a preliminary ability to understand graph structures and node features. However, the potential of LLMs in graph pattern mining remains largely unexplored. This is a key component in fields such as computational chemistry, biology, and social network analysis. To bridge this gap, this work introduces a comprehensive benchmark to assess LLMs' capabilities in graph pattern tasks. We have developed a benchmark that evaluates whether LLMs can understand graph patterns based on either terminological or topological descriptions. Additionally, our benchmark tests the LLMs' capacity to autonomously discover graph patterns from data. The benchmark encompasses both synthetic and real datasets, and a variety of models, with a total of 11 tasks and 7 models. Our experimental framework is designed for easy expansion to accommodate new models and datasets. Our findings reveal that: (1) LLMs have preliminary abilities to understand graph patterns, with O1-mini outperforming in the majority of tasks; (2) Formatting input graph data to align with the knowledge acquired during pretraining can enhance performance; (3) LLMs employ diverse potential algorithms to solve one task, with performance varying based on their execution capabilities.

Compression is at the heart of intelligence. A theoretically optimal way to compress any sequence of data is to find the shortest program that outputs that sequence and then halts. However, such Kolmogorov compression is uncomputable, and code generating LLMs struggle to approximate this theoretical ideal, as it requires reasoning, planning and search capabilities beyond those of current models. In this work, we introduce the *KoLMogorov-Test* (KT), a compression-as-intelligence intelligence test for code generation LLMs. In KT a model is presented with a sequence of data at inference time, and asked to generate the shortest program that produces the sequence. We identify several benefits of KT for both evaluation and training: an essentially infinite number of problem instances of varying difficulty is readily available, strong baselines already exist, the evaluation metric (compression) cannot be gamed, and pretraining data contamination is highly unlikely. To evaluate current models, we use audio, text, and DNA data, as well as sequences produced by random synthetic programs. Current flagship models perform poorly - both GPT4-o and Llama-3.1-405B struggle on our natural and synthetic sequences. On our synthetic distribution, we are able to train code generation models with lower compression rates than previous approaches. Moreover, we show that gains on synthetic data generalize poorly to real data, suggesting that new innovations are necessary for additional gains on KT.

Event-guided imaging has received significant attention due to its potential to revolutionize instant imaging systems. However, the prior methods primarily focus on enhancing RGB images in a post-processing manner, neglecting the challenges of image signal processor (ISP) dealing with event sensor and the benefits events provide for reforming the ISP process. To achieve this, we conduct the first research on event-guided ISP. First, we present a new event-RAW paired dataset, collected with a novel but still confidential sensor that records pixel-level aligned events and RAW images. This dataset includes 3373 RAW images with $2248\times 3264$ resolution and their corresponding events, spanning 24 scenes with 3 exposure modes and 3 lenses. Second, we propose a convential ISP pipeline to generate good RGB frames as reference. This convential ISP pipleline performs basic ISP operations, e.g., demosaicing, white balancing, denoising and color space transforming, with a ColorChecker as reference. Third, we classify the existing learnable ISP methods into 3 classes, and select multiple methods to train and evaluate on our new dataset. Lastly, since there is no prior work for reference, we propose a simple event-guided ISP method and test it on our dataset. We further put forward key technical challenges and future directions in RGB-Event ISP. In summary, to the best of our knowledge, this is the very first research focusing on event-guided ISP, and we hope it will inspire the community.

Predicting the future motion of surrounding agents is essential for autonomous vehicles (AVs) to operate safely in dynamic, human-robot-mixed environments. However, the scarcity of large-scale driving datasets has hindered the development of robust and generalizable motion prediction models, limiting their ability to capture complex interactions and road geometries. Inspired by recent advances in natural language processing (NLP) and computer vision (CV), self-supervised learning (SSL) has gained significant attention in the motion prediction community for learning rich and transferable scene representations. Nonetheless, existing pre-training methods for motion prediction have largely focused on specific model architectures and single dataset, limiting their scalability and generalizability. To address these challenges, we propose SmartPretrain, a general and scalable SSL framework for motion prediction that is both model-agnostic and dataset-agnostic. Our approach integrates contrastive and reconstructive SSL, leveraging the strengths of both generative and discriminative paradigms to effectively represent spatiotemporal evolution and interactions without imposing architectural constraints. Additionally, SmartPretrain employs a dataset-agnostic scenario sampling strategy that integrates multiple datasets, enhancing data volume, diversity, and robustness. Extensive experiments on multiple datasets demonstrate that SmartPretrain consistently improves the performance of state-of-the-art prediction models across datasets, data splits and main metrics. For instance, SmartPretrain significantly reduces the MissRate of Forecast-MAE by 10.6\%. These results highlight SmartPretrain's effectiveness as a unified, scalable solution for motion prediction, breaking free from the limitations of the small-data regime.

Large language models (LLMs) can solve an increasing number of complex reasoning tasks while making surprising mistakes in basic numerical understanding and processing (such as $9.11 > 9.9$). The latter ability is essential for tackling complex arithmetic and mathematical problems and serves as a foundation for most reasoning tasks, but previous work paid little attention to it or only discussed several restricted tasks (like integer addition). In this paper, we comprehensively investigate the numerical understanding and processing ability (NUPA) of LLMs. Firstly, we introduce a benchmark covering four common numerical representations and 17 distinct numerical tasks in four major categories, resulting in 41 meaningful combinations in total. These tasks are derived from primary and secondary education curricula, encompassing nearly all everyday numerical understanding and processing scenarios, and the rules of these tasks are very simple and clear. Through the benchmark, we find that current LLMs fail frequently in many of the tasks. To study the problem, we train small models with existing and potential techniques for enhancing NUPA (such as tokenizers, PEs, and number formats), comprehensively evaluating their effectiveness using our testbed. We also finetune practical-scale LLMs on our proposed NUPA tasks and find that 1) naive finetuning can improve NUPA a lot on many but not all tasks, and 2) surprisingly, techniques designed to enhance NUPA prove ineffective for finetuning pretrained models. We further explore the impact of chain-of-thought techniques on NUPA. Our work provides a more detailed and comprehensive understanding of NUPA in LLMs.

Constructing confidence intervals (CIs) for the average treatment effect (ATE) from patient records is crucial to assess the effectiveness and safety of drugs. However, patient records typically come from different hospitals, thus raising the question of how multiple observational/experimental datasets can be effectively combined for this purpose. In our paper, we propose a new method that estimates the ATE from multiple observational/experimental datasets and provides valid CIs. Our method makes little assumptions about the observational datasets and is thus widely applicable in medical practice. The key idea of our method is that we leverage prediction-powered inferences and thereby essentially `shrink' the CIs so that we offer more precise uncertainty quantification as compared to na{\"i}ve approaches. We further prove the unbiasedness of our method and the validity of our CIs. We confirm our theoretical results through various numerical experiments.

Large Language Models (LLMs) are recognized as systems that closely mimic aspects of human intelligence. This capability has attracted the attention of the social science community, who see the potential in leveraging LLMs to replace human participants in experiments, thereby reducing research costs and complexity. In this paper, we introduce a benchmark for LLMs personification, including a strategy for constructing virtual characters' life stories from the ground up, a Multi-Agent Cognitive Mechanism capable of simulating human cognitive processes, and a psychology-guided evaluation method to assess human simulations from both self and observational perspectives. Experimental results demonstrate that our constructed simulacra can produce personified responses that align with their target characters. We hope this work will serve as a benchmark in the field of human simulation, paving the way for future research.

Large Language Model (LLM) providers typically train a family of models, each of a different size targeting a specific deployment scenario. Models in the family are all trained from scratch, making the process extremely resource intensive. Recent work has successfully reduced the cost of training model families through a combination of structured pruning and knowledge distillation; here, only the largest model in the family is trained from scratch, and smaller models are obtained via pruning. We observe that while effective, this strategy must still perform pruning and distillation with hundreds of billions of training tokens for every new model, keeping overall training costs high. In this work, we introduce a novel nested weight-shared architecture named LLaMaFlex that can be pruned across both width and depth dimensions in a zero-shot manner to instantly yield a large number of highly accurate compressed models. LLaMaFlex starts from a pretrained model, and only requires a single continued training phase consisting of ~60B tokens, which trains the elastic network and an end-to-end Gumbel Softmax-based router; this router is able to interpolate smoothly across model sizes, enabling the "train once, deploy many'' paradigm. We train LLaMaFlex on Llama 3.1 8B and use it to zero-shot generate a family of compressed models that achieves accuracy on par with or better than state-of-the-art pruned, elastic/flexible, and trained-from-scratch models.

We present Uni$^2$Det, a brand new framework for unified and universal multi-dataset training on 3D detection, enabling robust performance across diverse domains and generalization to unseen domains. Due to substantial disparities in data distribution and variations in taxonomy across diverse domains, training such a detector by simply merging datasets poses a significant challenge. Motivated by this observation, we introduce multi-stage prompting modules for multi-dataset 3D detection, which leverages prompts based on the characteristics of corresponding datasets to mitigate existing differences. This elegant design facilitates seamless plug-and-play integration within various advanced 3D detection frameworks in a unified manner, while also allowing straightforward adaptation for universal applicability across datasets. Experiments are conducted across multiple dataset consolidation scenarios involving KITTI, Waymo, and nuScenes, demonstrating that our Uni$^2$Det outperforms existing methods by a large margin in multi-dataset training. Notably, results on zero-shot cross-dataset transfer validate the generalization capability of our proposed method. Our code is available at https://github.com/ThomasWangY/Uni2Det.

Bilevel optimization has recently attracted considerable attention due to its abundant applications in machine learning problems. However, existing methods rely on prior knowledge of problem parameters to determine stepsizes, resulting in significant effort in tuning stepsizes when these parameters are unknown. In this paper, we propose two novel tuning-free algorithms, D-TFBO and S-TFBO. D-TFBO employs a double-loop structure with stepsizes adaptively adjusted by the "inverse of cumulative gradient norms" strategy. S-TFBO features a simpler fully single-loop structure that updates three variables simultaneously with a theory-motivated joint design of adaptive stepsizes for all variables. We provide a comprehensive convergence analysis for both algorithms and show that D-TFBO and S-TFBO respectively require $\mathcal{O}(\frac{1}{\epsilon})$ and $\mathcal{O}(\frac{1}{\epsilon}\log^4(\frac{1}{\epsilon}))$ iterations to find an $\epsilon$-accurate stationary point, (nearly) matching their well-tuned counterparts using the information of problem parameters. Experiments on various problems show that our methods achieve performance comparable to existing well-tuned approaches, while being more robust to the selection of initial stepsizes. To the best of our knowledge, our methods are the first to completely eliminate the need for stepsize tuning, while achieving theoretical guarantees.

In this paper, we use the Fire Dynamics Simulator (FDS) combined with the {\fontfamily{lmtt}\selectfont \textit{supercomputer}} support to create a \textbf{C}ombustion \textbf{K}inetics (CK) dataset for machine learning and scientific research. This dataset captures the development of fires in industrial parks with high-precision Computational Fluid Dynamics (CFD) simulations. It includes various physical fields such as temperature and pressure, and covers multiple environmental combinations for exploring \underline{multi-physics} field coupling phenomena. Additionally, we evaluate several advanced machine learning architectures across our {\fontfamily{lmtt}\selectfont {Open-CK}} benchmark using a substantial computational setup of 64 NVIDIA A100 GPUs: \ding{182} vision backbone; \ding{183} spatio-temporal predictive models; \ding{184} operator learning frameworks. These architectures uniquely excel at handling complex physical field data. We also introduce three benchmarks to demonstrate their potential in enhancing the exploration of downstream tasks: (a) capturing continuous changes in combustion kinetics; (b) a neural partial differential equation solver for learning temperature fields and turbulence; (c) reconstruction of sparse physical observations. The Open-CK dataset and benchmarks aim to advance research in combustion kinetics driven by machine learning, providing a reliable baseline for developing and comparing cutting-edge technologies and models. We hope to further promote the application of deep learning in earth sciences. Our project is available at \url{https://github.com/whscience/Open-CK}.

How can we automatically select an out-of-distribution (OOD) detection model for various underlying tasks? This is crucial for maintaining the reliability of open-world applications by identifying data distribution shifts, particularly in critical domains such as online transactions, autonomous driving, and real-time patient diagnosis. Despite the availability of numerous OOD detection methods, the challenge of selecting an optimal model for diverse tasks remains largely underexplored, especially in scenarios lacking ground truth labels. In this work, we introduce MetaOOD, the first zero-shot, unsupervised framework that utilizes meta-learning to select an OOD detection model automatically. As a meta-learning approach, MetaOOD leverages historical performance data of existing methods across various benchmark OOD detection datasets, enabling the effective selection of a suitable model for new datasets without the need for labeled data at the test time. To quantify task similarities more accurately, we introduce language model-based embeddings that capture the distinctive OOD characteristics of both datasets and detection models. Through extensive experimentation with 24 unique test dataset pairs to choose from among 11 OOD detection models, we demonstrate that MetaOOD significantly outperforms existing methods and only brings marginal time overhead. Our results, validated by Wilcoxon statistical tests, show that MetaOOD surpasses a diverse group of 11 baselines, including established OOD detectors and advanced unsupervised selection methods.

Understanding human locomotion is crucial for AI agents such as robots, particularly in complex indoor home environments. Modeling human trajectories in these spaces requires insight into how individuals maneuver around physical obstacles and manage social navigation dynamics. These dynamics include subtle behaviors influenced by proxemics - the social use of space, such as stepping aside to allow others to pass or choosing longer routes to avoid collisions. Previous research has developed datasets of human motion in indoor scenes, but these are often limited in scale and lack the nuanced social navigation dynamics common in home environments. To address this, we present LocoVR, a dataset of 7000+ two-person trajectories captured in virtual reality from over 130 different indoor home environments. LocoVR provides accurate trajectory and precise spatial information, along with rich examples of socially-motivated movement behaviors. For example, the dataset captures instances of individuals navigating around each other in narrow spaces, adjusting paths to respect personal boundaries in living areas, and coordinating movements in high-traffic zones like entryways and kitchens. Our evaluation shows that LocoVR significantly enhances model performance in three practical indoor tasks utilizing human trajectories, and demonstrates predicting socially-aware navigation patterns in home environments.

Accurate prediction of RNA properties, such as stability and interactions, is crucial for advancing our understanding of biological processes and developing RNA-based therapeutics. RNA structures can be represented as 1D sequences, 2D topological graphs, or 3D all-atom models, each offering different insights into its function. Existing works predominantly focus on 1D sequence-based models, which overlook the geometric context provided by 2D and 3D geometries. This study presents the first systematic evaluation of incorporating explicit 2D and 3D geometric information into RNA property prediction, considering not only performance but also real-world challenges such as limited data availability, partial labeling, sequencing noise, and computational efficiency. To this end, we introduce a newly curated set of RNA datasets with enhanced 2D and 3D structural annotations, providing a resource for model evaluation on RNA data. Our findings reveal that models with explicit geometry encoding generally outperform sequence-based models, with an average prediction RMSE reduction of around 12% across all various RNA tasks and excelling in low-data and partial labeling regimes, underscoring the value of explicitly incorporating geometric context. On the other hand, geometry-unaware sequence-based models are more robust under sequencing noise but often require around 2-5x training data to match the performance of geometry-aware models. Our study offers further insights into the trade-offs between different RNA representations in practical applications and addresses a significant gap in evaluating deep learning models for RNA tasks.

Feedback data is widely used for fine-tuning and evaluating state-of-the-art AI models. Pairwise text preferences, where human or AI annotators select the “better” of two options, are particularly common. Such preferences are used to train (reward) models or to rank models with aggregate statistics. For many applications it is desirable to understand annotator preferences in addition to modelling them  – not least because extensive prior work has shown various unintended biases in preference datasets. Yet, preference datasets remain challenging to interpret. Neither black-box reward models nor statistics can answer why one text is preferred over another. Manual interpretation of the numerous (long) response pairs is usually equally infeasible. In this paper, we introduce the Inverse Constitutional AI (ICAI) problem, formulating the interpretation of pairwise text preference data as a compression task. In constitutional AI, a set of principles (a constitution) is used to provide feedback and fine-tune AI models. ICAI inverts this process: given a feedback dataset, we aim to extract a constitution that best enables a large language model (LLM) to reconstruct the original annotations. We propose a corresponding ICAI algorithm and validate its generated constitutions quantitatively based on annotation reconstruction accuracy on several datasets: (a) synthetic feedback data with known principles; (b) AlpacaEval cross-annotated human feedback data; (c) crowdsourced Chatbot Arena data; and (d) PRISM data from diverse demographic groups. As an example application, we further demonstrate the detection of biases in human feedback data. As a short and interpretable representation of the original dataset, generated constitutions have many potential use cases: they may help identify undesirable annotator biases, better understand model performance, scale feedback to unseen data, or assist with adapting AI models to individual user or group preferences. We release the source code for our algorithm and experiments at https://github.com/rdnfn/icai.

Recent advances in internet-scale video data pretraining have led to the development of text-to-video generative models that can create high-quality videos across a broad range of visual concepts, synthesize realistic motions and render complex objects. Hence, these generative models have the potential to become general-purpose simulators of the physical world. However, it is unclear how far we are from this goal with the existing text-to-video generative models. To this end, we present VideoPhy, a benchmark designed to assess whether the generated videos follow physical commonsense for real-world activities (e.g. marbles will roll down when placed on a slanted surface). Specifically, we curate diverse prompts that involve interactions between various material types in the physical world (e.g., solid-solid, solid-fluid, fluid-fluid). We then generate videos conditioned on these captions from diverse state-of-the-art text-to-video generative models, including open models (e.g., CogVideoX) and closed models (e.g., Lumiere, Dream Machine). Our human evaluation reveals that the existing models severely lack the ability to generate videos adhering to the given text prompts, while also lack physical commonsense. Specifically, the best performing model, CogVideoX-5B, generates videos that adhere to the caption and physical laws for 39.6% of the instances. VideoPhy thus highlights that the video generative models are far from accurately simulating the physical world. Finally, we propose an auto-evaluator, VideoCon-Physics, to assess the performance reliably for the newly released models. The code is available here: https://github.com/Hritikbansal/videophy.