NeurIPS 2024 Papers

Lip reading aims at transforming the videos of continuous lip movement into textual contents, and has achieved significant progress over the past decade. It serves as a critical yet practical assistance for speech-impaired individuals, with more practicability than speech recognition in noisy environments. With the increasing interpersonal communications in social media owing to globalization, the existing monolingual datasets for lip reading may not be sufficient to meet the exponential proliferation of bilingual and even multilingual users. However, to our best knowledge, research on code-switching is only explored in speech recognition, while the attempts in lip reading are seriously neglected. To bridge this gap, we have collected a bilingual code-switching lip reading benchmark composed of Chinese and English, dubbed CSLR. As the pioneering work, we recruited 62 speakers with proficient foundations in both spoken Chinese and English to express sentences containing both involved languages. Through rigorous criteria in data selection, CSLR benchmark has accumulated 85,560 video samples with a resolution of 1080x1920, totaling over 71.3 hours of high-quality code-switching lip movement data. To systematically evaluate the technical challenges in CSLR, we implement commonly-used lip reading backbones, as well as competitive solutions in code-switching speech for benchmark testing. Experiments show CSLR to be a challenging and under-explored lip reading task. We hope our proposed benchmark will extend the applicability of code-switching lip reading, and further contribute to the communities of cross-lingual communication and collaboration. Our dataset and benchmark are accessible at https://github.com/cslr-lipreading/CSLR.

Most real-world datasets consist of a natural hierarchy between classes or an inherent label structure that is either already available or can be constructed cheaply. However, most existing representation learning methods ignore this hierarchy, treating labels as permutation invariant. Recent work [Zeng et al., 2022] proposes using this structured information explicitly, but the use of Euclidean distance may distort the underlying semantic context [Chen et al., 2013]. In this work, motivated by the advantage of hyperbolic spaces in modeling hierarchical relationships, we propose a novel approach HypStructure: a Hyperbolic Structured regularization approach to accurately embed the label hierarchy into the learned representations. HypStructure is a simple-yet-effective regularizer that consists of a hyperbolic tree-based representation loss along with a centering loss, and can be combined with any standard task loss to learn hierarchy-informed features. Extensive experiments on several large-scale vision benchmarks demonstrate the efficacy of HypStructure in reducing distortion and boosting generalization performance especially under low dimensional scenarios. For a better understanding of structured representation, we perform eigenvalue analysis that links the representation geometry to improved Out-of-Distribution (OOD) detection performance seen empirically.

We present a novel multimodal preference dataset for creative tasks, consisting of over 250 million human votes on more than 2.2 million captions, collected through crowdsourcing rating data for The New Yorker's weekly cartoon caption contest over the past eight years. This unique dataset supports the development and evaluation of multimodal large language models and preference-based fine-tuning algorithms for humorous caption generation. We propose novel benchmarks for judging the quality of model-generated captions, utilizing both GPT4 and human judgments to establish ranking-based evaluation strategies. Our experimental results highlight the limitations of current fine-tuning methods, such as RLHF and DPO, when applied to creative tasks. Furthermore, we demonstrate that even state-of-the-art models like GPT4 and Claude currently underperform top human contestants in generating humorous captions. As we conclude this extensive data collection effort, we release the entire preference dataset to the research community, fostering further advancements in AI humor generation and evaluation.

Influential and popular benchmarks in AI are largely irrelevant to developing NLP tools for low-resource, Indigenous languages. With the primary goal of measuring the performance of general-purpose AI systems, these benchmarks fail to give due consideration and care to individual language communities, especially low-resource languages. The datasets contain numerous grammatical and orthographic errors, poor pronunciation, limited vocabulary, and the content lacks cultural relevance to the language community. To overcome the issues with these benchmarks, we have created a dataset for te reo Māori (the Indigenous language of Aotearoa/New Zealand) to pursue NLP tools that are ‘fit-for-our-purpose’. This paper demonstrates how low-resourced, Indigenous languages can develop tailored, high-quality benchmarks that; i. Consider the impact of colonisation on their language; ii. Reflect the diversity of speakers in the language community; iii. Support the aspirations for the tools they are developing and their language revitalisation efforts.

We introduce the Continuous Arcade Learning Environment (CALE), an extension of the well-known Arcade Learning Environment (ALE) [Bellemare et al., 2013]. The CALE uses the same underlying emulator of the Atari 2600 gaming system (Stella), but adds support for continuous actions. This enables the benchmarking and evaluation of continuous-control agents (such as PPO [Schulman et al., 2017] and SAC [Haarnoja et al., 2018]) and value-based agents (such as DQN [Mnih et al., 2015] and Rainbow [Hessel et al., 2018]) on the same environment suite. We provide a series of open questions and research directions that CALE enables, as well as initial baseline results using Soft Actor-Critic. CALE is available as part of the ALE athttps://github.com/Farama-Foundation/Arcade-Learning-Environment.

Compiler backends are tasked with generating executable machine code for processors. With the proliferation of diverse processors, it is imperative for programmers to tailor specific compiler backends to accommodate each one. Meanwhile, compiler backend development is a laborious and time-consuming task, lacking effective automation methods. Although language models have demonstrated strong abilities in code related tasks, the lack of appropriate datasets for compiler backend development limits the application of language models in this field. In this paper, we introduce ComBack, the first public dataset designed for improving compiler backend development capabilities of language models. ComBack includes 178 backends for mainstream compilers and three tasks including statement-level completion, next-statement suggestion and code generation, representing common development scenarios. We conducted experiments by fine-tuning six pre-trained language models with ComBack, demonstrating its effectiveness in enhancing model accuracy across the three tasks. We further evaluated the top-performing model(CodeT5+) across the three tasks for new targets, comparing its accuracy with conventional methods (Fork-Flow), ChatGPT-3.5-Turbo, and Code-LLaMA-34B-Instruct. Remarkably, fine-tuned CodeT5+ with only 220M parameters on ComBack outperformed Fork-Flow methods significantly and surpassed ChatGPT and Code-LLaMA. This suggests potential efficiency improvements in compiler development. ComBack is avaliable at https://huggingface.co/datasets/docz1105/ComBack.

While multi-modal large language models (MLLMs) have shown significant progress across popular visual reasoning benchmarks, whether they possess abstract visual reasoning abilities remains an open question. Similar to the Sudoku puzzles, abstract visual reasoning (AVR) problems require finding high-level patterns (e.g., repetition constraints on numbers) that control the input shapes (e.g., digits) in a specific task configuration (e.g., matrix). However, existing AVR benchmarks only consider a limited set of patterns (addition, conjunction), input shapes (rectangle, square), and task configurations (3 × 3 matrices). And they fail to capture all abstract reasoning patterns in human cognition necessary for addressing real-world tasks, such as geometric properties and object boundary understanding in real-world navigation. To evaluate MLLMs’ AVR abilities systematically, we introduce MARVEL founded on the core knowledge system in human cognition, a multi-dimensional AVR benchmark with 770 puzzles composed of six core knowledge patterns, geometric and abstract shapes, and five different task configurations. To inspect whether the model performance is grounded in perception or reasoning, MARVEL complements the standard AVR question with perception questions in a hierarchical evaluation framework. We conduct comprehensive experiments on MARVEL with ten representative MLLMs in zero-shot and few-shot settings. Our experiments reveal that all MLLMs show near-random performance on MARVEL, with significant performance gaps (40%) compared to humans across all patterns and task configurations. Further analysis of perception questions reveals that MLLMs struggle to comprehend the visual features (near-random performance). Although closed-source MLLMs, such as GPT-4V, show a promising understanding of reasoning patterns (on par with humans) after adding textual descriptions, this advantage is hindered by their weak perception abilities. We release our entire code and dataset at https://github.com/1171-jpg/MARVEL_AVR.

Respiratory audio, such as coughing and breathing sounds, has predictive power for a wide range of healthcare applications, yet is currently under-explored. The main problem for those applications arises from the difficulty in collecting large labeled task-specific data for model development. Generalizable respiratory acoustic foundation models pretrained with unlabeled data would offer appealing advantages and possibly unlock this impasse. However, given the safety-critical nature of healthcare applications, it is pivotal to also ensure openness and replicability for any proposed foundation model solution. To this end, we introduce OPERA, an OPEn Respiratory Acoustic foundation model pretraining and benchmarking system, as the first approach answering this need. We curate large-scale respiratory audio datasets ($\sim$136K samples, over 400 hours), pretrain three pioneering foundation models, and build a benchmark consisting of 19 downstream respiratory health tasks for evaluation. Our pretrained models demonstrate superior performance (against existing acoustic models pretrained with general audio on 16 out of 19 tasks) and generalizability (to unseen datasets and new respiratory audio modalities). This highlights the great promise of respiratory acoustic foundation models and encourages more studies using OPERA as an open resource to accelerate research on respiratory audio for health. The system is accessible from https://github.com/evelyn0414/OPERA.

Large language models (LLMs) have achieved unprecedented performances in various applications, yet evaluating them is still challenging. Existing benchmarks are either manually constructed or are automatic, but lack the ability to evaluate the thought process of LLMs with arbitrary complexity. We contend that utilizing existing relational databases based on the entity-relationship (ER) model is a promising approach for constructing benchmarks as they contain structured knowledge that can be used to question LLMs. Unlike knowledge graphs, which are also used to evaluate LLMs, relational databases have integrity constraints that can be used to better construct complex in-depth questions and verify answers: (1) functional dependencies can be used to pinpoint critical keywords that an LLM must know to properly answer a given question containing certain attribute values; and (2) foreign key constraints can be used to join relations and construct multi-hop questions, which can be arbitrarily long and used to debug intermediate answers. We thus propose ERBench, which uses these integrity constraints to convert any database into an LLM benchmark. ERBench supports continuous evaluation as databases change, multimodal questions, and various prompt engineering techniques. In our experiments, we construct LLM benchmarks using databases of multiple domains and make an extensive comparison of contemporary LLMs. We show how ERBench can properly evaluate any LLM by not only checking for answer correctness, but also effectively verifying the rationales by looking for the right keywords.

Question answering, asking, and assessment are three innate human traits crucial for understanding the world and acquiring knowledge. By enhancing these capabilities, humans can more effectively utilize data, leading to better comprehension and learning outcomes. However, current Multimodal Large Language Models (MLLMs) primarily focus on question answering, often neglecting the full potential of questioning and assessment skills. In this study, we introduce LOVA3, an innovative framework named ``Learning tO Visual Question Answering, Asking and Assessment,'' designed to equip MLLMs with these additional capabilities. Our approach involves the creation of two supplementary training tasks GenQA and EvalQA, aiming at fostering the skills of asking and assessing questions in the context of images. To develop the questioning ability, we compile a comprehensive set of multimodal foundational tasks. For assessment, we introduce a new benchmark called EvalQABench, comprising 64,000 training samples (split evenly between positive and negative samples) and 5,000 testing samples. We posit that enhancing MLLMs with the capabilities to answer, ask, and assess questions will enhance their multimodal comprehension, ultimately improving overall performance. To validate this hypothesis, we train MLLMs using the LOVA3 framework and evaluate them on a range of multimodal datasets and benchmarks. Our results demonstrate consistent performance gains, underscoring the critical role of these additional tasks in fostering comprehensive intelligence in MLLMs.

Despite significant advancements in text generation and reasoning, Large Language Models (LLMs) still face challenges in accurately performing complex arithmetic operations. Language model systems often enable LLMs to generate code for arithmetic operations to achieve accurate calculations. However, this approach compromises speed and security, and fine-tuning risks the language model losing prior capabilities. We propose a framework that enables exact arithmetic in *a single autoregressive step*, providing faster, more secure, and more interpretable LLM systems with arithmetic capabilities. We use the hidden states of a LLM to control a symbolic architecture that performs arithmetic. Our implementation using Llama 3 with OccamNet as a symbolic model (OccamLlama) achieves 100\% accuracy on single arithmetic operations ($+,-,\times,\div,\sin{},\cos{},\log{},\exp{},\sqrt{}$), outperforming GPT 4o with and without a code interpreter. Furthermore, OccamLlama outperforms GPT 4o with and without a code interpreter on average across a range of mathematical problem solving benchmarks, demonstrating that OccamLLMs can excel in arithmetic tasks, even surpassing much larger models. Code is available at https://github.com/druidowm/OccamLLM.

Steering vectors (SVs) are a new approach to efficiently adjust language model behaviour at inference time by intervening on intermediate model activations. They have shown promise in terms of improving both capabilities and model alignment. However, the reliability and generalisation properties of this approach are unknown. In this work, we rigorously investigate these properties, and show that steering vectors have substantial limitations both in- and out-of-distribution. In-distribution, steerability is highly variable across different inputs. Depending on the concept, spurious biases can substantially contribute to how effective steering is for each input, presenting a challenge for the widespread use of steering vectors. Out-of-distribution, while steering vectors often generalise well, for several concepts they are brittle to reasonable changes in the prompt, resulting in them failing to generalise well. Overall, our findings show that while steering can work well in the right circumstances, there remain many technical difficulties of applying steering vectors to guide models' behaviour at scale.

The sharp increase in data-related expenses has motivated research into condensing datasets while retaining the most informative features. Dataset distillation has thus recently come to the fore. This paradigm generates synthetic datasets that are representative enough to replace the original dataset in training a neural network. To avoid redundancy in these synthetic datasets, it is crucial that each element contains unique features and remains diverse from others during the synthesis stage. In this paper, we provide a thorough theoretical and empirical analysis of diversity within synthesized datasets. We argue that enhancing diversity can improve the parallelizable yet isolated synthesizing approach. Specifically, we introduce a novel method that employs dynamic and directed weight adjustment techniques to modulate the synthesis process, thereby maximizing the representativeness and diversity of each synthetic instance. Our method ensures that each batch of synthetic data mirrors the characteristics of a large, varying subset of the original dataset. Extensive experiments across multiple datasets, including CIFAR, Tiny-ImageNet, and ImageNet-1K, demonstrate the superior performance of our method, highlighting its effectiveness in producing diverse and representative synthetic datasets with minimal computational expense. Our code is available at https://github.com/AngusDujw/Diversity-Driven-Synthesis.

Datasets often suffer severe selection bias; clinical labels are only available on patients for whom doctors ordered medical exams. To assess model performance outside the support of available data, we present a computational framework for adaptive labeling, providing cost-efficient model evaluations under severe distribution shifts. We formulate the problem as a Markov Decision Process over states defined by posterior beliefs on model performance. Each batch of new labels incurs a “state transition” to sharper beliefs, and we choose batches to minimize uncertainty on model performance at the end of the label collection process. Instead of relying on high-variance REINFORCE policy gradient estimators that do not scale, our adaptive labeling policy is optimized using path-wise policy gradients computed by auto-differentiating through simulated roll-outs. Our framework is agnostic to different uncertainty quantification approaches and highlights the virtue of planning in adaptive labeling. On synthetic and real datasets, we empirically demonstrate even a one-step lookahead policy substantially outperforms active learning-inspired heuristics.

Polynomial-based learnable spectral graph neural networks (GNNs) utilize polynomial to approximate graph convolutions and have achieved impressive performance on graphs. Nevertheless, there are three progressive problems to be solved. Some models use polynomials with better approximation for approximating filters, yet perform worse on real-world graphs. Carefully crafted graph learning methods, sophisticated polynomial approximations, and refined coefficient constraints leaded to overfitting, which diminishes the generalization of the models. How to design a model that retains the ability of polynomial-based spectral GNNs to approximate filters while it possesses higher generalization and performance? In this paper, we propose a spectral GNN with triple filter ensemble (TFE-GNN), which extracts homophily and heterophily from graphs with different levels of homophily adaptively while utilizing the initial features. Specifically, the first and second ensembles are combinations of a set of base low-pass and high-pass filters, respectively, after which the third ensemble combines them with two learnable coefficients and yield a graph convolution (TFE-Conv). Theoretical analysis shows that the approximation ability of TFE-GNN is consistent with that of ChebNet under certain conditions, namely it can learn arbitrary filters. TFE-GNN can be viewed as a reasonable combination of two unfolded and integrated excellent spectral GNNs, which motivates it to perform well. Experiments show that TFE-GNN achieves high generalization and new state-of-the-art performance on various real-world datasets.

We focus on offline imitation learning (IL), which aims to mimic an expert's behavior using demonstrations without any interaction with the environment. One of the main challenges in offline IL is the limited support of expert demonstrations, which typically cover only a small fraction of the state-action space. While it may not be feasible to obtain numerous expert demonstrations, it is often possible to gather a larger set of sub-optimal demonstrations. For example, in treatment optimization problems, there are varying levels of doctor treatments available for different chronic conditions. These range from treatment specialists and experienced general practitioners to less experienced general practitioners. Similarly, when robots are trained to imitate humans in routine tasks, they might learn from individuals with different levels of expertise and efficiency. In this paper, we propose an offline IL approach that leverages the larger set of sub-optimal demonstrations while effectively mimicking expert trajectories. Existing offline IL methods based on behavior cloning or distribution matching often face issues such as overfitting to the limited set of expert demonstrations or inadvertently imitating sub-optimal trajectories from the larger dataset. Our approach, which is based on inverse soft-Q learning, learns from both expert and sub-optimal demonstrations. It assigns higher importance (through learned weights) to aligning with expert demonstrations and lower importance to aligning with sub-optimal ones. A key contribution of our approach, called SPRINQL, is transforming the offline IL problem into a convex optimization over the space of Q functions. Through comprehensive experimental evaluations, we demonstrate that the SPRINQL algorithm achieves state-of-the-art (SOTA) performance on offline IL benchmarks. Code is available at https://github.com/hmhuy0/SPRINQL .

We consider the dataset valuation problem, that is the problem of quantifying the incremental gain, to some relevant pre-defined utility of a machine learning task, of aggregating an individual dataset to others. The Shapley value is a natural tool to perform dataset valuation due to its formal axiomatic justification, which can be combined with Monte Carlo integration to overcome the computational tractability challenges. Such generic approximation methods, however, remain expensive in some cases. In this paper, we exploit the knowledge about the structure of the dataset valuation problem to devise more efficient Shapley value estimators. We propose a novel approximation, referred to as discrete uniform Shapley, which is expressed as an expectation under a discrete uniform distribution with support of reasonable size. We justify the relevancy of the proposed framework via asymptotic and non-asymptotic theoretical guarantees and illustrate its benefits via an extensive set of numerical experiments.

Linear RNN architectures, like Mamba, can be competitive with Transformer models in language modeling while having advantageous deployment characteristics. Given the focus on training large-scale Transformer models, we consider the challenge of converting these pretrained models for deployment. We demonstrate that it is feasible to distill large Transformers into linear RNNs by reusing the linear projection weights from attention layers with academic GPU resources. The resulting hybrid model, which incorporates a quarter of the attention layers, achieves performance comparable to the original Transformer in chat benchmarks and outperforms open-source hybrid Mamba models trained from scratch with trillions of tokens in both chat benchmarks and general benchmarks. Moreover, we introduce a hardware-aware speculative decoding algorithm that accelerates the inference speed of Mamba and hybrid models. Overall we show how, with limited computation resources, we can remove many of the original attention layers and generate from the resulting model more efficiently. Our top-performing model, distilled from Llama3-8B-Instruct, achieves a 29.61 length-controlled win rate on AlpacaEval 2 against GPT-4 and 7.35 on MT-Bench, surpassing the best 8B scale instruction-tuned linear RNN model. We also find that the distilled model has natural length extrapolation, showing almost perfect accuracy in the needle-in-a-haystack test at 20x the distillation length. Code and pre-trained checkpoints are open-sourced at [MambaInLlama](https://github.com/jxiw/MambaInLlama) for distillation and [SpeculativeMamba](https://github.com/itsdaniele/speculative\_mamba) for speculative decoding.

To make sense of massive data, we often first fit simplified models and then interpret the parameters; for example, we cluster the text embeddings and then interpret the mean parameters of each cluster. However, these parameters are often high-dimensional and hard to interpret. To make model parameters directly interpretable, we introduce a family of statistical models---including clustering, time series, and classification models---parameterized by *natural language predicates*. For example, a cluster of text about COVID could be parameterized by the predicate ``*discusses COVID*''. To learn these statistical models effectively, we develop a model-agnostic algorithm that optimizes continuous relaxations of predicate parameters with gradient descent and discretizes them by prompting language models (LMs). Finally, we apply our framework to a wide range of problems: taxonomizing user chat dialogues, characterizing how they evolve across time, finding categories where one language model is better than the other, clustering math problems based on subareas, and explaining visual features in memorable images. Our framework is highly versatile, applicable to both textual and visual domains, can be easily steered to focus on specific properties (e.g. subareas), and explains sophisticated concepts that classical methods (e.g. n-gram analysis) struggle to produce.