Meriem Boubdir, Edward Kim, Beyza Ermis et al. · nvidia
In Natural Language Processing (NLP), the Elo rating system, originally designed for ranking players in dynamic games such as chess, is increasingly being used to evaluate Large Language Models (LLMs) through "A vs B" paired comparisons. However, while popular, the system's suitability for assessing entities with constant skill levels, such as LLMs, remains relatively unexplored. We study two fundamental axioms that evaluation methods should adhere to: reliability and transitivity. We conduct extensive evaluation of Elo behaviour, illustrating that individual Elo computations exhibit volatility and delving into the impact of varying the Elo rating system's hyperparameters. We show that these axioms are not always satisfied raising questions about the reliability of current comparative evaluations of LLMs. If the current use of Elo scores is intended to substitute the costly head-to-head comparison of LLMs, it is crucial to ensure the ranking is as robust as possible. Guided by the axioms, our findings offer concrete guidelines for enhancing the reliability of LLM evaluation methods, suggesting a need for reassessment of existing comparative approaches.
Xiangyun Lei, Edward Kim, Viktoriia Baibakova et al.
Natural Language Processing (NLP), a cornerstone field within artificial intelligence, has been increasingly utilized in the field of materials science literature. Our study conducts a reproducibility analysis of two pioneering works within this domain: "Machine-learned and codified synthesis parameters of oxide materials" by Kim et al., and "Unsupervised word embeddings capture latent knowledge from materials science literature" by Tshitoyan et al. We aim to comprehend these studies from a reproducibility perspective, acknowledging their significant influence on the field of materials informatics, rather than critiquing them. Our study indicates that both papers offered thorough workflows, tidy and well-documented codebases, and clear guidance for model evaluation. This makes it easier to replicate their results successfully and partially reproduce their findings. In doing so, they set commendable standards for future materials science publications to aspire to. However, our analysis also highlights areas for improvement such as to provide access to training data where copyright restrictions permit, more transparency on model architecture and the training process, and specifications of software dependency versions. We also cross-compare the word embedding models between papers, and find that some key differences in reproducibility and cross-compatibility are attributable to design choices outside the bounds of the models themselves. In summary, our study appreciates the benchmark set by these seminal papers while advocating for further enhancements in research reproducibility practices in the field of NLP for materials science. This balance of understanding and continuous improvement will ultimately propel the intersecting domains of NLP and materials science literature into a future of exciting discoveries.
Creation and curation of knowledge graphs can accelerate disease discovery and analysis in real-world data. While disease ontologies aid in biological data annotation, codified categories (SNOMED-CT, ICD10, CPT) may not capture patient condition nuances or rare diseases. Multiple disease definitions across data sources complicate ontology mapping and disease clustering. We propose creating patient knowledge graphs using large language model extraction techniques, allowing data extraction via natural language rather than rigid ontological hierarchies. Our method maps to existing ontologies (MeSH, SNOMED-CT, RxNORM, HPO) to ground extracted entities. Using a large ambulatory care EHR database with 33.6M patients, we demonstrate our method through the patient search for Dravet syndrome, which received ICD10 recognition in October 2020. We describe our construction of patient-specific knowledge graphs and symptom-based patient searches. Using confirmed Dravet syndrome ICD10 codes as ground truth, we employ LLM-based entity extraction to characterize patients in grounded ontologies. We then apply this method to identify Beta-propeller protein-associated neurodegeneration (BPAN) patients, demonstrating real-world discovery where no ground truth exists.
Ambiguity is pervasive in real-world questions, yet large language models (LLMs) often respond with confident answers rather than seeking clarification. In this work, we show that question ambiguity is linearly encoded in the internal representations of LLMs and can be both detected and controlled at the neuron level. During the model's pre-filling stage, we identify that a small number of neurons, as few as one, encode question ambiguity information. Probes trained on these Ambiguity-Encoding Neurons (AENs) achieve strong performance on ambiguity detection and generalize across datasets, outperforming prompting-based and representation-based baselines. Layerwise analysis reveals that AENs emerge from shallow layers, suggesting early encoding of ambiguity signals in the model's processing pipeline. Finally, we show that through manipulating AENs, we can control LLM's behavior from direct answering to abstention. Our findings reveal that LLMs form compact internal representations of question ambiguity, enabling interpretable and controllable behavior.
Research has shown that neurons within the brain are selective to certain stimuli. For example, the fusiform face area (FFA) region is known by neuroscientists to selectively activate when people see faces over non-face objects. However, the mechanisms by which the primary visual system directs information to the correct higher levels of the brain are currently unknown. In our work, we mimic several high-level neural mechanisms of perception by creating a novel computational model that incorporates lateral and top down feedback in the form of hierarchical competition. Not only do we show that these elements can help explain the information flow and selectivity of high level areas within the brain, we also demonstrate that these neural mechanisms provide the foundation of a novel classification framework that rivals traditional supervised learning in computer vision. Additionally, we present both quantitative and qualitative results that demonstrate that our generative framework is consistent with neurological themes and enables simple, yet robust category level classification.
Materials discovery is often compared to the challenge of finding a needle in a haystack. While much work has focused on accurately predicting the properties of candidate materials with machine learning (ML), which amounts to evaluating whether a given candidate is a piece of straw or a needle, less attention has been paid to a critical question: Are we searching in the right haystack? We refer to the haystack as the design space for a particular materials discovery problem (i.e. the set of possible candidate materials to synthesize), and thus frame this question as one of design space selection. In this paper, we introduce two metrics, the Predicted Fraction of Improved Candidates (PFIC), and the Cumulative Maximum Likelihood of Improvement (CMLI), which we demonstrate can identify discovery-rich and discovery-poor design spaces, respectively. Using CMLI and PFIC together to identify optimal design spaces can significantly accelerate ML-driven materials discovery.
Materials science literature contains millions of materials synthesis procedures described in unstructured natural language text. Large-scale analysis of these synthesis procedures would facilitate deeper scientific understanding of materials synthesis and enable automated synthesis planning. Such analysis requires extracting structured representations of synthesis procedures from the raw text as a first step. To facilitate the training and evaluation of synthesis extraction models, we introduce a dataset of 230 synthesis procedures annotated by domain experts with labeled graphs that express the semantics of the synthesis sentences. The nodes in this graph are synthesis operations and their typed arguments, and labeled edges specify relations between the nodes. We describe this new resource in detail and highlight some specific challenges to annotating scientific text with shallow semantic structure. We make the corpus available to the community to promote further research and development of scientific information extraction systems.
Edward Kim, Zach Jensen, Alexander van Grootel et al.
Leveraging new data sources is a key step in accelerating the pace of materials design and discovery. To complement the strides in synthesis planning driven by historical, experimental, and computed data, we present an automated method for connecting scientific literature to synthesis insights. Starting from natural language text, we apply word embeddings from language models, which are fed into a named entity recognition model, upon which a conditional variational autoencoder is trained to generate syntheses for arbitrary materials. We show the potential of this technique by predicting precursors for two perovskite materials, using only training data published over a decade prior to their first reported syntheses. We demonstrate that the model learns representations of materials corresponding to synthesis-related properties, and that the model's behavior complements existing thermodynamic knowledge. Finally, we apply the model to perform synthesizability screening for proposed novel perovskite compounds.
Deep feed-forward convolutional neural networks (CNNs) have become ubiquitous in virtually all machine learning and computer vision challenges; however, advancements in CNNs have arguably reached an engineering saturation point where incremental novelty results in minor performance gains. Although there is evidence that object classification has reached human levels on narrowly defined tasks, for general applications, the biological visual system is far superior to that of any computer. Research reveals there are numerous missing components in feed-forward deep neural networks that are critical in mammalian vision. The brain does not work solely in a feed-forward fashion, but rather all of the neurons are in competition with each other; neurons are integrating information in a bottom up and top down fashion and incorporating expectation and feedback in the modeling process. Furthermore, our visual cortex is working in tandem with our parietal lobe, integrating sensory information from various modalities. In our work, we sought to improve upon the standard feed-forward deep learning model by augmenting them with biologically inspired concepts of sparsity, top-down feedback, and lateral inhibition. We define our model as a sparse coding problem using hierarchical layers. We solve the sparse coding problem with an additional top-down feedback error driving the dynamics of the neural network. While building and observing the behavior of our model, we were fascinated that multimodal, invariant neurons naturally emerged that mimicked, "Halle Berry neurons" found in the human brain. Furthermore, our sparse representation of multimodal signals demonstrates qualitative and quantitative superiority to the standard feed-forward joint embedding in common vision and machine learning tasks.
Computational synthesis planning approaches have achieved recent success in organic chemistry, where tabulated synthesis procedures are readily available for supervised learning. The syntheses of inorganic materials, however, exist primarily as natural language narratives contained within scientific journal articles. This synthesis information must first be extracted from the text in order to enable analogous synthesis planning methods for inorganic materials. In this work, we present a system for automatically extracting structured representations of synthesis procedures from the texts of materials science journal articles that describe explicit, experimental syntheses of inorganic compounds. We define the structured representation as a set of linked events made up of extracted scientific entities and evaluate two unsupervised approaches for extracting these structures on expert-annotated articles: a strong heuristic baseline and a generative model of procedural text. We also evaluate a variety of supervised models for extracting scientific entities. Our results provide insight into the nature of the data and directions for further work in this exciting new area of research.