Garrett Honke

Carl Edwards, Tuan Lai, Kevin Ros et al.

We present $\textbf{MolT5}$ $-$ a self-supervised learning framework for pretraining models on a vast amount of unlabeled natural language text and molecule strings. $\textbf{MolT5}$ allows for new, useful, and challenging analogs of traditional vision-language tasks, such as molecule captioning and text-based de novo molecule generation (altogether: translation between molecules and language), which we explore for the first time. Since $\textbf{MolT5}$ pretrains models on single-modal data, it helps overcome the chemistry domain shortcoming of data scarcity. Furthermore, we consider several metrics, including a new cross-modal embedding-based metric, to evaluate the tasks of molecule captioning and text-based molecule generation. Our results show that $\textbf{MolT5}$-based models are able to generate outputs, both molecules and captions, which in many cases are high quality.

Gheorghe Comanici, Eric Bieber, Mike Schaekermann et al. · amazon-science, baidu

In this report, we introduce the Gemini 2.X model family: Gemini 2.5 Pro and Gemini 2.5 Flash, as well as our earlier Gemini 2.0 Flash and Flash-Lite models. Gemini 2.5 Pro is our most capable model yet, achieving SoTA performance on frontier coding and reasoning benchmarks. In addition to its incredible coding and reasoning skills, Gemini 2.5 Pro is a thinking model that excels at multimodal understanding and it is now able to process up to 3 hours of video content. Its unique combination of long context, multimodal and reasoning capabilities can be combined to unlock new agentic workflows. Gemini 2.5 Flash provides excellent reasoning abilities at a fraction of the compute and latency requirements and Gemini 2.0 Flash and Flash-Lite provide high performance at low latency and cost. Taken together, the Gemini 2.X model generation spans the full Pareto frontier of model capability vs cost, allowing users to explore the boundaries of what is possible with complex agentic problem solving.

Paul Azunre, Craig Corcoran, Numa Dhamani et al.

A character-level convolutional neural network (CNN) motivated by applications in "automated machine learning" (AutoML) is proposed to semantically classify columns in tabular data. Simulated data containing a set of base classes is first used to learn an initial set of weights. Hand-labeled data from the CKAN repository is then used in a transfer-learning paradigm to adapt the initial weights to a more sophisticated representation of the problem (e.g., including more classes). In doing so, realistic data imperfections are learned and the set of classes handled can be expanded from the base set with reduced labeled data and computing power requirements. Results show the effectiveness and flexibility of this approach in three diverse domains: semantic classification of tabular data, age prediction from social media posts, and email spam classification. In addition to providing further evidence of the effectiveness of transfer learning in natural language processing (NLP), our experiments suggest that analyzing the semantic structure of language at the character level without additional metadata---i.e., network structure, headers, etc.---can produce competitive accuracy for type classification, spam classification, and social media age prediction. We present our open-source toolkit SIMON, an acronym for Semantic Inference for the Modeling of ONtologies, which implements this approach in a user-friendly and scalable/parallelizable fashion.

John Kirchenbauer, Garrett Honke, Gowthami Somepalli et al.

We develop a methodology for analyzing language model task performance at the individual example level based on training data density estimation. Experiments with paraphrasing as a controlled intervention on finetuning data demonstrate that increasing the support in the training distribution for specific test queries results in a measurable increase in density, which is also a significant predictor of the performance increase caused by the intervention. Experiments with pretraining data demonstrate that we can explain a significant fraction of the variance in model perplexity via density measurements. We conclude that our framework can provide statistical evidence of the dependence of a target model's predictions on subsets of its training data, and can more generally be used to characterize the support (or lack thereof) in the training data for a given test task.

Joel Niklaus, Lucia Zheng, Arya D. McCarthy et al.

Instruction tuning is an important step in making language models useful for direct user interaction. However, the legal domain is underrepresented in typical instruction datasets (e.g., only 10 out of 1600+ tasks in Super-NaturalInstructions). To study whether instruction tuning on legal datasets is necessary for strong legal reasoning, we aggregate 58 annotated legal datasets and write instructions for each, creating LawInstruct. LawInstruct covers 17 global jurisdictions, 24 languages and a total of 12M examples across diverse tasks such as legal QA, summarization of court cases, and legal argument mining. We evaluate our models on LegalBench, measuring legal reasoning across five categories in 162 challenging and realistic legal tasks, and MMLU, to measure potential drops in general reasoning capabilities. We find that legal-specific instruction tuning on Flan-T5 - yielding FLawN-T5 - improves performance on LegalBench across all model sizes, with an aggregate increase of 15 points or 50% over Flan-T5 for the base size. No model size shows performance drops in MMLU. We publish LawInstruct as a resource for further study of instruction tuning in the legal domain.

Ayse S. Cakmak, Nina Thigpen, Garrett Honke et al.

Depression and post-traumatic stress disorder (PTSD) are psychiatric conditions commonly associated with experiencing a traumatic event. Estimating mental health status through non-invasive techniques such as activity-based algorithms can help to identify successful early interventions. In this work, we used locomotor activity captured from 1113 individuals who wore a research grade smartwatch post-trauma. A convolutional variational autoencoder (VAE) architecture was used for unsupervised feature extraction from four weeks of actigraphy data. By using VAE latent variables and the participant's pre-trauma physical health status as features, a logistic regression classifier achieved an area under the receiver operating characteristic curve (AUC) of 0.64 to estimate mental health outcomes. The results indicate that the VAE model is a promising approach for actigraphy data analysis for mental health outcomes in long-term studies.

Garrett Honke, Irina Higgins, Nina Thigpen et al.

Despite extensive standardization, diagnostic interviews for mental health disorders encompass substantial subjective judgment. Previous studies have demonstrated that EEG-based neural measures can function as reliable objective correlates of depression, or even predictors of depression and its course. However, their clinical utility has not been fully realized because of 1) the lack of automated ways to deal with the inherent noise associated with EEG data at scale, and 2) the lack of knowledge of which aspects of the EEG signal may be markers of a clinical disorder. Here we adapt an unsupervised pipeline from the recent deep representation learning literature to address these problems by 1) learning a disentangled representation using $β$-VAE to denoise the signal, and 2) extracting interpretable features associated with a sparse set of clinical labels using a Symbol-Concept Association Network (SCAN). We demonstrate that our method is able to outperform the canonical hand-engineered baseline classification method on a number of factors, including participant age and depression diagnosis. Furthermore, our method recovers a representation that can be used to automatically extract denoised Event Related Potentials (ERPs) from novel, single EEG trajectories, and supports fast supervised re-mapping to various clinical labels, allowing clinicians to re-use a single EEG representation regardless of updates to the standardized diagnostic system. Finally, single factors of the learned disentangled representations often correspond to meaningful markers of clinical factors, as automatically detected by SCAN, allowing for human interpretability and post-hoc expert analysis of the recommendations made by the model.

Numa Dhamani, Paul Azunre, Jeffrey L. Gleason et al.

We apply an ensemble pipeline composed of a character-level convolutional neural network (CNN) and a long short-term memory (LSTM) as a general tool for addressing a range of disinformation problems. We also demonstrate the ability to use this architecture to transfer knowledge from labeled data in one domain to related (supervised and unsupervised) tasks. Character-level neural networks and transfer learning are particularly valuable tools in the disinformation space because of the messy nature of social media, lack of labeled data, and the multi-channel tactics of influence campaigns. We demonstrate their effectiveness in several tasks relevant for detecting disinformation: spam emails, review bombing, political sentiment, and conversation clustering.

Paul Azunre, Craig Corcoran, David Sullivan et al.

This paper describes an abstractive summarization method for tabular data which employs a knowledge base semantic embedding to generate the summary. Assuming the dataset contains descriptive text in headers, columns and/or some augmenting metadata, the system employs the embedding to recommend a subject/type for each text segment. Recommendations are aggregated into a small collection of super types considered to be descriptive of the dataset by exploiting the hierarchy of types in a pre-specified ontology. Using February 2015 Wikipedia as the knowledge base, and a corresponding DBpedia ontology as types, we present experimental results on open data taken from several sources--OpenML, CKAN and data.world--to illustrate the effectiveness of the approach.