Future Token Prediction -- Causal Language Modelling with Per-Token Semantic State Vector for Multi-Token Prediction

Causal decoder-only transformer models used for generative language modelling, such as Generative Pre-trained Transformers (GPT), are trained to predict the next token in a sequence based only on its previous tokens. Despite this simple training objective, they have proved to be powerful AI tools. However, only predicting the next token results in top layer embedding vectors that are highly token-focused. There may be benefits in generating embedding vectors at each token position that better capture the overall meaning of longer sequences of future text. Recent studies matching brain scans with deep language models suggest that humans also predict upcoming words when listening or reading but consider multiple future tokens rather than just one. This research investigates a new pretraining method called Future Token Prediction (FTP). In FTP, a large transformer encoder generates top layer embedding vectors for each token position, which, instead of being passed to a language head, are linearly and expansively projected to a pseudo-sequence, which is cross attended to by a small transformer decoder to predict the next N tokens forward from that position in the sequence. The top layer embedding vectors from FTP models exhibit distinct properties compared to those from standard GPT models, varying smoothly along a text sequence as measured by cosine similarity between adjacent tokens. Text generated by FTP models show improved topic coherence compared to standard GPT-like models trained with the same prediction perplexity for the next single token. The vectors are shown to better represent the topic of text based on the results of text classification examples. On a toy, but complex, coding problem, FTP networks produce significantly better results than GPT networks.

Extracting Structured Seed-Mediated Gold Nanorod Growth Procedures from Literature with GPT-3

Although gold nanorods have been the subject of much research, the pathways for controlling their shape and thereby their optical properties remain largely heuristically understood. Although it is apparent that the simultaneous presence of and interaction between various reagents during synthesis control these properties, computational and experimental approaches for exploring the synthesis space can be either intractable or too time-consuming in practice. This motivates an alternative approach leveraging the wealth of synthesis information already embedded in the body of scientific literature by developing tools to extract relevant structured data in an automated, high-throughput manner. To that end, we present an approach using the powerful GPT-3 language model to extract structured multi-step seed-mediated growth procedures and outcomes for gold nanorods from unstructured scientific text. GPT-3 prompt completions are fine-tuned to predict synthesis templates in the form of JSON documents from unstructured text input with an overall accuracy of $86\%$. The performance is notable, considering the model is performing simultaneous entity recognition and relation extraction. We present a dataset of 11,644 entities extracted from 1,137 papers, resulting in 268 papers with at least one complete seed-mediated gold nanorod growth procedure and outcome for a total of 332 complete procedures.

Structured information extraction from complex scientific text with fine-tuned large language models

Intelligently extracting and linking complex scientific information from unstructured text is a challenging endeavor particularly for those inexperienced with natural language processing. Here, we present a simple sequence-to-sequence approach to joint named entity recognition and relation extraction for complex hierarchical information in scientific text. The approach leverages a pre-trained large language model (LLM), GPT-3, that is fine-tuned on approximately 500 pairs of prompts (inputs) and completions (outputs). Information is extracted either from single sentences or across sentences in abstracts/passages, and the output can be returned as simple English sentences or a more structured format, such as a list of JSON objects. We demonstrate that LLMs trained in this way are capable of accurately extracting useful records of complex scientific knowledge for three representative tasks in materials chemistry: linking dopants with their host materials, cataloging metal-organic frameworks, and general chemistry/phase/morphology/application information extraction. This approach represents a simple, accessible, and highly-flexible route to obtaining large databases of structured knowledge extracted from unstructured text. An online demo is available at http://www.matscholar.com/info-extraction.