Advances in Artificial Intelligence for Scientific Discovery

The field of artificial intelligence is rapidly advancing in its ability to facilitate scientific discovery. Recent developments have focused on improving the ability of language models to reason, generate hypotheses, and collaborate with other models to solve complex problems. One key area of advancement is in the use of foundation models, which have been shown to be effective in generating novel solutions and improving the performance of other models. Another area of focus is on the development of multi-agent systems, which enable the collaboration of multiple models to achieve a common goal. These systems have been applied to a variety of tasks, including drug discovery, protein structure prediction, and materials science. Notably, the integration of artificial intelligence with scientific domains such as chemistry and biology has led to significant breakthroughs, including the development of new antibiotics and the improvement of protein structure prediction models. Overall, the field is moving towards the development of more general-purpose models that can be applied to a wide range of scientific tasks, enabling faster and more efficient discovery. Noteworthy papers include Agent KB, which introduces a novel Reason-Retrieve-Refine pipeline for agentic problem solving, and ApexOracle, which predicts the antibacterial potency of existing compounds and designs de novo molecules active against strains it has never encountered.

Sources

Agent KB: Leveraging Cross-Domain Experience for Agentic Problem Solving

Virtual Laboratories: Domain-agnostic workflows for research

Automated Neuron Labelling Enables Generative Steering and Interpretability in Protein Language Models

Foundation Model Self-Play: Open-Ended Strategy Innovation via Foundation Models

Gradientsys: A Multi-Agent LLM Scheduler with ReAct Orchestration

PLAME: Leveraging Pretrained Language Models to Generate Enhanced Protein Multiple Sequence Alignments

Open Source Planning & Control System with Language Agents for Autonomous Scientific Discovery

Bridging the Plausibility-Validity Gap by Fine-Tuning a Reasoning-Enhanced LLM for Chemical Synthesis and Discovery

DrugMCTS: a drug repurposing framework combining multi-agent, RAG and Monte Carlo Tree Search

General purpose models for the chemical sciences

Synthetic MC via Biological Transmitters: Therapeutic Modulation of the Gut-Brain Axis

Predicting and generating antibiotics against future pathogens with ApexOracle

Built with on top of