Self-improving agent MoLRSI hits 15.9% in antibiotic screening versus 8.7% for an expert workflow

The arXiv preprint says the agent redesigned its own screening workflow using the same training data and library as the earlier expert-led one, and also produced eight experimentally validated peptide-MHC binders; results are not yet peer reviewed.

Read at arXiv preprint

More from Snape Report

Anthropic finds Claude can write training data that passes a production-style monitor yet makes a model misaligned
OpenAI has about $11 billion committed for its round, less than half its first $30 billion target, WSJ reports
Nasdaq slides 2.9% as Intel has its worst day since the 1970s and chip stocks fall
OpenAI releases GPT-Rosalind-6, a life-sciences model with fewer dual-use biology refusals to vetted groups only
Apple confirms Apple Intelligence launches in China in December with iOS 27.3, with Alibaba's Qwen
Wiz says one key stolen in AI platform breaches made 1.7 million Bedrock calls in 13 hours
OpenAI lists gpt-6-astra-law on its API pricing page at $10 per million input tokens and $50 per million output
S&P revises Meta's outlook to negative over AI capital spending and negative free cash flow

Front page →