Operon-S1 speeds up decision making for biology language agents
Operon-S1: a fast decision controller for long-horizon biomedical agents
Artificial IntelligenceMachine Learning
Summary
Working with complex biological tasks often requires language-based AI agents to make many small decisions during their reasoning process. These decisions are usually made internally, which makes them hard to improve or measure directly. The authors created Operon-S1, a quick decision controller that separates judgment from action by posing specific yes/no questions. This system speeds up judgments dramatically without losing accuracy and allows decisions to be retrained or replaced independently from the agent’s main reasoning. Like a biological system that controls gene activity without producing proteins, Operon-S1 guides the agent without doing the work itself.
What this means in practice
- •For biomedical ai developers: Integrate Operon-S1 to speed up and modularize decision making in long biomedical AI tasks without reducing accuracy.
- •For laboratory automation engineers: Use Operon-S1’s decision framework to improve troubleshooting and design of laboratory protocols with measurable and trainable judgments.
Authors
Genoria AI Team, Yuejie Hou, Yueyuxiao Yang, Weiting Tang, Meng Yang
Abstract
Language-model agents now carry out long scientific work in biology, but their judgements about conventions, skills, errors, evidence and stopping are made inside the reasoning that does the work, where they cannot be measured, replaced or improved on their own. Here we present Operon-S1, a fast decision system, a System 1 for the agent, that separates judging from doing. A decision harness poses the agent's recurring judgements as closed questions at 13 decision points and acts only through fixed messages that never contain an answer, so any model that answers the questions can serve as the judge. Across seven life-science benchmarks, a local decision model took over these judgements in 0.6 s at the median for an in-loop request on CPU at no request cost, with no measurable change in the agent's score on five of six benchmarks, whereas the general language model that drives the agent, asked the same questions, accounted for 8.0% to 35.2% of task cost. Every decision is recorded with its consequence. Retraining on labels from task construction, verifiers or execution raised decision accuracy on every question in each of three trained versions, and a new version enters service only through a promotion gate. Task-level results in this study use the public checkpoint. The same closed decisions serve wet-laboratory protocol troubleshooting and the design of a germline target enrichment panel. Like the regulator of an operon, which switches genes on and off without making their products, Operon-S1 steers the agent without doing its work, and makes the agent's judgements a layer that can be measured, replaced and trained.