Agents use habits to give consistent results on repeated tasks
Making Agents More Consistent: Skills Should Form Habits for Repeat Tasks
Artificial IntelligenceCryptography and Security
Summary
When computer agents do the same task multiple times, they often give different answers, which can be a problem for users who rely on consistent results. The authors found that these agents waste a lot of effort redoing the same reasoning each time. They propose a new method where agents create habits—fixed scripts learned from their own past work—to handle common cases faster and more reliably. This approach greatly improves consistency in outputs and reduces the effort spent on repeated tasks, while still allowing careful reasoning when needed.
What this means in practice
- •For enterprise ai teams: Ensure AI agents produce reliable and auditable outputs on repeated business queries by implementing skill habit formation to reduce redundant reasoning.
- •For software engineers: Build more efficient and consistent automated task agents by separating routing logic from parameter extraction and using deterministic skill habits.
Authors
Travis Weber, Rohit Taneja
Abstract
On repeated work, agents are inconsistent. We ran 42 tasks three times each and found that, depending on the model, 38% to 74% returned answers that did not agree. Consistency is what a buyer, an auditor, or a regulator requires, and agents do not have it. They are wasteful too: 95.3% to 97.2% of what an agent generates goes to re-deriving a plan the system already knows. We propose skill habit formation. An agent mines its own execution history for candidate skills, deterministic variants that compete against the incumbent rather than replacing it. A candidate declares the region of input space it claims, so the common case runs as a script and the rest falls through to reasoning. Four gates of ascending cost admit candidates; the central one tests a candidate's execution trace against a retained reference, within a tolerance measured from that reference's own run-to-run variability. On text-to-SQL, three of four reasoning arms reproduced their own output on 11 to 13 of 42 repeated questions and the fourth on 26 of 42, while a habit-formed variant reproduced on all 456 dispatches we repeated and was non-inferior to every arm it replaced (p<0.0001). It also used 14% to 56% fewer tokens, turning net positive after 7 to 53 reuses. We measured what this costs in accuracy. The guard admitted work it should have deferred on 2.6% of natural paraphrases and 26% of inputs near its boundary, and 11 of 13 such failures were invisible to the trace-conformance gate at any threshold. Deterministic errors repeat exactly: a bad habit is as reliable as a good one, and that is the price of the property that makes the system auditable. Separating routing from parameter extraction raised end-to-end accuracy from 0.888 to 0.952 at 43% of the cost.