Papers for

enterprise ai platform operators

Papers whose findings have a practical use for this group, as judged from the abstract. Open a paper to read what it means in practice.

Budgetverify improves efficiency in financial question answering verification

BudgetVerify: Budget-Tiered Verification for Financial QA

Abstract: Financial question answering often requires precise numerical extraction, unit handling, and arithmetic over tables and text, but applying expensive verification uniformly wastes test-time compute. We propose BudgetVerify, a budget-tiered generator-verifier framework that routes each generated answer to one of three verification tiers: no verification, lightweight check-and-revise, or higher-cost solve-first-then-compare verification. The router is trained from offline correctness and token-cost outcomes and, at test time, selects a verification tier using information available before verification, including the question, context statistics, the generated answer, and associated generator metadata. The selected tier either returns the generated answer directly or invokes the corresponding verifier. Across six commercial and open-weight base models, BudgetVerify consistently produces more efficient accuracy-cost Pareto frontiers than fixed verification policies by selectively allocating stronger verification only when it is useful. Although absolute performance varies across models, these efficiency gains and the resulting qualitative frontier shape are consistent across generator models.

Sun 27 SeptArtificial IntelligenceMachine Learning
The gist
Answering financial questions accurately often means checking if numbers and calculations are right, which can take a lot of computer power. The authors propose BudgetVerify, a system that decides how much checking is needed for each answer, sometimes skipping checks and other times doing detailed verification. This approach uses information about the question and the generated answer to save computing resources while keeping answers accurate. The system was tested with different AI models and consistently balanced accuracy and cost better than using the same checking method every time.
Open → 2609.33052v1

Environment-probing memory curation improves agent task success and efficiency

Grounding Agent Memory: Environment-Probing Curation for Enterprise Agents

Abstract: Persistent memory is entering production-oriented agent platforms to help long-horizon agents accumulate experience across sessions. Yet a post-task curator agent restricted to completed trajectories can preserve errors, overgeneralize partial evidence, or retain stale knowledge. We introduce environment-probing curation, a deployment-compatible extension that gives an existing asynchronous curator agent least-privilege, read-only world tools to check, scope, and refresh candidate memories. It requires no model retraining and leaves the task agent, retriever, memory representation, and production write authority unchanged. In a production-like GitHub Copilot (GHCP) harness built on its SDK, we compare stateless execution, full in-context learning, GHCP + Mem, and GHCP + Mem (w/ Env Probing) on CLBench database exploration and 90 adapted APEX management-consulting tasks. On CLBench, probing raises pass rate from 39% to 73% and pass-discounted reward from 8.60 to 22.60 while reducing queries from 8.8 to 4.7 per question and task-agent cost from \$3.38 to \$1.68. Across six APEX worlds, all 18 memory-versus-baseline mean reward comparisons are positive and task-agent tool calls fall by 16--75%; probing gives the best task-agent reward gain per dollar in five worlds. Probing also attains higher mean reward than GHCP + Mem on both Sonnet 4.6 and Opus 4.7 without schema drift. Environment probing therefore turns existing agent-memory curation into an environment-informed, auditable process while preserving a compact task-time interface.

Thu 10 SeptArtificial IntelligenceSoftware Engineering
The gist
Long-term computer agents try to remember past experiences to perform better over time, but sometimes their memories can be wrong or outdated. The authors propose a method that lets a helper agent peek back into the environment in a limited way to check and update these memories without changing the main agent or retraining the system. This approach was tested with a coding assistant setup and several management task simulations, showing much better success rates and fewer unnecessary questions. Their method makes agent memory updating more accurate, efficient, and easier to audit.
Open → 2609.11060v1