Forage V2: Knowledge Evolution and Transfer in Autonomous Agent Organizations

TL;DR

Forage V2 introduces an organizational architecture for autonomous agents that accumulates, transfers, and safeguards knowledge across runs, improving reliability, efficiency, and evaluation calibration in open-world tasks.

Contribution

It extends Forage to a learning organization with institutional safeguards, enabling knowledge transfer and organizational memory that enhance agent performance and evaluation accuracy.

Findings

Knowledge entries grow from 0 to 54 over six runs.

Seeded agents reduce coverage gap from 6.6pp to 1.1pp.

Organizational knowledge calibrates evaluation, converging on the same denominator estimate.

Abstract

Autonomous agents operating in open-world tasks -- where the completion boundary is not given in advance -- face denominator blindness: they systematically underestimate the scope of the target space. Forage V1 addressed this through co-evolving evaluation (an independent Evaluator discovers what "complete" means) and method isolation (Evaluator and Planner cannot see each other's code). V2 extends the architecture from a single expedition to a learning organization: experience accumulates across runs, transfers across model capabilities, and institutional safeguards prevent knowledge degradation. We demonstrate two claims across three task types (web scraping, API queries, mathematical reasoning). Knowledge accumulation: over six runs, knowledge entries grow from 0 to 54, and denominator estimates stabilize as domain understanding deepens. Knowledge transfer: a weaker agent (Sonnet) seeded with a stronger agent's (Opus) knowledge narrows a 6.6pp coverage gap to 1.1pp, halves cost (9.40 to 5.13 USD), converges in half the rounds (mean 4.5 vs. 7.0), and three independent seeded runs arrive at exactly the same denominator estimate (266), suggesting organizational knowledge calibrates evaluation itself. V2's contribution is architectural: it designs institutions -- audit separation, contract protocols, organizational memory -- that make any agent more reliable upon entry. The accumulated experience is organizational, model-agnostic, and transferable, stored as readable documents that any future agent inherits regardless of provider or capability level.