Agent objective
- Interact with the supplied stateful environment.
- Produce verifier-checkable actions or artifacts.
- Maximise scalar reward under the package contract.
Five metareasoning environments where the world worsens while an agent thinks. Every reasoning, inspection, hold, checkpoint, or commitment advances hidden state first, so computation can improve a plan while making it infeasible. Exact simulators test information gathering, option preservation, stopping, and action across negotiation, infrastructure, auctions, forensics, and distributed recovery.
Five metareasoning environments where the world worsens while an agent thinks. Every reasoning, inspection, hold, checkpoint, or commitment advances hidden state first, so computation can improve a plan while making it infeasible. Exact simulators test information gathering, option preservation, stopping, and action across negotiation, infrastructure, auctions, forensics, and distributed recovery.
Enough detail to understand the intellectual terrain; generated instances, hidden mechanisms, and solution paths remain inside the private package.
| Environment | Mathematical or technical frontier | Adaptive research problem |
|---|---|---|
| TreatyForge | Dynamic coalition negotiation | Inspect decisive preferences, bind short-lived support, and commit before reservations harden and parties withdraw. |
| CascadeLab | Nonlinear infrastructure intervention | Inspect sensors, contain temporarily, simulate, and deploy a small robust portfolio before failures propagate. |
| AuctionStorm | Online combinatorial auctions | Value clause inspection, factor updates, option freezing, and stopping while bids reprice, arrive, cancel, or expire. |
| Chronicle | Time-decaying incident forensics | Preserve decisive records and reconstruct an exact causal chain amid corruption, expiration, echoes, and forgeries. |
| QuorumStorm | Partitioned replicated recovery | Fence risky nodes and balance availability with safety while writes, leases, elections, crashes, and partitions evolve. |
A frozen policy scored 55.25 with 48% exact success across 100 public standard episodes, versus 21.49 and 2% for greedy. It excelled at coalition and infrastructure cases but remained weak on auctions and quorum recovery, yielding a useful capability profile rather than saturation. Scores are public development results with semantic ticks but no independently metered internal compute; release validation passed 48 tests and 320 property invariants.
We publish aggregate behavior and task structure, while withholding generated instances, hidden labels, exact successful probes, private checks, and solution trajectories.
Shown with its provenance and limitations; it is not a performance guarantee.
Public development standard suite. Success rate was 48%, versus 2% for the baseline.
As identified by the supplied artifact.
100 reported runs.
redqueen2_score_bundle.zip:reports/baseline_controls_v1.json
Machine-readable provenance and the exact displayed metric are available in results.json.
The paid ZIP will live in a private R2 bucket. Vercel authorizes the buyer and issues a 2–5 minute object URL; R2 serves the bytes directly.
Authenticated buyer + entitlement check
+ private R2 object + 2–5 minute signed URL
= direct, auditable download
Package SHA-256
a5f5f11980461940a5fb0d97c570552f96f376a4e0652a55148357d50ea50fb8One purchase licenses this identified item to one legal organisation for worldwide, perpetual commercial model training, evaluation, research and development. Redistribution and resale of the package are not permitted.