E006 — Diversity Search

ACTIVE at checkpointPOLICY SEARCHREPRODUCED

Diversity Search improved the published verified real-time milestone from 5,050 to 5,750 points. Search uses zero model calls.

A new verified milestone

VERIFIED FROM INGESTED ARTIFACTSAUTOMATED POLICY SEARCHMODEL CALLS: 0REAL-TIMEREPLAY VERIFIED
Run ID
pleiads-20261005-131958-e6ceb4
Experiment / controller
E006 / native5
Score
5,750 points
Survival
216.760 s (emulated game time)
Real-time / replay
Confirmed / exact recorded-input replay
Screenshot matches
2,239 / 2,239 captured samples
Original manifest SHA-256
9df9cf29b00e0b2ad6f2d209b462bff7f27495a0c01c38b14e83a7c013ec3943

All captured samples were checked, at one capture every six emulated frames. Screenshot bytes and game software are excluded. CSV preserves the original TSV action order. This is a laboratory milestone, not a world-record claim.

A wider behavioural search space

  • Separate behavioural champions
  • Local mutations
  • Structural mutations
  • Crossover
  • New policy families
  • Behavioural diversity
  • Temporal score windows
  • Effective-policy canonicalisation
  • Exploration and exploitation branches

Validated checkpoint: 2026-10-05. Data last synchronized: .
Latest record verified from reviewed published run evidence. Registry counts retain the last full dataset checkpoint; record publication does not refresh the complete dataset. This is a snapshot, not a live counter.

Registry count checkpoint: 2026-10-05T08:28:16.420013+00:00. The latest reviewed record and its evidence update separately from full dataset totals.

Why change the method?

Small timing mutations can produce different policy files that execute the same actions before the run ends. E006 therefore retains separate behavioural champions and tests structural changes, using the E005 best score as its baseline.

Exploration and exploitation

An exploitation branch improves an existing champion. An exploration branch searches new families and behaviours. Temporal score windows help locate where progress is made or lost; canonicalisation compares policies by the actions that actually occurred before the run ended.

First E006 milestone / 5 October 2026

The new native5 run reached 5,510 points, with 202.211 seconds of survival, zero model calls and an exact recorded-input replay. A fresh read-only comparison matched all 2,091 captured screenshots, at six-frame intervals. Original manifests, actions, metrics and screenshot hashes accompany this checkpoint.

Two selected replay-verified real-time Search runs: E005 5,050 points and E006 5,510 points.
Two selected confirmed milestones, with zero model calls. This figure does not describe all run scores or establish a causal success rate. Run identities and chart provenance.

The retained plan descends from the E005 5,050 baseline and changes late-run behaviour. This demonstrates an observed improvement in one confirmed run, not a controlled estimate of the new search method’s success rate. The search remains active at the checkpoint. Accelerated candidates, real-time confirmation and replay verification continue to be reported separately.

Future diary entries will follow reviewed bundles with a new maximum or a useful failure. Score, run identity and timeline will update together; unvalidated internal runs do not create public pages.