POWM / Octave: eight learning observers / Report

OCTAVE — School training and benchmark results

Current01 Oct 2026 · saved

The archived source is available to read.

01 / THE QUESTION

Editorial guide

When the runnable eight-observer model is trained from scratch on a broader pulse school, does Level 2 training improve Level 1 forecasts, and how does it compare with a dense baseline?

02 / WHAT THE RECORD SAYS

Source passage

L2 improves held-out L1 bit accuracy in all three recurrent seeds: mean +1.66 percentage points. Exact L1 improves in seed 117 but declines in seeds 228 and 339: mean +2.34 points, driven by one seed.

03 / LIMITS & OPEN POINTS

Source passage

This is a preliminary fixed-budget experiment, not a mastery or frontier-performance certificate. More updates could change the comparison.

04 / THE NEXT QUESTION

Source passage

Eight-observer superiority requires matched observer-count ablations.

05 / CHECKS IN THE RECORD

Source passage

Final exams did not select checkpoints, alter teaching or drive training. Train bands, held-out bands, noise protocol and reproduction commands are documented in README.txt. This is a preliminary fixed-budget experiment, not a mastery or frontier-performance certificate. More updates could change the comparison. Same data/update budget does not mean same compute: architectures, learning rates and gradient clipping differ. Eight-observer superiority requires matched observer-count ablations. Complete L2 forecasts require all 64 A+B bits correct, whereas complete L1 forecasts require 32 A bits. Log loss and Brier scores appear in raw results. No uncertainty calibration is claimed. Historical recurrence/copy reference uses strong integer-lag structure and separately trained weights.

Read this passage in context ↗

Previous connections

POWM / OCTAVE — Eight-observer predictive memory architecturereading (verified)

Source & provenance

Access: Public
Dates declared in the record: Not recorded
Project basis: Explicit project source citation
Origin: d0re research collection (snapshot 1 October 2026)
Original path: journey/sources/octave-school-benchmark.txt
Authorship basis: Collection attribution; original record retains its own credits
Title basis: Editorial title
Date: 01 Oct 2026 · saved
Status basis: This is the latest OCTAVE result in the corpus. r-288afeeb8dda closes with 'No level is automatically unlocked by these results.' No later record re-runs it.
Size: 6,635 bytes
Word count: 921 words
SHA-256 of indexed source: 288afeeb8dda6449f735745fb14913d78be023a16fdd1f330c6448e32d11811f

Open the source file ↗

View citation

OCTAVE — School training and benchmark results
2026-10-01
d0re research record r-288afeeb8dda
https://www.d0re.com/library/records/r-288afeeb8dda.html
Authorship basis: Collection attribution; original record retains its own credits
Date basis: saved
Source SHA-256: 288afeeb8dda6449f735745fb14913d78be023a16fdd1f330c6448e32d11811f