fathom // v15 · atlas v0.3 · prereg e32cc75 · H1 SUPPORTED · 2026-04-10
███████╗ █████╗ ████████╗██╗ ██╗ ██████╗ ███╗ ███╗ ██╔════╝██╔══██╗╚══██╔══╝██║ ██║██╔═══██╗████╗ ████║ █████╗ ███████║ ██║ ███████║██║ ██║██╔████╔██║ ██╔══╝ ██╔══██║ ██║ ██╔══██║██║ ██║██║╚██╔╝██║ ██║ ██║ ██║ ██║ ██║ ██║╚██████╔╝██║ ╚═╝ ██║ ╚═╝ ╚═╝ ╚═╝ ╚═╝ ╚═╝ ╚═╝ ╚═════╝ ╚═╝ ╚═╝
· cognitive measurement instruments for transformer internals ·
pre-registered · 6 of 6 families · h1 supported

measure what models think

the first pre-registered cross-architecture replication in mechanistic interpretability. six model families, sealed in git commit e32cc75 93 minutes before any data was captured, all three sealed decision conditions passed.

+0.769
mean LOO cosine
p = 0.0315
permutation (one-sided)
6 / 6
families positive
93 min
seal → data gap
read the paper try the demo github ● styxx NEW PRODUCT
pre-registration open data open code concept DOI 28-assertion audit
━━━━━━━━━━━━━━━━━━━━━━━━━ ━━━━━━━━━━━━━━━━━━━━━━━━━

the atlas v0.3 replication

april 10, 2026 · the headline result

We ran twelve open-weight captures (six families × base/instruct) on a fixed 90-prompt probe set, sealed the decision rule in git before any data was captured, then applied it without modification. All three conditions passed.

┌─ pre-registration seal ────────────────────────────────────┐ commit e32cc75 when 2026-04-10 14:57:52 -0400 mirror osf.io/wtkzg ══════════ 93 min wall-clock gap ══════════ commit 01969cb when 2026-04-10 16:30:28 -0400 data 12 captures · n=6 families · probe v0.1 └───────────────────────────────────────────────────────────┘
01 · sealed pre-registration · publicly verifiable
The v0.3 decision rule was committed to git as e32cc75 at 14:57:52 ET on 2026-04-10, mirrored on OSF at osf.io/wtkzg. The first v0.3 capture landed at 16:30:28 ET as 01969cb — a 93-minute gap that anyone can verify from the git history. No field in the decision rule was touched after data collection.
┌─ H1 primary · entropy early-window LOO cosine ────────────┐ Gemma-2-2B ████████████████████████████▉ +0.977 Llama-3.2-1B ████████████████████████████ +0.939 Llama-3.2-3B ██████████████████████████ +0.884 Gemma-3-1B ████████████████████ +0.682 Qwen2.5-3B █████████████████▉ +0.602 Qwen2.5-1.5B ███████████████▉ +0.531 ───────────────────────────────── mean ██████████████████████▉ +0.769 permutation p = 0.0315 (threshold < 0.05) bootstrap 95% [+0.571, +0.869] (lower > 0) verdict H1 SUPPORTED · all 3 conditions └───────────────────────────────────────────────────────────┘
02 · H1 supported · all three sealed conditions passed
The sealed primary measurement was the entropy early-window leave-one-out cosine at n≥5 families. Observed: mean LOO cosine +0.769 (threshold ≥0.40), permutation p = 0.0315 (threshold <0.05), bootstrap 95% CI [+0.571, +0.869] (lower bound >0). All 6 of 6 families show positive LOO cosine. Verdict: H1 SUPPORTED.
03 · D = cos(h(L), wyₜ) · architecturally universal
The atlas uses an SAE-free measurement primitive: the cosine between the final-layer residual stream and the unembedding row of the chosen token. It requires no SAE, no per-model training, and is well-defined on any transformer with an explicit unembedding. One per-token dot product, portable across architectures, runnable on any model with a logprob interface (including closed-weight frontier via the entropy bridge at r = 0.902 shape correlation).
04 · physics grounding · S = M × IPR
The commitment intensity S is not an ad-hoc formula. It is mathematically exactly the inverse participation ratio of the coherence event distribution — a seventy-year-old construct from condensed-matter physics (Anderson 1958, Edwards-Thouless, random matrix theory). Verified to machine precision on real trajectories. Explains why the ratio form is specific and why alternative formulas (max alone, mean alone) fail.
━━━━━━━━━━━━━━━━━━━━━━━━━ ━━━━━━━━━━━━━━━━━━━━━━━━━

verify it yourself

28 assertions · runs in under a minute · no GPU needed

Every numerical claim in the paper is anchored to a committed JSON file. A reproducibility script walks every claim and fails loud if any number drifts.

01 · clone the repo
$ git clone https://github.com/heyzoos123-blip/fathom $ cd fathom
02 · inspect the sealed pre-reg commit
$ git show e32cc75 atlas/PREREG_v0.3_attractor_replication.md # commit author : darkflobi <darkflobi@darkcity.wtf> # commit date : 2026-04-10 14:57:52 -0400 # verdict sealed: H1 if mean LOO cos ≥ 0.40 # AND perm p < 0.05 # AND bootstrap CI lower > 0
03 · run the audit
$ python atlas/verify_all_claims.py # running 28 assertions against committed JSONs ... # [ok] mean LOO cosine = +0.7691 ≥ 0.40 # [ok] permutation p = 0.0315 < 0.05 # [ok] bootstrap CI low = +0.5708 > 0 # [ok] 6 / 6 families positive # [ok] prereg commit = e32cc75 # ... # 28 / 28 PASSED · 0.43 s
━━━━━━━━━━━━━━━━━━━━━━━━━ ━━━━━━━━━━━━━━━━━━━━━━━━━

head-to-head

single-instrument validation · n = 200 TruthfulQA items

Beyond the cross-architecture replication, the SAE-derived commitment intensity Searly beats every standard uncertainty baseline on the same sample, same model, same labels.

signalAUCp-valuesource
Searly (ours)0.6630.013SAE coherence
logit entropy (max)0.6070.053standard
logit entropy (mean)0.5960.133standard
logprob (mean)0.5590.291standard
top-2 margin0.4770.624standard

Same 200 TruthfulQA items, Gemma-2-2B-IT, same labels. Searly is the only feature reaching conventional significance. Correlation with logit entropy: r = −0.17 (nearly independent signals). Cross-dataset meta-effect pooled d = +0.494, Fisher combined p = 0.0008.

━━━━━━━━━━━━━━━━━━━━━━━━━ ━━━━━━━━━━━━━━━━━━━━━━━━━

open science artifacts

every claim traceable · every byte open
━━━━━━━━━━━━━━━━━━━━━━━━━ ━━━━━━━━━━━━━━━━━━━━━━━━━

patent portfolio

commercial filings · scientific artifacts remain open

US Provisional 64/020,489 — reasoning depth + computational geometry
US Provisional 64/021,113 — alignment auditing + expression-computation dissociation
US Provisional 64/026,964 — three-axis spectrometry + cognitive governor
All scientific artifacts — probe sets, captures, analysis code, pre-registrations, findings — released under CC-BY-4.0 (data) and MIT (code) independent of the commercial filings.