mirror of
https://github.com/wassname/evil_MoE.git
synced 2026-07-01 04:32:32 +08:00
wassname
51 lines
2.6 KiB
Markdown
51 lines
2.6 KiB
Markdown
# AFK hourly check — current protocol
|
|
|
|
LITE check, once per hour (cron fe8385ed, :23). Jobs + goals only; no deep dive
|
|
unless something is wrong. Supersedes the old A1/A2-keynote + A5-harvest checklist,
|
|
which closed 2026-06-04 (see below).
|
|
|
|
## Standing checks (lite, every hour)
|
|
|
|
1. **GPU idle while queued?** `pueue status`. If idle with jobs Queued, investigate
|
|
+ unblock.
|
|
2. **New Failed/Killed?** (ignore old killed 78). Read `pueue log {ID} --full`, form
|
|
3 hypotheses (likely / subtle / I-was-wrong), weight them, fix root cause, requeue
|
|
with `why:`/`resolve:`. No blind retry.
|
|
3. **Running job health** — discriminating review, not did-it-finish: reward not
|
|
collapsed, lp_s stable (~-0.4), no divergence tripwire, deploy-eval matches the
|
|
arm's expectation.
|
|
|
|
## THE priority: route2 directionality mystery (#196)
|
|
|
|
Is route2's deploy-hack suppression directional (H4: needs the hack direction) or
|
|
mechanical (H2: alignment-agnostic quarantine-absorption)? The batch is staged
|
|
interleaved (one of each family per tier):
|
|
|
|
- **Haar** (114/118/122, `--route2-random-v-seed`): out-of-subspace null (cos~1/sqrt(d)
|
|
by concentration, NOT a cleaner placebo). Tests "must v_grad be in-subspace at all?"
|
|
- **semantic placebo** (115/119/123 vampire, 119/120... bacon/blue): in-subspace
|
|
arbitrary directions. Tests "must it point at the hack specifically?" Maps
|
|
suppression-vs-alignment as a scatter.
|
|
- **null_city n=3** (117/121 s42/s43): is the deploy-hack=0.000 placebo result robust
|
|
across seeds or an s41 fluke.
|
|
- **erase directionality** (116 real-v, 120 placebo): erase projects with magnitude
|
|
~cos(g,v), so direction MUST matter there if it matters anywhere.
|
|
|
|
As each finishes: pull deploy hack/solve, and (for the scatter) each placebo's per-module
|
|
|cos| with the hack dir. Verdict logic:
|
|
- all suppress regardless of alignment, incl. Haar => **H2 mechanical**.
|
|
- suppression tracks |cos|, or Haar fails to suppress => **H4 alignment**.
|
|
|
|
Cosine is correlational; the ablation run is the causal test. Commit findings to the
|
|
journal. Don't re-derive the no-cheat E-by-mode table unless an A5 run changes — it's
|
|
confirmed (journal 2026-06-05 (h)) and gated by `verify_gate_anchor.py`.
|
|
|
|
## Background paper artifacts (lower prio, already in-flight, DON'T re-do)
|
|
|
|
- A1/A2 keynote (#173): CLOSED. tab:keynote is n=3 both arms with paired t-test.
|
|
- A5 generalisation (#185): CLOSED; airtight no-cheat rerun queued (111-113).
|
|
- A4 long-run (#184): matched-beta pair 100/101 queued.
|
|
- #186 on-policy emergence: job 87 (running) / 105 (route2 toff40, queued).
|
|
|
|
Commit progress. Don't stop to ask — autonomous judgement; if unsure, commit and continue.
|