diff options
| author | Yuren Hao <yurenh2@illinois.edu> | 2026-07-16 17:22:04 -0500 |
|---|---|---|
| committer | Yuren Hao <yurenh2@illinois.edu> | 2026-07-16 17:22:04 -0500 |
| commit | 0b5e8659ba0d053e8635589b1bf6f574afab8d2f (patch) | |
| tree | 04a286f438b0c8800c88d46f9e25d2721168fcac /docs/campaign | |
| parent | 44e4238f7b3e1bca91faa7aec5e78109d7dd2966 (diff) | |
RESULT 33: triple verdict — A/B neutral at 72M, dose-response monotone thru 1e-2, ride governor beats all static betas (1.2562); stage1b_ride launched
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014FAPDWQ49M5Ye3NpTndTpn
Diffstat (limited to 'docs/campaign')
| -rw-r--r-- | docs/campaign/CASCADE_ABLATION_PLAN.md | 17 |
1 files changed, 17 insertions, 0 deletions
diff --git a/docs/campaign/CASCADE_ABLATION_PLAN.md b/docs/campaign/CASCADE_ABLATION_PLAN.md index bd5ec4f..84197bc 100644 --- a/docs/campaign/CASCADE_ABLATION_PLAN.md +++ b/docs/campaign/CASCADE_ABLATION_PLAN.md @@ -568,6 +568,23 @@ direction), not training-under-fault; wave-2 = co-training with faults injected +### RESULT 33 (2026-07-16): TRIPLE VERDICT — centered CE-neutral at 72M too; CE-vs-beta monotone through 1e-2; the RIDE GOVERNOR beats every static beta. +1. 72M A/B (resume fw72m_cent s175000 -> 185000, single-GPU B24, identical schedule): + plain 3.4288 vs centered 3.4289 — IDENTICAL. The 42M decoupling replicates at crown scale: + the recipe does not need centered. plain + governor = 1.0x cost at 72M. +2. Static beta dose-response (plain, 42M tail): 2e-3 1.2611 / 3e-3 1.2590 / 5e-3 1.2578 / + 1e-2 1.2571; centered@1e-2 1.2569 (neutral again). CE STILL IMPROVING at 1e-2 — no bias bite + anywhere in the explored range; the 42M window extends beyond 1e-2. +3. arm_ride30 (two-sided governor, 30x headroom): best 1.2562 — BEATS every static arm. The + governor surfed dynamically: 3e-3 -> ~6e-2 excursions mid-tail -> taper to ~7e-4 at the end, + ZERO guard events, cos 0.992 held. It discovered a dynamic beta schedule no hand-tuning found + (high-beta bulk + late taper aligned with the cosine-LR endgame). +- RECIPE CONSEQUENCE: estimator upgrades (centered/centfast/centmirror) demote to cos-telemetry + tools and insurance; the working recipe trends to PLAIN single-sided + ride governor = 1.0x + estimator cost. 8B ledger returns to the 3.2x base. +- Full-epoch validations in flight: stage1b_plain3e3 (46k, tracking cent-full within 0.004), + stage1b_est15 (26k), stage1b_ride (LAUNCHED, GPU3 — the recipe candidate). + ### RESULT 32 (2026-07-16): CROWN RESEALED — fw72m_cent 3.3318 vs BP 3.2884: honest gap 0.043, 10x below the blown-schedule number. fw72m_cent complete: 234k steps / 1.44B FineWeb tokens, from scratch, fully BP-free, DDP 2xA6000. | | best val CE | note | |