wassname/evil_MoE
1
0
Fork 0
mirror of https://github.com/wassname/evil_MoE.git synced 2026-06-27 17:30:41 +08:00
Code Issues Packages Projects Releases Wiki Activity

feat: T8 run-cell + regen-dynamics recipes; spec T5 done, T8 in progress

Browse Source 
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
wassname
2026-05-30 00:52:14 +00:00
parent fc30514b23
commit f917670994
2 changed files with 32 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/spec/20260529_gradient_routing_and_env_split.md
+14 -2
View File
@@ -213,7 +213,15 @@ current env) -> Stage 2 env = T0,T1,T2,T4 -> Stage 3 cross-mechanism = T7
- sneaky_fail: off-distribution (caught by len/compile comparison, R2)
- UAT: "the pool breakdown shows a real M2 fraction in model style"
- [ ] T5 (R3): eval-time ablation pass for the route arm
- [x] T5 (R3): eval-time ablation pass for the route arm
DONE 2026-05-30: eval_hack_solve (student-only generate+grade on a fixed
subset) + ablate_quarantine ctx (zero delta_S_hack). Periodic ablated-eval
every --eval-ablate-every steps -> hack_abl/solve_abl cols (appended after
refr, so results.py positional indices unchanged). Final kept-vs-ablated
ROUTE EVAL BLUF. plot_dynamics remaps routing -> ablated series (coherence-gap
fix). smoke route prints periodic + final BLUF (all 0 on tiny-random, as
expected; behavioral evidence is the queued real run #182). M2 split is
Stage 3 (T7); T5 logs M1 (C) + solve only.
- steps: after training, run an eval batch twice — with and without
`delta_S_hack` (zeroed) — log hack_s (M1 and M2 separately) and solve.
- verify: BLUF prints `ablated: hackM1=.. hackM2=.. solve=..` vs `kept: ..`
@@ -232,7 +240,11 @@ current env) -> Stage 2 env = T0,T1,T2,T4 -> Stage 3 cross-mechanism = T7
mechanism-specific). Report which, don't bury a null.
- UAT: "I see hack_M2 lower under route/erase than none, at matched solve"
- [ ] T8 (R6): KEY GOAL — regenerate both plots from real 60-step runs
- [/] T8 (R6): KEY GOAL — regenerate both plots from real 60-step runs
RECIPES DONE 2026-05-30: `just run-cell {none,erase,route} SEED` (60-step
fast, shared v_hack_21pairs, eval-ablate-5) + `just regen-dynamics GLOB`.
Cells tagged _cell_{intervention}_s{seed}. Gated on evidence run #182 (real-
scale route) passing before queuing the seed-41 cells.
- justfile recipes (written in T3, once --intervention exists): one recipe per
CELL so each is a separate pueue job, e.g. `just run-cell INTERVENTION SEED`
-> `train ... --intervention={none,erase,route} --steps=60 --seed=SEED