gustavlangstroem
/

Microexpert_NG

Text Generation

Model card Files Files and versions

gustavlangstroem commited on 4 days ago

Commit

c2b229f

·

verified ·

1 Parent(s): 55a2700

Update README.md

Files changed (1) hide show

README.md +1 -1

README.md CHANGED Viewed

@@ -210,7 +210,7 @@ Implemented in Python/PyTorch. No custom CUDA required except BitLinear kernels
 | Risk | Mitigation |
 |------|-----------|
 | **Cannibalization signal too noisy** | Dual-EMA smoothing validated by GCond [6]; cooldown timer; min age |
-| **Merge collapse** still no soloution want to avoid replay buffer|
 | **Router instability** | Embedding continuity on split; cooldown between events |
 | **Expert starvation at 100K** | Death mechanism; pressure system|
 | **Split/merge oscillation** | min age before merge; hysteresis; cooldown |

 | Risk | Mitigation |
 |------|-----------|
 | **Cannibalization signal too noisy** | Dual-EMA smoothing validated by GCond [6]; cooldown timer; min age |
+| **Merge collapse** |still no soloution want to avoid replay buffer|
 | **Router instability** | Embedding continuity on split; cooldown between events |
 | **Expert starvation at 100K** | Death mechanism; pressure system|
 | **Split/merge oscillation** | min age before merge; hysteresis; cooldown |