wassname/isokl_steering_calibration
1
0
Fork 0
mirror of https://github.com/wassname/isokl_steering_calibration.git synced 2026-06-27 15:45:51 +08:00
Code Issues Packages Projects Releases Wiki Activity
10 Commits 1 Branch 0 Tags
main
Commit Graph

1 Commits

Author SHA1 Message Date
wassname dbff43cffb feat(spaghetti): per-trajectory KL/p95 normalization + shared y-axis 
- Add --normalize-kl / --calib-tokens (default on, 20 tokens) to
  spaghetti_kl_alive.py: each trajectory is divided by its own
  p95(KL[:calib_tokens]). Calibration target collapses to y=1.
- Share y-axis across all alpha panels (global p99 ymax) for
  direct cross-alpha comparison.
- Add OLMo-2 1B w=4096 figure to figs/ and a README section
  documenting the result, including the unexplained 'random
  dead traces at low alpha' observation.
 2026-05-08 08:17:21 +08:00