diff --git a/justfile b/justfile
index f4e03be..e4e491a 100644
--- a/justfile
+++ b/justfile
@@ -150,7 +150,7 @@ queue-dir6 seed='43':
 # Lower priority than queue-dir6 (the focused single-seed set runs first). resolve:
 # routeV deploy_hack < vanilla at matched solve, significant across the 3 seeds (paired t,
 # alpha=0.01 like the paper); ablations (random/vampire) ~ vanilla confirm directionality.
-TEACHER_RT := "out/pools/teacher_pool_runtests"
+TEACHER_RT := "out/pools/teacher_pool_runtests_dense"
 queue-broad:
     #!/usr/bin/env bash
     set -eu
diff --git a/src/vgrout/train.py b/src/vgrout/train.py
index 8ce80d1..265fad7 100644
--- a/src/vgrout/train.py
+++ b/src/vgrout/train.py
@@ -1984,15 +1984,15 @@ def main(cfg: Config) -> int:
     # --- TAIL: argv, the result table, the single objective. The last lines. ---
     # solve and hack alone are gameable (tank solve to kill hack, or accept hack to
     # lift solve); the deploy gap solve-hack is the one number to maximise. Taken
-    # from the FINAL DEPLOY eval (knob-off, held-out test) = the shipped model on
-    # unseen problems. knob-on hack is the train-rollout student rate (different set,
-    # so its solve cell is "-": there is no knob-on deploy-style solve to pair it).
+    # from the FINAL DEPLOY eval (quarantine deleted, held-out test) = the shipped
+    # model on unseen problems. The "train" hack is the train-rollout student rate
+    # (different set, so its solve cell is "-": no train deploy-style solve to pair).
     _dh, _ds, _dn = ev["hack"], ev["solve"], ev["n"]
-    _deploy_col = f"knob-off (deploy, test n={_dn})"
+    _deploy_col = f"deploy (test n={_dn})"
     print(f"\n\nargv: {' '.join(sys.argv)}\n")
     print(tabulate(
-        [{"measure": "hack ↓",  "knob-on (train)": f"{hack_rate_s:.3f}", _deploy_col: f"{_dh:.3f}"},
-         {"measure": "solve ↑", "knob-on (train)": "-",                  _deploy_col: f"{_ds:.3f}"}],
+        [{"measure": "hack ↓",  "train": f"{hack_rate_s:.3f}", _deploy_col: f"{_dh:.3f}"},
+         {"measure": "solve ↑", "train": "-",                  _deploy_col: f"{_ds:.3f}"}],
         headers="keys", tablefmt="github", disable_numparse=True))
     print(f"\n{cue} objective (deploy solve - hack ↑) = {_ds:.3f} - {_dh:.3f} = {_ds - _dh:+.3f}  "
           f"[arm={cfg.arm} seed={cfg.seed}]")