exp76: full phrase probes [48,36,28,20,16] (PR openai#880 set)

Each additional probe length adds ~0.005 BPB.
probe[28] → -0.007, probe[36] → -0.005.
Testing if probe[48] captures even longer verbatim patterns.

Author: sofiabod

results.tsv

@@ -46,3 +46,4 @@ b224b23	1.1323	15.88	keep	TTT AdamW 5ep lr=0.0005 DDP-synced	BEATS SOTA! sw=1.13
 49aaca9	0.2532	15.67	discard	hierarchical Dirichlet c=5.0 + order-13	no improvement, eval=627s OVER BUDGET
 1a8ee89	0.2534	15.26	discard	hierarchical Dirichlet c=0.5 order-9	slightly worse than c=1.0, eval=532s
 cd10ecb	0.2463	15.39	keep	flat Dirichlet c=1.0 + phrase[28,20,16]	NEW BEST! phrase[28] adds -0.007, eval=529s
+fc5f627	0.2417	15.39	keep	flat Dirichlet c=1.0 + phrase[36,28,20,16]	NEW BEST! phrase[36] adds -0.005, eval=548s

train_gpt.py

@@ -896,7 +896,7 @@ def eval_val_sliding(
 class LongPhraseCache:
     """variable-length suffix matcher for verbatim repetition (PR #880).
     probes at lengths [48,36,28,20,16] using rolling hashes."""
-    PROBE_LENGTHS = [36, 28, 20, 16]  # extended probes for more phrase matching
+    PROBE_LENGTHS = [48, 36, 28, 20, 16]  # full probe set (matching PR #880)
     PRIMES = [np.uint64(p) for p in [
         36313, 27191, 51647, 81929, 131071, 174763, 233017, 299993, 350377,
         412391, 479909, 541267, 613651, 700897, 786433, 850001, 921587,