Cubric cadence accumulator: N/N/N/C pattern for n-gram table optimization

The real Cubric concept: periodic C-step fires every K batches to optimize
the n-gram hash tables using neural model knowledge from already-scored data.

C-step operations (all score-first legal):
1. Decay stale counts (make room for fresh patterns)
2. Boost confident: inject +1 count where model AND ngram agree (>0.5, >0.3)
3. Prune noisy: zero out high-collision buckets (hit rate < 1% with >20 counts)
4. Reweight orders: scale counts by per-order accuracy (±5%)

4-arm A/B: control vs cadence 4/10/20

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: Octavian

concepts/cubric_ngram/run_cadence_ab.sh

@@ -0,0 +1,109 @@
+#!/bin/bash
+set -euo pipefail
+# ══════════════════════════════════════════════════════════════
+# CUBRIC CADENCE ACCUMULATOR — N/N/N/C pattern
+#
+# HYPOTHESIS: Periodic neural optimization of n-gram hash tables
+# will improve BPP over static tables. The C-step uses already-scored
+# data to: (1) decay stale counts, (2) boost patterns where model and
+# n-gram agree, (3) prune noisy hash collisions, (4) reweight orders
+# by tracked accuracy. This transforms the n-gram system from a
+# static counter into an adaptive pattern reservoir.
+#
+# EXPECTED: 0.003-0.010 BPP improvement over baseline n-gram.
+# The improvement should grow over the eval pass as the C-step
+# accumulates more signal about the document.
+#
+# RISK: C-step could corrupt the tables if pruning/boosting is
+# miscalibrated. Count decay could erase good patterns.
+#
+# ARMS:
+#   A: Baseline (n-gram, no cubric)
+#   B: Cubric cadence=4 (C every 4 batches, frequent optimization)
+#   C: Cubric cadence=10 (C every 10 batches, balanced)
+#   D: Cubric cadence=20 (C every 20 batches, conservative)
+#
+# Score-first legal: C-step only reads from already-scored segments.
+# ══════════════════════════════════════════════════════════════
+
+SCRIPT_DIR="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" && pwd)"
+REPO_ROOT="$(cd -- "${SCRIPT_DIR}/../.." && pwd)"
+cd "${REPO_ROOT}"
+
+if [ -d "flash-attention/hopper" ]; then
+    export PYTHONPATH="${REPO_ROOT}/flash-attention/hopper:${PYTHONPATH:-}"
+elif [ -d "local_shims" ]; then
+    export PYTHONPATH="${REPO_ROOT}/local_shims:${PYTHONPATH:-}"
+fi
+
+SEED="${SEED:-1337}"
+NPROC="${NPROC_PER_NODE:-8}"
+TRAIN_SCRIPT="${SCRIPT_DIR}/train_gpt_cadence.py"
+
+COMMON_ENV=(
+  SEED="${SEED}"
+  MLP_ACT=leaky_relu_sq MLP_LEAKY_SLOPE=0.5
+  XSA_LAST_N=4 BIGRAM_VOCAB_SIZE=1536
+  ROPE_DIMS=24
+  COMPILE_ENABLED=1 COMPILE_FULLGRAPH=0
+  NGRAM_EVAL_ORDER=5
+  NGRAM_EVAL_ALPHA=0.30
+  NGRAM_EVAL_MIN_COUNT=2
+  NGRAM_EVAL_BUCKETS=4194304
+  NGRAM_EVAL_ADAPTIVE=1
+  NGRAM_EVAL_ALPHA_MIN=0.05
+  NGRAM_EVAL_ALPHA_MAX=0.60
+)
+
+run_arm() {
+    local arm_id="$1"
+    local desc="$2"
+    shift 2
+    local run_id="cubcad_${arm_id}_s${SEED}_$(date +%Y%m%d_%H%M%S)"
+    echo ""
+    echo "═══════════════════════════════════════"
+    echo "  [${arm_id}] ${desc}"
+    echo "  RUN_ID: ${run_id}"
+    echo "═══════════════════════════════════════"
+    env "${COMMON_ENV[@]}" "$@" \
+      RUN_ID="$run_id" \
+      torchrun --standalone --nproc_per_node="$NPROC" \
+        "$TRAIN_SCRIPT" \
+        2>&1 | tee "logs/${run_id}.log"
+    echo "── [${arm_id}] result ──"
+    grep -E "final_int6_sliding_window_ngram.*exact|c_steps=" \
+      "logs/${run_id}.log" 2>/dev/null | tail -3
+    echo ""
+}
+
+mkdir -p logs
+
+echo "══════════════════════════════════════════════════"
+echo "  CUBRIC CADENCE — N/N/N/C ACCUMULATOR A/B"
+echo "══════════════════════════════════════════════════"
+
+run_arm "A" "CONTROL: static n-gram, no cubric" \
+  CUBRIC_CADENCE=0
+
+run_arm "B" "H: C every 4 batches (aggressive optimization)" \
+  CUBRIC_CADENCE=4 CUBRIC_COUNT_DECAY=0.02 \
+  CUBRIC_BOOST_CONFIDENT=1 CUBRIC_PRUNE_NOISY=1 CUBRIC_REWEIGHT_ORDERS=1
+
+run_arm "C" "H: C every 10 batches (balanced)" \
+  CUBRIC_CADENCE=10 CUBRIC_COUNT_DECAY=0.02 \
+  CUBRIC_BOOST_CONFIDENT=1 CUBRIC_PRUNE_NOISY=1 CUBRIC_REWEIGHT_ORDERS=1
+
+run_arm "D" "H: C every 20 batches (conservative)" \
+  CUBRIC_CADENCE=20 CUBRIC_COUNT_DECAY=0.02 \
+  CUBRIC_BOOST_CONFIDENT=1 CUBRIC_PRUNE_NOISY=1 CUBRIC_REWEIGHT_ORDERS=1
+
+echo "══════════════════════════════════════════════════"
+echo "  SUMMARY"
+echo "══════════════════════════════════════════════════"
+for f in logs/cubcad_*_s${SEED}_*.log; do
+    arm=$(basename "$f" | sed 's/cubcad_\([A-D]\)_.*/\1/')
+    bpb=$(grep "final_int6_sliding_window_ngram.*exact" "$f" 2>/dev/null | grep -oP 'val_bpb:\K[0-9.]+' || echo "N/A")
+    csteps=$(grep -oP 'c_steps=\K[0-9]+' "$f" 2>/dev/null | tail -1 || echo "0")
+    echo "  [$arm] sliding_ngram_bpb=$bpb c_steps=$csteps"
+done
+echo "══════════════════════════════════════════════════"