# Эксперимент: Markov-coded synchronization-robust latent streams (#1242)

{
  "worked": true,
  "confidence": 6,
  "verdict": "Built a reproducible iid-versus-Markov latent-stream benchmark with exact insertion-capacity/rate computation, sampled subsequence list-collision checks, and a small GRU classifier trained on inserted streams. At delta=0.25, n=12, M=8, the rate 0.25 satisfies the 10% safety-margin bound 0.3128; q=0.25 improved accuracy at 25% insertions (0.950 vs 0.936) and 50% insertions (0.826 vs 0.714), despite slightly more sampled list collisions. The signal is promising but q-dependent: q=0.4 improved 25% robustness but was worse than iid at 50%, so this is not evidence of a universal win.",
  "metrics": {
    "baseline": "iid q=0.5: clean 1.000, 25% insertion 0.936, 50% insertion 0.714; sampled mean list size 1.084, collision fraction 0.083",
    "idea": "Markov q=0.25: clean 1.000, 25% insertion 0.950, 50% insertion 0.826; sampled mean list size 1.151, collision fraction 0.148. Capacity(delta=.25)=0.3476, safe rate at epsilon=.1=0.3128, tested rate=.25."
  },
  "how_to_run": "/home/maxwelhelp/main/bin/python3 experiment.py",
  "files": [
    "experiment.py",
    "results.json"
  ],
  "limitations": "Only a tiny synthetic binary identification task was tested, with n=12, M=8, two random repetitions per configuration, and a GRU rather than a transformer. The experiment used random insertions only, did not learn q, did not add the Markov negative-log-prior objective, and did not establish asymptotic list-decodability or FLOP/sample-efficiency gains. CUDA cuDNN allocation failed and the training automatically fell back to CPU."
}