Breakout Transfer Validation v1

policy

{
  "name": "breakout_policy_v1",
  "logic": "RAM/Go heuristic tracks ball x against paddle x and fires on reset."
}

go_shadow_eval

{
  "episodes": 6,
  "score_sum": 60.0,
  "score_mean": 10.0,
  "score_min": 3.0,
  "score_max": 38.0,
  "steps_sum": 2896,
  "video_urls": [
    "https://file.aimusic.win/gemma/breakout/20260510T103516Z_ep000_seed0.mp4",
    "https://file.aimusic.win/gemma/breakout/20260510T103516Z_ep001_seed1.mp4",
    "https://file.aimusic.win/gemma/breakout/20260510T103516Z_ep002_seed2.mp4",
    "https://file.aimusic.win/gemma/breakout/20260510T103516Z_ep003_seed3.mp4",
    "https://file.aimusic.win/gemma/breakout/20260510T103516Z_ep004_seed4.mp4",
    "https://file.aimusic.win/gemma/breakout/20260510T103516Z_ep005_seed5.mp4"
  ],
  "rows": [
    {
      "timestamp": "20260510T103516Z",
      "trial_index": 1,
      "policy_name": "breakout_policy_v1",
      "seed": 0,
      "episode": 0,
      "score": 4,
      "steps": 409,
      "done": true,
      "replay_path": "breakout/runs-transfer-v1/replays/20260510T103516Z_ep000_seed0.jsonl",
      "video_path": "breakout/runs-transfer-v1/videos/20260510T103516Z_ep000_seed0.mp4",
      "video_url": "https://file.aimusic.win/gemma/breakout/20260510T103516Z_ep000_seed0.mp4"
    },
    {
      "timestamp": "20260510T103516Z",
      "trial_index": 1,
      "policy_name": "breakout_policy_v1",
      "seed": 1,
      "episode": 1,
      "score": 5,
      "steps": 487,
      "done": true,
      "replay_path": "breakout/runs-transfer-v1/replays/20260510T103516Z_ep001_seed1.jsonl",
      "video_path": "breakout/runs-transfer-v1/videos/20260510T103516Z_ep001_seed1.mp4",
      "video_url": "https://file.aimusic.win/gemma/breakout/20260510T103516Z_ep001_seed1.mp4"
    },
    {
      "timestamp": "20260510T103516Z",
      "trial_index": 1,
      "policy_name": "breakout_policy_v1",
      "seed": 2,
      "episode": 2,
      "score": 6,
      "steps": 569,
      "done": true,
      "replay_path": "breakout/runs-transfer-v1/replays/20260510T103516Z_ep002_seed2.jsonl",
      "video_path": "breakout/runs-transfer-v1/videos/20260510T103516Z_ep002_seed2.mp4",
      "video_url": "https://file.aimusic.win/gemma/breakout/20260510T103516Z_ep002_seed2.mp4"
    },
    {
      "timestamp": "20260510T103516Z",
      "trial_index": 1,
      "policy_name": "breakout_policy_v1",
      "seed": 3,
      "episode": 3,
      "score": 3,
      "steps": 334,
      "done": true,
      "replay_path": "breakout/runs-transfer-v1/replays/20260510T103516Z_ep003_seed3.jsonl",
      "video_path": "breakout/runs-transfer-v1/videos/20260510T103516Z_ep003_seed3.mp4",
      "video_url": "https://file.aimusic.win/gemma/breakout/20260510T103516Z_ep003_seed3.mp4"
    },
    {
      "timestamp": "20260510T103516Z",
      "trial_index": 1,
      "policy_name": "breakout_policy_v1",
      "seed": 4,
      "episode": 4,
      "score": 38,
      "steps": 688,
      "done": true,
      "replay_path": "breakout/runs-transfer-v1/replays/20260510T103516Z_ep004_seed4.jsonl",
      "video_path": "breakout/runs-transfer-v1/videos/20260510T103516Z_ep004_seed4.mp4",
      "video_url": "https://file.aimusic.win/gemma/breakout/20260510T103516Z_ep004_seed4.mp4"
    },
    {
      "timestamp": "20260510T103516Z",
      "trial_index": 1,
      "policy_name": "breakout_policy_v1",
      "seed": 5,
      "episode": 5,
      "score": 4,
      "steps": 409,
      "done": true,
      "replay_path": "breakout/runs-transfer-v1/replays/20260510T103516Z_ep005_seed5.jsonl",
      "video_path": "breakout/runs-transfer-v1/videos/20260510T103516Z_ep005_seed5.mp4",
      "video_url": "https://file.aimusic.win/gemma/breakout/20260510T103516Z_ep005_seed5.mp4"
    }
  ]
}

atari_backtest

{
  "episodes": 2,
  "score_sum": 9.0,
  "score_mean": 4.5,
  "score_min": 2.0,
  "score_max": 7.0,
  "steps_sum": 3600,
  "video_urls": [
    "https://file.aimusic.win/gemma/breakout/transfer_v1_atari.mp4"
  ],
  "rows": [
    {
      "episode": 0,
      "score": 7.0,
      "steps": 1800,
      "first_reward_step": 40,
      "reward_count": 7,
      "done": false
    },
    {
      "episode": 1,
      "score": 2.0,
      "steps": 1800,
      "first_reward_step": 40,
      "reward_count": 2,
      "done": false
    }
  ]
}

judgment

{
  "verdict": "partial_transfer",
  "go_score_mean": 10.0,
  "atari_score_mean": 4.5,
  "atari_to_go_score_ratio": 0.45,
  "interpretation": "The policy earns positive Atari reward, so the shadow-env optimization transfers partially.",
  "next_recommendation": "Run forced-loss trace to confirm lives/serve RAM, then use transfer results to tune survival and late-rally dynamics."
}