{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.009718172983479106,
  "eval_steps": 500,
  "global_step": 90,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "completion_length": 135.5,
      "epoch": 0.00010797969981643452,
      "grad_norm": 3.9876437187194824,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": -0.0,
      "reward": 3.65625,
      "reward_std": 0.3837142586708069,
      "rewards/gpt4o_acoustic_reward": 1.78125,
      "rewards/gpt4o_semantic_reward": 1.875,
      "step": 1
    },
    {
      "completion_length": 86.25,
      "epoch": 0.00021595939963286903,
      "grad_norm": 5.973817348480225,
      "kl": 0.0,
      "learning_rate": 5.000000000000001e-07,
      "loss": 0.0,
      "reward": 3.8125,
      "reward_std": 0.5610761046409607,
      "rewards/gpt4o_acoustic_reward": 1.96875,
      "rewards/gpt4o_semantic_reward": 1.84375,
      "step": 2
    },
    {
      "completion_length": 136.0,
      "epoch": 0.00032393909944930353,
      "grad_norm": 4.262232780456543,
      "kl": 0.00010204315185546875,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 0.0,
      "reward": 3.84375,
      "reward_std": 0.41466206312179565,
      "rewards/gpt4o_acoustic_reward": 1.875,
      "rewards/gpt4o_semantic_reward": 1.96875,
      "step": 3
    },
    {
      "completion_length": 86.5,
      "epoch": 0.00043191879926573806,
      "grad_norm": 14.690574645996094,
      "kl": 0.00030517578125,
      "learning_rate": 1.5e-06,
      "loss": 0.0,
      "reward": 3.53125,
      "reward_std": 0.4703739583492279,
      "rewards/gpt4o_acoustic_reward": 1.90625,
      "rewards/gpt4o_semantic_reward": 1.625,
      "step": 4
    },
    {
      "completion_length": 217.25,
      "epoch": 0.0005398984990821725,
      "grad_norm": 5.1103596687316895,
      "kl": 0.0002307891845703125,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 0.0,
      "reward": 3.71875,
      "reward_std": 0.4893767237663269,
      "rewards/gpt4o_acoustic_reward": 1.78125,
      "rewards/gpt4o_semantic_reward": 1.9375,
      "step": 5
    },
    {
      "completion_length": 257.75,
      "epoch": 0.0006478781988986071,
      "grad_norm": 5.311628341674805,
      "kl": 0.00016498565673828125,
      "learning_rate": 2.5e-06,
      "loss": 0.0,
      "reward": 3.21875,
      "reward_std": 0.5316166281700134,
      "rewards/gpt4o_acoustic_reward": 1.6875,
      "rewards/gpt4o_semantic_reward": 1.53125,
      "step": 6
    },
    {
      "completion_length": 261.0,
      "epoch": 0.0007558578987150416,
      "grad_norm": 4.986703395843506,
      "kl": 0.0002002716064453125,
      "learning_rate": 3e-06,
      "loss": 0.0,
      "reward": 3.59375,
      "reward_std": 0.650750994682312,
      "rewards/gpt4o_acoustic_reward": 1.6875,
      "rewards/gpt4o_semantic_reward": 1.90625,
      "step": 7
    },
    {
      "completion_length": 86.75,
      "epoch": 0.0008638375985314761,
      "grad_norm": 4.431477069854736,
      "kl": 0.0003185272216796875,
      "learning_rate": 3.5e-06,
      "loss": 0.0,
      "reward": 3.75,
      "reward_std": 0.5062625408172607,
      "rewards/gpt4o_acoustic_reward": 1.875,
      "rewards/gpt4o_semantic_reward": 1.875,
      "step": 8
    },
    {
      "completion_length": 131.25,
      "epoch": 0.0009718172983479105,
      "grad_norm": 4.666789531707764,
      "kl": 0.000743865966796875,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.0,
      "reward": 4.09375,
      "reward_std": 0.41466206312179565,
      "rewards/gpt4o_acoustic_reward": 1.9375,
      "rewards/gpt4o_semantic_reward": 2.15625,
      "step": 9
    },
    {
      "completion_length": 163.0,
      "epoch": 0.001079796998164345,
      "grad_norm": 4.67036771774292,
      "kl": 0.00101470947265625,
      "learning_rate": 4.5e-06,
      "loss": 0.0,
      "reward": 3.46875,
      "reward_std": 0.30195677280426025,
      "rewards/gpt4o_acoustic_reward": 1.90625,
      "rewards/gpt4o_semantic_reward": 1.5625,
      "step": 10
    },
    {
      "completion_length": 105.0,
      "epoch": 0.0011877766979807797,
      "grad_norm": 5.638863563537598,
      "kl": 0.0023193359375,
      "learning_rate": 5e-06,
      "loss": 0.0,
      "reward": 4.09375,
      "reward_std": 0.34684044122695923,
      "rewards/gpt4o_acoustic_reward": 1.9375,
      "rewards/gpt4o_semantic_reward": 2.15625,
      "step": 11
    },
    {
      "completion_length": 148.75,
      "epoch": 0.0012957563977972141,
      "grad_norm": 3.818448066711426,
      "kl": 0.0029296875,
      "learning_rate": 5.500000000000001e-06,
      "loss": 0.0,
      "reward": 3.75,
      "reward_std": 0.455963134765625,
      "rewards/gpt4o_acoustic_reward": 1.8125,
      "rewards/gpt4o_semantic_reward": 1.9375,
      "step": 12
    },
    {
      "completion_length": 146.0,
      "epoch": 0.0014037360976136485,
      "grad_norm": 3.814582586288452,
      "kl": 0.0072021484375,
      "learning_rate": 6e-06,
      "loss": 0.0001,
      "reward": 3.78125,
      "reward_std": 0.2961430847644806,
      "rewards/gpt4o_acoustic_reward": 1.8125,
      "rewards/gpt4o_semantic_reward": 1.96875,
      "step": 13
    },
    {
      "completion_length": 143.75,
      "epoch": 0.0015117157974300832,
      "grad_norm": 5.144272327423096,
      "kl": 0.01171875,
      "learning_rate": 6.5000000000000004e-06,
      "loss": 0.0001,
      "reward": 4.0625,
      "reward_std": 0.4435809254646301,
      "rewards/gpt4o_acoustic_reward": 2.03125,
      "rewards/gpt4o_semantic_reward": 2.03125,
      "step": 14
    },
    {
      "completion_length": 111.5,
      "epoch": 0.0016196954972465176,
      "grad_norm": 5.313261985778809,
      "kl": 0.01495361328125,
      "learning_rate": 7e-06,
      "loss": 0.0001,
      "reward": 3.5,
      "reward_std": 0.5165305137634277,
      "rewards/gpt4o_acoustic_reward": 1.84375,
      "rewards/gpt4o_semantic_reward": 1.65625,
      "step": 15
    },
    {
      "completion_length": 161.25,
      "epoch": 0.0017276751970629522,
      "grad_norm": 5.388939380645752,
      "kl": 0.020751953125,
      "learning_rate": 7.500000000000001e-06,
      "loss": 0.0002,
      "reward": 3.71875,
      "reward_std": 0.4040188491344452,
      "rewards/gpt4o_acoustic_reward": 1.875,
      "rewards/gpt4o_semantic_reward": 1.84375,
      "step": 16
    },
    {
      "completion_length": 534.0,
      "epoch": 0.0018356548968793867,
      "grad_norm": 6.198124408721924,
      "kl": 0.031494140625,
      "learning_rate": 8.000000000000001e-06,
      "loss": 0.0003,
      "reward": 3.09375,
      "reward_std": 0.7942933440208435,
      "rewards/gpt4o_acoustic_reward": 1.65625,
      "rewards/gpt4o_semantic_reward": 1.4375,
      "step": 17
    },
    {
      "completion_length": 184.25,
      "epoch": 0.001943634596695821,
      "grad_norm": 5.088284492492676,
      "kl": 0.033203125,
      "learning_rate": 8.5e-06,
      "loss": 0.0003,
      "reward": 3.40625,
      "reward_std": 0.4718404710292816,
      "rewards/gpt4o_acoustic_reward": 1.65625,
      "rewards/gpt4o_semantic_reward": 1.75,
      "step": 18
    },
    {
      "completion_length": 113.0,
      "epoch": 0.0020516142965122555,
      "grad_norm": 5.60780668258667,
      "kl": 0.048828125,
      "learning_rate": 9e-06,
      "loss": 0.0005,
      "reward": 3.875,
      "reward_std": 0.5530447959899902,
      "rewards/gpt4o_acoustic_reward": 1.96875,
      "rewards/gpt4o_semantic_reward": 1.90625,
      "step": 19
    },
    {
      "completion_length": 110.0,
      "epoch": 0.00215959399632869,
      "grad_norm": 4.834724426269531,
      "kl": 0.060791015625,
      "learning_rate": 9.5e-06,
      "loss": 0.0006,
      "reward": 3.53125,
      "reward_std": 0.4558830261230469,
      "rewards/gpt4o_acoustic_reward": 1.875,
      "rewards/gpt4o_semantic_reward": 1.65625,
      "step": 20
    },
    {
      "completion_length": 77.0,
      "epoch": 0.0022675736961451248,
      "grad_norm": 4.40089750289917,
      "kl": 0.103515625,
      "learning_rate": 1e-05,
      "loss": 0.001,
      "reward": 3.1875,
      "reward_std": 0.41582149267196655,
      "rewards/gpt4o_acoustic_reward": 1.53125,
      "rewards/gpt4o_semantic_reward": 1.65625,
      "step": 21
    },
    {
      "completion_length": 172.25,
      "epoch": 0.0023755533959615594,
      "grad_norm": 3.9682302474975586,
      "kl": 0.138671875,
      "learning_rate": 9.999999393824489e-06,
      "loss": 0.0014,
      "reward": 3.90625,
      "reward_std": 0.35455697774887085,
      "rewards/gpt4o_acoustic_reward": 1.875,
      "rewards/gpt4o_semantic_reward": 2.03125,
      "step": 22
    },
    {
      "completion_length": 87.75,
      "epoch": 0.0024835330957779936,
      "grad_norm": 5.275251865386963,
      "kl": 0.1259765625,
      "learning_rate": 9.9999975752981e-06,
      "loss": 0.0013,
      "reward": 3.40625,
      "reward_std": 0.6111315488815308,
      "rewards/gpt4o_acoustic_reward": 1.71875,
      "rewards/gpt4o_semantic_reward": 1.6875,
      "step": 23
    },
    {
      "completion_length": 142.0,
      "epoch": 0.0025915127955944283,
      "grad_norm": 3.5632264614105225,
      "kl": 0.140625,
      "learning_rate": 9.999994544421277e-06,
      "loss": 0.0014,
      "reward": 3.71875,
      "reward_std": 0.35455700755119324,
      "rewards/gpt4o_acoustic_reward": 1.84375,
      "rewards/gpt4o_semantic_reward": 1.875,
      "step": 24
    },
    {
      "completion_length": 116.0,
      "epoch": 0.002699492495410863,
      "grad_norm": 3.38319993019104,
      "kl": 0.23046875,
      "learning_rate": 9.99999030119475e-06,
      "loss": 0.0023,
      "reward": 3.78125,
      "reward_std": 0.40410637855529785,
      "rewards/gpt4o_acoustic_reward": 1.96875,
      "rewards/gpt4o_semantic_reward": 1.8125,
      "step": 25
    },
    {
      "completion_length": 129.0,
      "epoch": 0.002807472195227297,
      "grad_norm": 3.809497833251953,
      "kl": 0.255859375,
      "learning_rate": 9.999984845619553e-06,
      "loss": 0.0026,
      "reward": 3.6875,
      "reward_std": 0.25444716215133667,
      "rewards/gpt4o_acoustic_reward": 1.96875,
      "rewards/gpt4o_semantic_reward": 1.71875,
      "step": 26
    },
    {
      "completion_length": 137.5,
      "epoch": 0.0029154518950437317,
      "grad_norm": 4.824902057647705,
      "kl": 0.234375,
      "learning_rate": 9.999978177697008e-06,
      "loss": 0.0023,
      "reward": 3.65625,
      "reward_std": 0.37412554025650024,
      "rewards/gpt4o_acoustic_reward": 1.9375,
      "rewards/gpt4o_semantic_reward": 1.71875,
      "step": 27
    },
    {
      "completion_length": 106.0,
      "epoch": 0.0030234315948601664,
      "grad_norm": 3.999227523803711,
      "kl": 0.2470703125,
      "learning_rate": 9.99997029742873e-06,
      "loss": 0.0025,
      "reward": 4.34375,
      "reward_std": 0.4553835391998291,
      "rewards/gpt4o_acoustic_reward": 2.0,
      "rewards/gpt4o_semantic_reward": 2.34375,
      "step": 28
    },
    {
      "completion_length": 181.0,
      "epoch": 0.003131411294676601,
      "grad_norm": 3.2852933406829834,
      "kl": 0.22265625,
      "learning_rate": 9.999961204816628e-06,
      "loss": 0.0022,
      "reward": 3.28125,
      "reward_std": 0.44162869453430176,
      "rewards/gpt4o_acoustic_reward": 1.59375,
      "rewards/gpt4o_semantic_reward": 1.6875,
      "step": 29
    },
    {
      "completion_length": 468.25,
      "epoch": 0.003239390994493035,
      "grad_norm": 3.166698694229126,
      "kl": 0.2119140625,
      "learning_rate": 9.999950899862911e-06,
      "loss": 0.0021,
      "reward": 3.8125,
      "reward_std": 0.38640254735946655,
      "rewards/gpt4o_acoustic_reward": 1.75,
      "rewards/gpt4o_semantic_reward": 2.0625,
      "step": 30
    },
    {
      "completion_length": 200.5,
      "epoch": 0.00334737069430947,
      "grad_norm": 3.77912974357605,
      "kl": 0.283203125,
      "learning_rate": 9.999939382570075e-06,
      "loss": 0.0028,
      "reward": 3.84375,
      "reward_std": 0.4040188789367676,
      "rewards/gpt4o_acoustic_reward": 1.9375,
      "rewards/gpt4o_semantic_reward": 1.90625,
      "step": 31
    },
    {
      "completion_length": 200.75,
      "epoch": 0.0034553503941259045,
      "grad_norm": 3.542156934738159,
      "kl": 0.384765625,
      "learning_rate": 9.999926652940914e-06,
      "loss": 0.0039,
      "reward": 3.8125,
      "reward_std": 0.5613945722579956,
      "rewards/gpt4o_acoustic_reward": 1.9375,
      "rewards/gpt4o_semantic_reward": 1.875,
      "step": 32
    },
    {
      "completion_length": 141.5,
      "epoch": 0.0035633300939423387,
      "grad_norm": 4.484449863433838,
      "kl": 0.4453125,
      "learning_rate": 9.999912710978512e-06,
      "loss": 0.0044,
      "reward": 4.0,
      "reward_std": 0.535681962966919,
      "rewards/gpt4o_acoustic_reward": 1.8125,
      "rewards/gpt4o_semantic_reward": 2.1875,
      "step": 33
    },
    {
      "completion_length": 270.75,
      "epoch": 0.0036713097937587733,
      "grad_norm": 4.242273330688477,
      "kl": 0.36328125,
      "learning_rate": 9.999897556686254e-06,
      "loss": 0.0036,
      "reward": 3.9375,
      "reward_std": 0.5912878513336182,
      "rewards/gpt4o_acoustic_reward": 1.71875,
      "rewards/gpt4o_semantic_reward": 2.21875,
      "step": 34
    },
    {
      "completion_length": 143.75,
      "epoch": 0.003779289493575208,
      "grad_norm": 4.18234920501709,
      "kl": 0.41796875,
      "learning_rate": 9.999881190067809e-06,
      "loss": 0.0042,
      "reward": 3.5,
      "reward_std": 0.7955806255340576,
      "rewards/gpt4o_acoustic_reward": 1.53125,
      "rewards/gpt4o_semantic_reward": 1.96875,
      "step": 35
    },
    {
      "completion_length": 391.0,
      "epoch": 0.003887269193391642,
      "grad_norm": 2.8110883235931396,
      "kl": 0.3046875,
      "learning_rate": 9.999863611127149e-06,
      "loss": 0.003,
      "reward": 2.875,
      "reward_std": 0.37944716215133667,
      "rewards/gpt4o_acoustic_reward": 1.4375,
      "rewards/gpt4o_semantic_reward": 1.4375,
      "step": 36
    },
    {
      "completion_length": 176.25,
      "epoch": 0.003995248893208077,
      "grad_norm": 3.60547137260437,
      "kl": 0.298828125,
      "learning_rate": 9.999844819868535e-06,
      "loss": 0.003,
      "reward": 3.5,
      "reward_std": 0.6833221912384033,
      "rewards/gpt4o_acoustic_reward": 1.625,
      "rewards/gpt4o_semantic_reward": 1.875,
      "step": 37
    },
    {
      "completion_length": 146.75,
      "epoch": 0.004103228593024511,
      "grad_norm": 4.145380020141602,
      "kl": 0.416015625,
      "learning_rate": 9.999824816296527e-06,
      "loss": 0.0042,
      "reward": 3.84375,
      "reward_std": 0.5861972570419312,
      "rewards/gpt4o_acoustic_reward": 1.90625,
      "rewards/gpt4o_semantic_reward": 1.9375,
      "step": 38
    },
    {
      "completion_length": 145.75,
      "epoch": 0.004211208292840946,
      "grad_norm": 4.236860275268555,
      "kl": 0.462890625,
      "learning_rate": 9.99980360041597e-06,
      "loss": 0.0046,
      "reward": 3.9375,
      "reward_std": 0.4884645938873291,
      "rewards/gpt4o_acoustic_reward": 2.03125,
      "rewards/gpt4o_semantic_reward": 1.90625,
      "step": 39
    },
    {
      "completion_length": 153.5,
      "epoch": 0.00431918799265738,
      "grad_norm": 4.617465496063232,
      "kl": 0.5703125,
      "learning_rate": 9.99978117223201e-06,
      "loss": 0.0057,
      "reward": 3.875,
      "reward_std": 0.6747629046440125,
      "rewards/gpt4o_acoustic_reward": 2.03125,
      "rewards/gpt4o_semantic_reward": 1.84375,
      "step": 40
    },
    {
      "completion_length": 143.0,
      "epoch": 0.004427167692473815,
      "grad_norm": 4.216564178466797,
      "kl": 0.490234375,
      "learning_rate": 9.999757531750086e-06,
      "loss": 0.0049,
      "reward": 3.21875,
      "reward_std": 0.3697783946990967,
      "rewards/gpt4o_acoustic_reward": 1.625,
      "rewards/gpt4o_semantic_reward": 1.59375,
      "step": 41
    },
    {
      "completion_length": 58.0,
      "epoch": 0.0045351473922902496,
      "grad_norm": 5.301109313964844,
      "kl": 0.53125,
      "learning_rate": 9.999732678975928e-06,
      "loss": 0.0053,
      "reward": 3.25,
      "reward_std": 0.6608318090438843,
      "rewards/gpt4o_acoustic_reward": 1.6875,
      "rewards/gpt4o_semantic_reward": 1.5625,
      "step": 42
    },
    {
      "completion_length": 119.25,
      "epoch": 0.004643127092106684,
      "grad_norm": 3.858963966369629,
      "kl": 0.455078125,
      "learning_rate": 9.999706613915567e-06,
      "loss": 0.0046,
      "reward": 3.90625,
      "reward_std": 0.5087142586708069,
      "rewards/gpt4o_acoustic_reward": 1.84375,
      "rewards/gpt4o_semantic_reward": 2.0625,
      "step": 43
    },
    {
      "completion_length": 165.0,
      "epoch": 0.004751106791923119,
      "grad_norm": 4.549792289733887,
      "kl": 0.5546875,
      "learning_rate": 9.999679336575317e-06,
      "loss": 0.0056,
      "reward": 4.0,
      "reward_std": 0.3292241394519806,
      "rewards/gpt4o_acoustic_reward": 1.96875,
      "rewards/gpt4o_semantic_reward": 2.03125,
      "step": 44
    },
    {
      "completion_length": 55.75,
      "epoch": 0.004859086491739553,
      "grad_norm": 5.7472734451293945,
      "kl": 0.73046875,
      "learning_rate": 9.999650846961797e-06,
      "loss": 0.0073,
      "reward": 3.9375,
      "reward_std": 0.4665188491344452,
      "rewards/gpt4o_acoustic_reward": 1.84375,
      "rewards/gpt4o_semantic_reward": 2.09375,
      "step": 45
    },
    {
      "completion_length": 48.25,
      "epoch": 0.004967066191555987,
      "grad_norm": 3.262253522872925,
      "kl": 0.8984375,
      "learning_rate": 9.99962114508191e-06,
      "loss": 0.009,
      "reward": 3.15625,
      "reward_std": 0.417702317237854,
      "rewards/gpt4o_acoustic_reward": 1.34375,
      "rewards/gpt4o_semantic_reward": 1.8125,
      "step": 46
    },
    {
      "completion_length": 401.5,
      "epoch": 0.005075045891372422,
      "grad_norm": 10.104180335998535,
      "kl": 0.7421875,
      "learning_rate": 9.999590230942861e-06,
      "loss": 0.0074,
      "reward": 2.75,
      "reward_std": 0.6032058596611023,
      "rewards/gpt4o_acoustic_reward": 1.4375,
      "rewards/gpt4o_semantic_reward": 1.3125,
      "step": 47
    },
    {
      "completion_length": 183.5,
      "epoch": 0.0051830255911888565,
      "grad_norm": 5.417222023010254,
      "kl": 0.5859375,
      "learning_rate": 9.999558104552146e-06,
      "loss": 0.0058,
      "reward": 3.96875,
      "reward_std": 0.5666286945343018,
      "rewards/gpt4o_acoustic_reward": 1.8125,
      "rewards/gpt4o_semantic_reward": 2.15625,
      "step": 48
    },
    {
      "completion_length": 118.0,
      "epoch": 0.005291005291005291,
      "grad_norm": 4.999019622802734,
      "kl": 0.640625,
      "learning_rate": 9.999524765917554e-06,
      "loss": 0.0064,
      "reward": 3.25,
      "reward_std": 0.5622053742408752,
      "rewards/gpt4o_acoustic_reward": 1.6875,
      "rewards/gpt4o_semantic_reward": 1.5625,
      "step": 49
    },
    {
      "completion_length": 137.0,
      "epoch": 0.005398984990821726,
      "grad_norm": 4.787111759185791,
      "kl": 0.875,
      "learning_rate": 9.999490215047167e-06,
      "loss": 0.0088,
      "reward": 3.8125,
      "reward_std": 0.6087885499000549,
      "rewards/gpt4o_acoustic_reward": 1.8125,
      "rewards/gpt4o_semantic_reward": 2.0,
      "step": 50
    },
    {
      "completion_length": 136.25,
      "epoch": 0.00550696469063816,
      "grad_norm": 5.265024185180664,
      "kl": 0.61328125,
      "learning_rate": 9.999454451949364e-06,
      "loss": 0.0061,
      "reward": 3.4375,
      "reward_std": 0.6203576326370239,
      "rewards/gpt4o_acoustic_reward": 1.6875,
      "rewards/gpt4o_semantic_reward": 1.75,
      "step": 51
    },
    {
      "completion_length": 67.0,
      "epoch": 0.005614944390454594,
      "grad_norm": 5.786508083343506,
      "kl": 0.7578125,
      "learning_rate": 9.999417476632818e-06,
      "loss": 0.0076,
      "reward": 3.5,
      "reward_std": 0.40393877029418945,
      "rewards/gpt4o_acoustic_reward": 1.78125,
      "rewards/gpt4o_semantic_reward": 1.71875,
      "step": 52
    },
    {
      "completion_length": 138.0,
      "epoch": 0.005722924090271029,
      "grad_norm": 5.882197380065918,
      "kl": 0.984375,
      "learning_rate": 9.999379289106492e-06,
      "loss": 0.0098,
      "reward": 2.84375,
      "reward_std": 0.46389293670654297,
      "rewards/gpt4o_acoustic_reward": 1.5625,
      "rewards/gpt4o_semantic_reward": 1.28125,
      "step": 53
    },
    {
      "completion_length": 90.75,
      "epoch": 0.0058309037900874635,
      "grad_norm": 5.502197742462158,
      "kl": 0.8046875,
      "learning_rate": 9.999339889379647e-06,
      "loss": 0.008,
      "reward": 3.9375,
      "reward_std": 0.5373806953430176,
      "rewards/gpt4o_acoustic_reward": 1.84375,
      "rewards/gpt4o_semantic_reward": 2.09375,
      "step": 54
    },
    {
      "completion_length": 134.75,
      "epoch": 0.005938883489903898,
      "grad_norm": 5.883394241333008,
      "kl": 0.765625,
      "learning_rate": 9.999299277461835e-06,
      "loss": 0.0077,
      "reward": 3.75,
      "reward_std": 0.5033453106880188,
      "rewards/gpt4o_acoustic_reward": 1.84375,
      "rewards/gpt4o_semantic_reward": 1.90625,
      "step": 55
    },
    {
      "completion_length": 92.5,
      "epoch": 0.006046863189720333,
      "grad_norm": 5.420223236083984,
      "kl": 0.7265625,
      "learning_rate": 9.999257453362903e-06,
      "loss": 0.0073,
      "reward": 3.625,
      "reward_std": 0.5520563125610352,
      "rewards/gpt4o_acoustic_reward": 1.75,
      "rewards/gpt4o_semantic_reward": 1.875,
      "step": 56
    },
    {
      "completion_length": 384.75,
      "epoch": 0.006154842889536767,
      "grad_norm": 5.169672012329102,
      "kl": 0.765625,
      "learning_rate": 9.999214417092993e-06,
      "loss": 0.0077,
      "reward": 2.8125,
      "reward_std": 0.5581127405166626,
      "rewards/gpt4o_acoustic_reward": 1.40625,
      "rewards/gpt4o_semantic_reward": 1.40625,
      "step": 57
    },
    {
      "completion_length": 110.0,
      "epoch": 0.006262822589353202,
      "grad_norm": 5.439882755279541,
      "kl": 0.73828125,
      "learning_rate": 9.99917016866254e-06,
      "loss": 0.0074,
      "reward": 3.3125,
      "reward_std": 0.38853639364242554,
      "rewards/gpt4o_acoustic_reward": 1.65625,
      "rewards/gpt4o_semantic_reward": 1.65625,
      "step": 58
    },
    {
      "completion_length": 138.75,
      "epoch": 0.006370802289169636,
      "grad_norm": 4.6155219078063965,
      "kl": 0.671875,
      "learning_rate": 9.999124708082272e-06,
      "loss": 0.0067,
      "reward": 3.5625,
      "reward_std": 0.5936527252197266,
      "rewards/gpt4o_acoustic_reward": 1.8125,
      "rewards/gpt4o_semantic_reward": 1.75,
      "step": 59
    },
    {
      "completion_length": 108.75,
      "epoch": 0.00647878198898607,
      "grad_norm": 5.069967746734619,
      "kl": 0.90625,
      "learning_rate": 9.999078035363214e-06,
      "loss": 0.0091,
      "reward": 3.5625,
      "reward_std": 0.5204293727874756,
      "rewards/gpt4o_acoustic_reward": 1.6875,
      "rewards/gpt4o_semantic_reward": 1.875,
      "step": 60
    },
    {
      "completion_length": 234.75,
      "epoch": 0.0065867616888025055,
      "grad_norm": 4.814848899841309,
      "kl": 0.9453125,
      "learning_rate": 9.999030150516681e-06,
      "loss": 0.0095,
      "reward": 3.5,
      "reward_std": 0.554833710193634,
      "rewards/gpt4o_acoustic_reward": 1.78125,
      "rewards/gpt4o_semantic_reward": 1.71875,
      "step": 61
    },
    {
      "completion_length": 115.25,
      "epoch": 0.00669474138861894,
      "grad_norm": 4.598373889923096,
      "kl": 0.6484375,
      "learning_rate": 9.998981053554285e-06,
      "loss": 0.0065,
      "reward": 3.4375,
      "reward_std": 0.37944716215133667,
      "rewards/gpt4o_acoustic_reward": 1.71875,
      "rewards/gpt4o_semantic_reward": 1.71875,
      "step": 62
    },
    {
      "completion_length": 107.75,
      "epoch": 0.006802721088435374,
      "grad_norm": 5.762064456939697,
      "kl": 1.015625,
      "learning_rate": 9.998930744487927e-06,
      "loss": 0.0102,
      "reward": 3.75,
      "reward_std": 0.5572783946990967,
      "rewards/gpt4o_acoustic_reward": 1.6875,
      "rewards/gpt4o_semantic_reward": 2.0625,
      "step": 63
    },
    {
      "completion_length": 88.75,
      "epoch": 0.006910700788251809,
      "grad_norm": 5.052530765533447,
      "kl": 0.90625,
      "learning_rate": 9.998879223329809e-06,
      "loss": 0.0091,
      "reward": 3.40625,
      "reward_std": 0.587338924407959,
      "rewards/gpt4o_acoustic_reward": 1.71875,
      "rewards/gpt4o_semantic_reward": 1.6875,
      "step": 64
    },
    {
      "completion_length": 210.75,
      "epoch": 0.007018680488068243,
      "grad_norm": 4.801774024963379,
      "kl": 0.578125,
      "learning_rate": 9.99882649009242e-06,
      "loss": 0.0058,
      "reward": 3.4375,
      "reward_std": 0.454224169254303,
      "rewards/gpt4o_acoustic_reward": 1.71875,
      "rewards/gpt4o_semantic_reward": 1.71875,
      "step": 65
    },
    {
      "completion_length": 145.5,
      "epoch": 0.007126660187884677,
      "grad_norm": 5.015283107757568,
      "kl": 0.84765625,
      "learning_rate": 9.998772544788552e-06,
      "loss": 0.0085,
      "reward": 2.875,
      "reward_std": 0.744390606880188,
      "rewards/gpt4o_acoustic_reward": 1.53125,
      "rewards/gpt4o_semantic_reward": 1.34375,
      "step": 66
    },
    {
      "completion_length": 72.5,
      "epoch": 0.0072346398877011124,
      "grad_norm": 5.724644184112549,
      "kl": 0.9765625,
      "learning_rate": 9.99871738743128e-06,
      "loss": 0.0098,
      "reward": 3.09375,
      "reward_std": 0.7175944447517395,
      "rewards/gpt4o_acoustic_reward": 1.71875,
      "rewards/gpt4o_semantic_reward": 1.375,
      "step": 67
    },
    {
      "completion_length": 61.25,
      "epoch": 0.007342619587517547,
      "grad_norm": 4.83007287979126,
      "kl": 0.765625,
      "learning_rate": 9.99866101803398e-06,
      "loss": 0.0077,
      "reward": 3.34375,
      "reward_std": 0.3844503164291382,
      "rewards/gpt4o_acoustic_reward": 1.71875,
      "rewards/gpt4o_semantic_reward": 1.625,
      "step": 68
    },
    {
      "completion_length": 102.5,
      "epoch": 0.007450599287333981,
      "grad_norm": 7.941281795501709,
      "kl": 0.65625,
      "learning_rate": 9.99860343661032e-06,
      "loss": 0.0066,
      "reward": 3.4375,
      "reward_std": 0.6357907056808472,
      "rewards/gpt4o_acoustic_reward": 1.75,
      "rewards/gpt4o_semantic_reward": 1.6875,
      "step": 69
    },
    {
      "completion_length": 149.5,
      "epoch": 0.007558578987150416,
      "grad_norm": 4.062788009643555,
      "kl": 0.5625,
      "learning_rate": 9.998544643174261e-06,
      "loss": 0.0056,
      "reward": 3.59375,
      "reward_std": 0.5205094814300537,
      "rewards/gpt4o_acoustic_reward": 1.78125,
      "rewards/gpt4o_semantic_reward": 1.8125,
      "step": 70
    },
    {
      "completion_length": 298.75,
      "epoch": 0.00766655868696685,
      "grad_norm": 4.381830215454102,
      "kl": 0.52734375,
      "learning_rate": 9.998484637740058e-06,
      "loss": 0.0053,
      "reward": 3.15625,
      "reward_std": 0.6124106645584106,
      "rewards/gpt4o_acoustic_reward": 1.71875,
      "rewards/gpt4o_semantic_reward": 1.4375,
      "step": 71
    },
    {
      "completion_length": 102.0,
      "epoch": 0.007774538386783284,
      "grad_norm": 4.673843860626221,
      "kl": 1.1171875,
      "learning_rate": 9.998423420322264e-06,
      "loss": 0.0112,
      "reward": 2.25,
      "reward_std": 0.38853639364242554,
      "rewards/gpt4o_acoustic_reward": 1.1875,
      "rewards/gpt4o_semantic_reward": 1.0625,
      "step": 72
    },
    {
      "completion_length": 69.75,
      "epoch": 0.007882518086599719,
      "grad_norm": 5.254530906677246,
      "kl": 0.98046875,
      "learning_rate": 9.998360990935716e-06,
      "loss": 0.0098,
      "reward": 3.03125,
      "reward_std": 0.6602689027786255,
      "rewards/gpt4o_acoustic_reward": 1.59375,
      "rewards/gpt4o_semantic_reward": 1.4375,
      "step": 73
    },
    {
      "completion_length": 135.75,
      "epoch": 0.007990497786416154,
      "grad_norm": 4.070398807525635,
      "kl": 1.15625,
      "learning_rate": 9.998297349595558e-06,
      "loss": 0.0115,
      "reward": 2.375,
      "reward_std": 0.3908371925354004,
      "rewards/gpt4o_acoustic_reward": 1.15625,
      "rewards/gpt4o_semantic_reward": 1.21875,
      "step": 74
    },
    {
      "completion_length": 181.0,
      "epoch": 0.008098477486232589,
      "grad_norm": 3.7110443115234375,
      "kl": 0.58984375,
      "learning_rate": 9.998232496317217e-06,
      "loss": 0.0059,
      "reward": 3.21875,
      "reward_std": 0.5467228889465332,
      "rewards/gpt4o_acoustic_reward": 1.53125,
      "rewards/gpt4o_semantic_reward": 1.6875,
      "step": 75
    },
    {
      "completion_length": 813.25,
      "epoch": 0.008206457186049022,
      "grad_norm": 2.858513116836548,
      "kl": 0.546875,
      "learning_rate": 9.998166431116421e-06,
      "loss": 0.0055,
      "reward": 2.625,
      "reward_std": 0.864783763885498,
      "rewards/gpt4o_acoustic_reward": 1.3125,
      "rewards/gpt4o_semantic_reward": 1.3125,
      "step": 76
    },
    {
      "completion_length": 1024.0,
      "epoch": 0.008314436885865457,
      "grad_norm": 2.668346643447876,
      "kl": 0.56640625,
      "learning_rate": 9.998099154009186e-06,
      "loss": 0.0057,
      "reward": 2.3125,
      "reward_std": 0.4607051610946655,
      "rewards/gpt4o_acoustic_reward": 1.3125,
      "rewards/gpt4o_semantic_reward": 1.0,
      "step": 77
    },
    {
      "completion_length": 830.0,
      "epoch": 0.008422416585681892,
      "grad_norm": 2.78299880027771,
      "kl": 0.578125,
      "learning_rate": 9.998030665011827e-06,
      "loss": 0.0058,
      "reward": 2.25,
      "reward_std": 0.38640254735946655,
      "rewards/gpt4o_acoustic_reward": 1.15625,
      "rewards/gpt4o_semantic_reward": 1.09375,
      "step": 78
    },
    {
      "completion_length": 465.5,
      "epoch": 0.008530396285498325,
      "grad_norm": 2.7636075019836426,
      "kl": 0.578125,
      "learning_rate": 9.997960964140946e-06,
      "loss": 0.0058,
      "reward": 2.78125,
      "reward_std": 0.6020145416259766,
      "rewards/gpt4o_acoustic_reward": 1.5,
      "rewards/gpt4o_semantic_reward": 1.28125,
      "step": 79
    },
    {
      "completion_length": 265.25,
      "epoch": 0.00863837598531476,
      "grad_norm": 5.034330368041992,
      "kl": 0.8203125,
      "learning_rate": 9.99789005141345e-06,
      "loss": 0.0082,
      "reward": 2.90625,
      "reward_std": 0.5861972570419312,
      "rewards/gpt4o_acoustic_reward": 1.375,
      "rewards/gpt4o_semantic_reward": 1.53125,
      "step": 80
    },
    {
      "completion_length": 150.75,
      "epoch": 0.008746355685131196,
      "grad_norm": 3.18487286567688,
      "kl": 0.8359375,
      "learning_rate": 9.997817926846528e-06,
      "loss": 0.0084,
      "reward": 2.6875,
      "reward_std": 0.8567711114883423,
      "rewards/gpt4o_acoustic_reward": 1.34375,
      "rewards/gpt4o_semantic_reward": 1.34375,
      "step": 81
    },
    {
      "completion_length": 134.75,
      "epoch": 0.00885433538494763,
      "grad_norm": 3.657865047454834,
      "kl": 0.90234375,
      "learning_rate": 9.997744590457672e-06,
      "loss": 0.009,
      "reward": 2.625,
      "reward_std": 0.7562625408172607,
      "rewards/gpt4o_acoustic_reward": 1.40625,
      "rewards/gpt4o_semantic_reward": 1.21875,
      "step": 82
    },
    {
      "completion_length": 88.0,
      "epoch": 0.008962315084764064,
      "grad_norm": 4.395254135131836,
      "kl": 0.90625,
      "learning_rate": 9.997670042264661e-06,
      "loss": 0.009,
      "reward": 3.28125,
      "reward_std": 0.2597687840461731,
      "rewards/gpt4o_acoustic_reward": 1.5625,
      "rewards/gpt4o_semantic_reward": 1.71875,
      "step": 83
    },
    {
      "completion_length": 185.25,
      "epoch": 0.009070294784580499,
      "grad_norm": 3.564257860183716,
      "kl": 0.7578125,
      "learning_rate": 9.997594282285572e-06,
      "loss": 0.0076,
      "reward": 3.21875,
      "reward_std": 0.5205094814300537,
      "rewards/gpt4o_acoustic_reward": 1.75,
      "rewards/gpt4o_semantic_reward": 1.46875,
      "step": 84
    },
    {
      "completion_length": 285.0,
      "epoch": 0.009178274484396934,
      "grad_norm": 3.515869617462158,
      "kl": 0.90234375,
      "learning_rate": 9.997517310538774e-06,
      "loss": 0.009,
      "reward": 2.78125,
      "reward_std": 0.5465551018714905,
      "rewards/gpt4o_acoustic_reward": 1.4375,
      "rewards/gpt4o_semantic_reward": 1.34375,
      "step": 85
    },
    {
      "completion_length": 345.25,
      "epoch": 0.009286254184213368,
      "grad_norm": 3.76198673248291,
      "kl": 0.8828125,
      "learning_rate": 9.99743912704293e-06,
      "loss": 0.0088,
      "reward": 2.375,
      "reward_std": 0.9464328289031982,
      "rewards/gpt4o_acoustic_reward": 1.34375,
      "rewards/gpt4o_semantic_reward": 1.03125,
      "step": 86
    },
    {
      "completion_length": 470.75,
      "epoch": 0.009394233884029803,
      "grad_norm": 3.8601486682891846,
      "kl": 0.70703125,
      "learning_rate": 9.997359731816998e-06,
      "loss": 0.0071,
      "reward": 3.0625,
      "reward_std": 0.6021620631217957,
      "rewards/gpt4o_acoustic_reward": 1.6875,
      "rewards/gpt4o_semantic_reward": 1.375,
      "step": 87
    },
    {
      "completion_length": 330.25,
      "epoch": 0.009502213583846238,
      "grad_norm": 4.483823299407959,
      "kl": 0.953125,
      "learning_rate": 9.99727912488023e-06,
      "loss": 0.0095,
      "reward": 2.5625,
      "reward_std": 0.45129746198654175,
      "rewards/gpt4o_acoustic_reward": 1.46875,
      "rewards/gpt4o_semantic_reward": 1.09375,
      "step": 88
    },
    {
      "completion_length": 325.5,
      "epoch": 0.009610193283662671,
      "grad_norm": 3.386530876159668,
      "kl": 0.7578125,
      "learning_rate": 9.997197306252168e-06,
      "loss": 0.0076,
      "reward": 2.65625,
      "reward_std": 0.6569532155990601,
      "rewards/gpt4o_acoustic_reward": 1.4375,
      "rewards/gpt4o_semantic_reward": 1.21875,
      "step": 89
    },
    {
      "completion_length": 100.75,
      "epoch": 0.009718172983479106,
      "grad_norm": 4.329235553741455,
      "kl": 1.2734375,
      "learning_rate": 9.997114275952653e-06,
      "loss": 0.0127,
      "reward": 2.34375,
      "reward_std": 0.5661739110946655,
      "rewards/gpt4o_acoustic_reward": 1.1875,
      "rewards/gpt4o_semantic_reward": 1.15625,
      "step": 90
    }
  ],
  "logging_steps": 1,
  "max_steps": 6400,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 5,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}