{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.014037360976136486,
  "eval_steps": 500,
  "global_step": 130,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "completion_length": 149.0,
      "epoch": 0.00010797969981643452,
      "grad_norm": 4.355660438537598,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": 0.0,
      "reward": 3.75,
      "reward_std": 0.4128488302230835,
      "rewards/gpt4o_acoustic_reward": 0.699999988079071,
      "rewards/gpt4o_semantic_reward": 3.049999952316284,
      "step": 1
    },
    {
      "completion_length": 87.25,
      "epoch": 0.00021595939963286903,
      "grad_norm": 10.339611053466797,
      "kl": 0.0,
      "learning_rate": 1e-07,
      "loss": 0.0,
      "reward": 3.950000047683716,
      "reward_std": 0.46606144309043884,
      "rewards/gpt4o_acoustic_reward": 0.75,
      "rewards/gpt4o_semantic_reward": 3.1999998092651367,
      "step": 2
    },
    {
      "completion_length": 129.25,
      "epoch": 0.00032393909944930353,
      "grad_norm": 9.756179809570312,
      "kl": 0.0001125335693359375,
      "learning_rate": 2e-07,
      "loss": 0.0,
      "reward": 3.6875,
      "reward_std": 0.4336637258529663,
      "rewards/gpt4o_acoustic_reward": 0.737500011920929,
      "rewards/gpt4o_semantic_reward": 2.950000047683716,
      "step": 3
    },
    {
      "completion_length": 75.75,
      "epoch": 0.00043191879926573806,
      "grad_norm": 5.357244968414307,
      "kl": 0.000377655029296875,
      "learning_rate": 3e-07,
      "loss": 0.0,
      "reward": 3.4000000953674316,
      "reward_std": 0.4117715358734131,
      "rewards/gpt4o_acoustic_reward": 0.699999988079071,
      "rewards/gpt4o_semantic_reward": 2.700000047683716,
      "step": 4
    },
    {
      "completion_length": 188.75,
      "epoch": 0.0005398984990821725,
      "grad_norm": 5.540310382843018,
      "kl": 8.058547973632812e-05,
      "learning_rate": 4e-07,
      "loss": 0.0,
      "reward": 3.6875,
      "reward_std": 0.48131972551345825,
      "rewards/gpt4o_acoustic_reward": 0.737500011920929,
      "rewards/gpt4o_semantic_reward": 2.950000047683716,
      "step": 5
    },
    {
      "completion_length": 393.75,
      "epoch": 0.0006478781988986071,
      "grad_norm": 9.635984420776367,
      "kl": 0.00014019012451171875,
      "learning_rate": 5e-07,
      "loss": 0.0,
      "reward": 3.4625000953674316,
      "reward_std": 0.8044465184211731,
      "rewards/gpt4o_acoustic_reward": 0.7124999761581421,
      "rewards/gpt4o_semantic_reward": 2.75,
      "step": 6
    },
    {
      "completion_length": 217.75,
      "epoch": 0.0007558578987150416,
      "grad_norm": 4.512867450714111,
      "kl": 4.100799560546875e-05,
      "learning_rate": 6e-07,
      "loss": 0.0,
      "reward": 3.875,
      "reward_std": 0.6544983386993408,
      "rewards/gpt4o_acoustic_reward": 0.625,
      "rewards/gpt4o_semantic_reward": 3.25,
      "step": 7
    },
    {
      "completion_length": 104.0,
      "epoch": 0.0008638375985314761,
      "grad_norm": 4.235530376434326,
      "kl": 0.0001773834228515625,
      "learning_rate": 7e-07,
      "loss": 0.0,
      "reward": 4.049999713897705,
      "reward_std": 0.5000368356704712,
      "rewards/gpt4o_acoustic_reward": 0.800000011920929,
      "rewards/gpt4o_semantic_reward": 3.25,
      "step": 8
    },
    {
      "completion_length": 139.75,
      "epoch": 0.0009718172983479105,
      "grad_norm": 5.992825984954834,
      "kl": 0.00017547607421875,
      "learning_rate": 8e-07,
      "loss": 0.0,
      "reward": 4.150000095367432,
      "reward_std": 0.4549916386604309,
      "rewards/gpt4o_acoustic_reward": 0.75,
      "rewards/gpt4o_semantic_reward": 3.3999998569488525,
      "step": 9
    },
    {
      "completion_length": 172.5,
      "epoch": 0.001079796998164345,
      "grad_norm": 5.302628040313721,
      "kl": 0.000141143798828125,
      "learning_rate": 9e-07,
      "loss": 0.0,
      "reward": 3.4625000953674316,
      "reward_std": 0.6239354610443115,
      "rewards/gpt4o_acoustic_reward": 0.8125,
      "rewards/gpt4o_semantic_reward": 2.6500000953674316,
      "step": 10
    },
    {
      "completion_length": 98.5,
      "epoch": 0.0011877766979807797,
      "grad_norm": 5.695831775665283,
      "kl": 5.698204040527344e-05,
      "learning_rate": 1e-06,
      "loss": 0.0,
      "reward": 4.237500190734863,
      "reward_std": 0.4028981328010559,
      "rewards/gpt4o_acoustic_reward": 0.7874999642372131,
      "rewards/gpt4o_semantic_reward": 3.4499998092651367,
      "step": 11
    },
    {
      "completion_length": 176.0,
      "epoch": 0.0012957563977972141,
      "grad_norm": 3.156630039215088,
      "kl": 0.00012302398681640625,
      "learning_rate": 9.999999395720266e-07,
      "loss": 0.0,
      "reward": 4.012499809265137,
      "reward_std": 0.29057008028030396,
      "rewards/gpt4o_acoustic_reward": 0.7125000357627869,
      "rewards/gpt4o_semantic_reward": 3.299999952316284,
      "step": 12
    },
    {
      "completion_length": 142.25,
      "epoch": 0.0014037360976136485,
      "grad_norm": 4.895700454711914,
      "kl": 0.00019359588623046875,
      "learning_rate": 9.999997582881211e-07,
      "loss": 0.0,
      "reward": 4.0,
      "reward_std": 0.354991614818573,
      "rewards/gpt4o_acoustic_reward": 0.75,
      "rewards/gpt4o_semantic_reward": 3.249999761581421,
      "step": 13
    },
    {
      "completion_length": 138.5,
      "epoch": 0.0015117157974300832,
      "grad_norm": 6.870318412780762,
      "kl": 0.00029754638671875,
      "learning_rate": 9.999994561483274e-07,
      "loss": 0.0,
      "reward": 4.099999904632568,
      "reward_std": 0.3661222457885742,
      "rewards/gpt4o_acoustic_reward": 0.7999999523162842,
      "rewards/gpt4o_semantic_reward": 3.3000001907348633,
      "step": 14
    },
    {
      "completion_length": 127.5,
      "epoch": 0.0016196954972465176,
      "grad_norm": 6.080170154571533,
      "kl": 0.000301361083984375,
      "learning_rate": 9.999990331527185e-07,
      "loss": 0.0,
      "reward": 3.125000238418579,
      "reward_std": 0.4335150122642517,
      "rewards/gpt4o_acoustic_reward": 0.625,
      "rewards/gpt4o_semantic_reward": 2.5,
      "step": 15
    },
    {
      "completion_length": 147.25,
      "epoch": 0.0017276751970629522,
      "grad_norm": 4.455915927886963,
      "kl": 0.0003414154052734375,
      "learning_rate": 9.999984893013964e-07,
      "loss": 0.0,
      "reward": 4.150000095367432,
      "reward_std": 0.3805276155471802,
      "rewards/gpt4o_acoustic_reward": 0.75,
      "rewards/gpt4o_semantic_reward": 3.3999998569488525,
      "step": 16
    },
    {
      "completion_length": 534.0,
      "epoch": 0.0018356548968793867,
      "grad_norm": 4.104217529296875,
      "kl": 0.00018024444580078125,
      "learning_rate": 9.999978245944931e-07,
      "loss": 0.0,
      "reward": 3.387500047683716,
      "reward_std": 0.4947923421859741,
      "rewards/gpt4o_acoustic_reward": 0.7374999523162842,
      "rewards/gpt4o_semantic_reward": 2.6500000953674316,
      "step": 17
    },
    {
      "completion_length": 206.0,
      "epoch": 0.001943634596695821,
      "grad_norm": 4.355428695678711,
      "kl": 0.0005950927734375,
      "learning_rate": 9.999970390321687e-07,
      "loss": 0.0001,
      "reward": 3.5625,
      "reward_std": 0.6289784908294678,
      "rewards/gpt4o_acoustic_reward": 0.6625000238418579,
      "rewards/gpt4o_semantic_reward": 2.9000000953674316,
      "step": 18
    },
    {
      "completion_length": 103.0,
      "epoch": 0.0020516142965122555,
      "grad_norm": 5.251166820526123,
      "kl": 0.00048828125,
      "learning_rate": 9.999961326146134e-07,
      "loss": 0.0,
      "reward": 3.549999952316284,
      "reward_std": 0.4274619519710541,
      "rewards/gpt4o_acoustic_reward": 0.75,
      "rewards/gpt4o_semantic_reward": 2.8000001907348633,
      "step": 19
    },
    {
      "completion_length": 117.5,
      "epoch": 0.00215959399632869,
      "grad_norm": 4.956709384918213,
      "kl": 0.0003566741943359375,
      "learning_rate": 9.99995105342046e-07,
      "loss": 0.0,
      "reward": 3.012500286102295,
      "reward_std": 0.7860726714134216,
      "rewards/gpt4o_acoustic_reward": 0.762499988079071,
      "rewards/gpt4o_semantic_reward": 2.25,
      "step": 20
    },
    {
      "completion_length": 66.5,
      "epoch": 0.0022675736961451248,
      "grad_norm": 3.629720687866211,
      "kl": 0.00066375732421875,
      "learning_rate": 9.999939572147152e-07,
      "loss": 0.0001,
      "reward": 3.5875000953674316,
      "reward_std": 0.5995372533798218,
      "rewards/gpt4o_acoustic_reward": 0.6875000596046448,
      "rewards/gpt4o_semantic_reward": 2.9000000953674316,
      "step": 21
    },
    {
      "completion_length": 171.75,
      "epoch": 0.0023755533959615594,
      "grad_norm": 4.503050804138184,
      "kl": 0.0007476806640625,
      "learning_rate": 9.999926882328982e-07,
      "loss": 0.0001,
      "reward": 3.7125000953674316,
      "reward_std": 0.5707715153694153,
      "rewards/gpt4o_acoustic_reward": 0.762499988079071,
      "rewards/gpt4o_semantic_reward": 2.950000047683716,
      "step": 22
    },
    {
      "completion_length": 82.0,
      "epoch": 0.0024835330957779936,
      "grad_norm": 4.001644611358643,
      "kl": 0.00089263916015625,
      "learning_rate": 9.999912983969018e-07,
      "loss": 0.0001,
      "reward": 3.2874999046325684,
      "reward_std": 0.29747867584228516,
      "rewards/gpt4o_acoustic_reward": 0.7874999642372131,
      "rewards/gpt4o_semantic_reward": 2.5,
      "step": 23
    },
    {
      "completion_length": 122.0,
      "epoch": 0.0025915127955944283,
      "grad_norm": 3.9283950328826904,
      "kl": 0.000701904296875,
      "learning_rate": 9.999897877070623e-07,
      "loss": 0.0001,
      "reward": 3.9000000953674316,
      "reward_std": 0.5509018898010254,
      "rewards/gpt4o_acoustic_reward": 0.75,
      "rewards/gpt4o_semantic_reward": 3.1500000953674316,
      "step": 24
    },
    {
      "completion_length": 62.75,
      "epoch": 0.002699492495410863,
      "grad_norm": 4.654013156890869,
      "kl": 0.000957489013671875,
      "learning_rate": 9.999881561637442e-07,
      "loss": 0.0001,
      "reward": 3.7249999046325684,
      "reward_std": 0.5639642477035522,
      "rewards/gpt4o_acoustic_reward": 0.824999988079071,
      "rewards/gpt4o_semantic_reward": 2.9000000953674316,
      "step": 25
    },
    {
      "completion_length": 83.75,
      "epoch": 0.002807472195227297,
      "grad_norm": 5.039306163787842,
      "kl": 0.001708984375,
      "learning_rate": 9.999864037673423e-07,
      "loss": 0.0002,
      "reward": 3.6624999046325684,
      "reward_std": 0.5867879390716553,
      "rewards/gpt4o_acoustic_reward": 0.762499988079071,
      "rewards/gpt4o_semantic_reward": 2.9000000953674316,
      "step": 26
    },
    {
      "completion_length": 109.75,
      "epoch": 0.0029154518950437317,
      "grad_norm": 5.404584884643555,
      "kl": 0.0012969970703125,
      "learning_rate": 9.9998453051828e-07,
      "loss": 0.0001,
      "reward": 3.5374999046325684,
      "reward_std": 0.5478276014328003,
      "rewards/gpt4o_acoustic_reward": 0.7874999642372131,
      "rewards/gpt4o_semantic_reward": 2.75,
      "step": 27
    },
    {
      "completion_length": 111.0,
      "epoch": 0.0030234315948601664,
      "grad_norm": 4.432938098907471,
      "kl": 0.00122833251953125,
      "learning_rate": 9.999825364170103e-07,
      "loss": 0.0001,
      "reward": 4.149999618530273,
      "reward_std": 0.3389889597892761,
      "rewards/gpt4o_acoustic_reward": 0.800000011920929,
      "rewards/gpt4o_semantic_reward": 3.3499999046325684,
      "step": 28
    },
    {
      "completion_length": 122.75,
      "epoch": 0.003131411294676601,
      "grad_norm": 5.11518669128418,
      "kl": 0.00152587890625,
      "learning_rate": 9.99980421464015e-07,
      "loss": 0.0002,
      "reward": 3.6374998092651367,
      "reward_std": 0.38321733474731445,
      "rewards/gpt4o_acoustic_reward": 0.6875000596046448,
      "rewards/gpt4o_semantic_reward": 2.950000047683716,
      "step": 29
    },
    {
      "completion_length": 161.0,
      "epoch": 0.003239390994493035,
      "grad_norm": 4.147963523864746,
      "kl": 0.0016326904296875,
      "learning_rate": 9.999781856598053e-07,
      "loss": 0.0002,
      "reward": 4.175000190734863,
      "reward_std": 0.2711343467235565,
      "rewards/gpt4o_acoustic_reward": 0.7749999761581421,
      "rewards/gpt4o_semantic_reward": 3.4000000953674316,
      "step": 30
    },
    {
      "completion_length": 125.25,
      "epoch": 0.00334737069430947,
      "grad_norm": 5.0930562019348145,
      "kl": 0.0027313232421875,
      "learning_rate": 9.999758290049217e-07,
      "loss": 0.0003,
      "reward": 3.4375,
      "reward_std": 0.6973484754562378,
      "rewards/gpt4o_acoustic_reward": 0.737500011920929,
      "rewards/gpt4o_semantic_reward": 2.6999998092651367,
      "step": 31
    },
    {
      "completion_length": 84.75,
      "epoch": 0.0034553503941259045,
      "grad_norm": 4.429357528686523,
      "kl": 0.0017547607421875,
      "learning_rate": 9.99973351499934e-07,
      "loss": 0.0002,
      "reward": 3.2875001430511475,
      "reward_std": 0.5658307075500488,
      "rewards/gpt4o_acoustic_reward": 0.7374999523162842,
      "rewards/gpt4o_semantic_reward": 2.5500001907348633,
      "step": 32
    },
    {
      "completion_length": 96.0,
      "epoch": 0.0035633300939423387,
      "grad_norm": 5.403738498687744,
      "kl": 0.0024566650390625,
      "learning_rate": 9.999707531454407e-07,
      "loss": 0.0002,
      "reward": 3.262500047683716,
      "reward_std": 0.5316473841667175,
      "rewards/gpt4o_acoustic_reward": 0.7125000357627869,
      "rewards/gpt4o_semantic_reward": 2.549999952316284,
      "step": 33
    },
    {
      "completion_length": 92.0,
      "epoch": 0.0036713097937587733,
      "grad_norm": 5.378174304962158,
      "kl": 0.00179290771484375,
      "learning_rate": 9.9996803394207e-07,
      "loss": 0.0002,
      "reward": 4.462500095367432,
      "reward_std": 0.41377225518226624,
      "rewards/gpt4o_acoustic_reward": 0.762499988079071,
      "rewards/gpt4o_semantic_reward": 3.6999998092651367,
      "step": 34
    },
    {
      "completion_length": 88.0,
      "epoch": 0.003779289493575208,
      "grad_norm": 13.495048522949219,
      "kl": 0.0030975341796875,
      "learning_rate": 9.999651938904792e-07,
      "loss": 0.0003,
      "reward": 4.324999809265137,
      "reward_std": 0.4066784977912903,
      "rewards/gpt4o_acoustic_reward": 0.8250000476837158,
      "rewards/gpt4o_semantic_reward": 3.5,
      "step": 35
    },
    {
      "completion_length": 117.5,
      "epoch": 0.003887269193391642,
      "grad_norm": 3.707780361175537,
      "kl": 0.00421142578125,
      "learning_rate": 9.999622329913548e-07,
      "loss": 0.0004,
      "reward": 4.050000190734863,
      "reward_std": 0.46861180663108826,
      "rewards/gpt4o_acoustic_reward": 0.75,
      "rewards/gpt4o_semantic_reward": 3.3000001907348633,
      "step": 36
    },
    {
      "completion_length": 206.25,
      "epoch": 0.003995248893208077,
      "grad_norm": 3.7213122844696045,
      "kl": 0.0020904541015625,
      "learning_rate": 9.999591512454125e-07,
      "loss": 0.0002,
      "reward": 3.7249999046325684,
      "reward_std": 0.6525067090988159,
      "rewards/gpt4o_acoustic_reward": 0.7749999761581421,
      "rewards/gpt4o_semantic_reward": 2.950000286102295,
      "step": 37
    },
    {
      "completion_length": 183.5,
      "epoch": 0.004103228593024511,
      "grad_norm": 4.1779465675354,
      "kl": 0.002410888671875,
      "learning_rate": 9.99955948653397e-07,
      "loss": 0.0002,
      "reward": 3.9000000953674316,
      "reward_std": 0.6134613156318665,
      "rewards/gpt4o_acoustic_reward": 0.7999999523162842,
      "rewards/gpt4o_semantic_reward": 3.1000001430511475,
      "step": 38
    },
    {
      "completion_length": 135.25,
      "epoch": 0.004211208292840946,
      "grad_norm": 5.88501501083374,
      "kl": 0.00311279296875,
      "learning_rate": 9.999526252160825e-07,
      "loss": 0.0003,
      "reward": 3.8000001907348633,
      "reward_std": 0.5650771856307983,
      "rewards/gpt4o_acoustic_reward": 0.800000011920929,
      "rewards/gpt4o_semantic_reward": 3.0,
      "step": 39
    },
    {
      "completion_length": 101.75,
      "epoch": 0.00431918799265738,
      "grad_norm": 5.475719928741455,
      "kl": 0.0032501220703125,
      "learning_rate": 9.999491809342725e-07,
      "loss": 0.0003,
      "reward": 3.6625001430511475,
      "reward_std": 0.5419445037841797,
      "rewards/gpt4o_acoustic_reward": 0.7625000476837158,
      "rewards/gpt4o_semantic_reward": 2.8999998569488525,
      "step": 40
    },
    {
      "completion_length": 109.5,
      "epoch": 0.004427167692473815,
      "grad_norm": 4.616827011108398,
      "kl": 0.00372314453125,
      "learning_rate": 9.999456158087993e-07,
      "loss": 0.0004,
      "reward": 3.799999952316284,
      "reward_std": 0.562907874584198,
      "rewards/gpt4o_acoustic_reward": 0.75,
      "rewards/gpt4o_semantic_reward": 3.0500001907348633,
      "step": 41
    },
    {
      "completion_length": 80.5,
      "epoch": 0.0045351473922902496,
      "grad_norm": 4.345307350158691,
      "kl": 0.0023193359375,
      "learning_rate": 9.999419298405247e-07,
      "loss": 0.0002,
      "reward": 3.75,
      "reward_std": 0.43463942408561707,
      "rewards/gpt4o_acoustic_reward": 0.75,
      "rewards/gpt4o_semantic_reward": 3.0,
      "step": 42
    },
    {
      "completion_length": 151.0,
      "epoch": 0.004643127092106684,
      "grad_norm": 11.16802978515625,
      "kl": 0.007354736328125,
      "learning_rate": 9.999381230303397e-07,
      "loss": 0.0007,
      "reward": 3.9124999046325684,
      "reward_std": 0.7695527076721191,
      "rewards/gpt4o_acoustic_reward": 0.762499988079071,
      "rewards/gpt4o_semantic_reward": 3.1500000953674316,
      "step": 43
    },
    {
      "completion_length": 175.75,
      "epoch": 0.004751106791923119,
      "grad_norm": 4.1300482749938965,
      "kl": 0.0037841796875,
      "learning_rate": 9.999341953791646e-07,
      "loss": 0.0004,
      "reward": 3.6000001430511475,
      "reward_std": 0.5399592518806458,
      "rewards/gpt4o_acoustic_reward": 0.75,
      "rewards/gpt4o_semantic_reward": 2.8499999046325684,
      "step": 44
    },
    {
      "completion_length": 123.5,
      "epoch": 0.004859086491739553,
      "grad_norm": 4.736630916595459,
      "kl": 0.005859375,
      "learning_rate": 9.999301468879483e-07,
      "loss": 0.0006,
      "reward": 4.012499809265137,
      "reward_std": 0.5317236185073853,
      "rewards/gpt4o_acoustic_reward": 0.762499988079071,
      "rewards/gpt4o_semantic_reward": 3.25,
      "step": 45
    },
    {
      "completion_length": 120.25,
      "epoch": 0.004967066191555987,
      "grad_norm": 4.6078596115112305,
      "kl": 0.003692626953125,
      "learning_rate": 9.999259775576694e-07,
      "loss": 0.0004,
      "reward": 4.387499809265137,
      "reward_std": 0.5126998424530029,
      "rewards/gpt4o_acoustic_reward": 0.7874999642372131,
      "rewards/gpt4o_semantic_reward": 3.5999999046325684,
      "step": 46
    },
    {
      "completion_length": 128.25,
      "epoch": 0.005075045891372422,
      "grad_norm": 127.92069244384766,
      "kl": 0.01190185546875,
      "learning_rate": 9.999216873893363e-07,
      "loss": 0.0012,
      "reward": 3.6374998092651367,
      "reward_std": 0.4805222749710083,
      "rewards/gpt4o_acoustic_reward": 0.737500011920929,
      "rewards/gpt4o_semantic_reward": 2.9000000953674316,
      "step": 47
    },
    {
      "completion_length": 233.75,
      "epoch": 0.0051830255911888565,
      "grad_norm": 4.082341194152832,
      "kl": 0.003662109375,
      "learning_rate": 9.999172763839854e-07,
      "loss": 0.0004,
      "reward": 3.549999952316284,
      "reward_std": 0.345066636800766,
      "rewards/gpt4o_acoustic_reward": 0.7000000476837158,
      "rewards/gpt4o_semantic_reward": 2.8500001430511475,
      "step": 48
    },
    {
      "completion_length": 88.25,
      "epoch": 0.005291005291005291,
      "grad_norm": 4.746938228607178,
      "kl": 0.004119873046875,
      "learning_rate": 9.999127445426832e-07,
      "loss": 0.0004,
      "reward": 3.1500000953674316,
      "reward_std": 0.3950536549091339,
      "rewards/gpt4o_acoustic_reward": 0.75,
      "rewards/gpt4o_semantic_reward": 2.4000000953674316,
      "step": 49
    },
    {
      "completion_length": 117.0,
      "epoch": 0.005398984990821726,
      "grad_norm": 4.920814037322998,
      "kl": 0.00421142578125,
      "learning_rate": 9.999080918665248e-07,
      "loss": 0.0004,
      "reward": 4.050000190734863,
      "reward_std": 0.6103179454803467,
      "rewards/gpt4o_acoustic_reward": 0.75,
      "rewards/gpt4o_semantic_reward": 3.3000001907348633,
      "step": 50
    },
    {
      "completion_length": 149.75,
      "epoch": 0.00550696469063816,
      "grad_norm": 6.0530619621276855,
      "kl": 0.0069580078125,
      "learning_rate": 9.999033183566353e-07,
      "loss": 0.0007,
      "reward": 3.7125000953674316,
      "reward_std": 0.6462839841842651,
      "rewards/gpt4o_acoustic_reward": 0.762499988079071,
      "rewards/gpt4o_semantic_reward": 2.950000047683716,
      "step": 51
    },
    {
      "completion_length": 101.25,
      "epoch": 0.005614944390454594,
      "grad_norm": 5.342473983764648,
      "kl": 0.007476806640625,
      "learning_rate": 9.998984240141678e-07,
      "loss": 0.0007,
      "reward": 3.512500047683716,
      "reward_std": 0.9265444278717041,
      "rewards/gpt4o_acoustic_reward": 0.7125000357627869,
      "rewards/gpt4o_semantic_reward": 2.799999952316284,
      "step": 52
    },
    {
      "completion_length": 157.75,
      "epoch": 0.005722924090271029,
      "grad_norm": 4.94399356842041,
      "kl": 0.006072998046875,
      "learning_rate": 9.998934088403062e-07,
      "loss": 0.0006,
      "reward": 3.325000286102295,
      "reward_std": 0.4148828983306885,
      "rewards/gpt4o_acoustic_reward": 0.7250000238418579,
      "rewards/gpt4o_semantic_reward": 2.6000001430511475,
      "step": 53
    },
    {
      "completion_length": 108.0,
      "epoch": 0.0058309037900874635,
      "grad_norm": 6.271359443664551,
      "kl": 0.00469970703125,
      "learning_rate": 9.998882728362618e-07,
      "loss": 0.0005,
      "reward": 3.3500001430511475,
      "reward_std": 0.40565162897109985,
      "rewards/gpt4o_acoustic_reward": 0.699999988079071,
      "rewards/gpt4o_semantic_reward": 2.6500000953674316,
      "step": 54
    },
    {
      "completion_length": 176.0,
      "epoch": 0.005938883489903898,
      "grad_norm": 4.691992282867432,
      "kl": 0.007080078125,
      "learning_rate": 9.99883016003277e-07,
      "loss": 0.0007,
      "reward": 3.7624998092651367,
      "reward_std": 0.3887961506843567,
      "rewards/gpt4o_acoustic_reward": 0.8125,
      "rewards/gpt4o_semantic_reward": 2.9499998092651367,
      "step": 55
    },
    {
      "completion_length": 101.5,
      "epoch": 0.006046863189720333,
      "grad_norm": 4.526642322540283,
      "kl": 0.006072998046875,
      "learning_rate": 9.998776383426216e-07,
      "loss": 0.0006,
      "reward": 3.9250001907348633,
      "reward_std": 0.3565652370452881,
      "rewards/gpt4o_acoustic_reward": 0.7250000238418579,
      "rewards/gpt4o_semantic_reward": 3.1999998092651367,
      "step": 56
    },
    {
      "completion_length": 216.0,
      "epoch": 0.006154842889536767,
      "grad_norm": 4.5564141273498535,
      "kl": 0.00689697265625,
      "learning_rate": 9.998721398555956e-07,
      "loss": 0.0007,
      "reward": 3.987499952316284,
      "reward_std": 0.4106605648994446,
      "rewards/gpt4o_acoustic_reward": 0.7875000238418579,
      "rewards/gpt4o_semantic_reward": 3.200000047683716,
      "step": 57
    },
    {
      "completion_length": 139.75,
      "epoch": 0.006262822589353202,
      "grad_norm": 5.731558322906494,
      "kl": 0.007415771484375,
      "learning_rate": 9.998665205435286e-07,
      "loss": 0.0007,
      "reward": 3.0250000953674316,
      "reward_std": 0.8147455453872681,
      "rewards/gpt4o_acoustic_reward": 0.7250000238418579,
      "rewards/gpt4o_semantic_reward": 2.3000001907348633,
      "step": 58
    },
    {
      "completion_length": 192.25,
      "epoch": 0.006370802289169636,
      "grad_norm": 4.051910400390625,
      "kl": 0.005645751953125,
      "learning_rate": 9.998607804077782e-07,
      "loss": 0.0006,
      "reward": 3.8625001907348633,
      "reward_std": 0.5302895307540894,
      "rewards/gpt4o_acoustic_reward": 0.7124999761581421,
      "rewards/gpt4o_semantic_reward": 3.1499998569488525,
      "step": 59
    },
    {
      "completion_length": 126.75,
      "epoch": 0.00647878198898607,
      "grad_norm": 4.635890960693359,
      "kl": 0.00811767578125,
      "learning_rate": 9.998549194497321e-07,
      "loss": 0.0008,
      "reward": 4.287500381469727,
      "reward_std": 0.4707535207271576,
      "rewards/gpt4o_acoustic_reward": 0.7874999642372131,
      "rewards/gpt4o_semantic_reward": 3.5,
      "step": 60
    },
    {
      "completion_length": 151.75,
      "epoch": 0.0065867616888025055,
      "grad_norm": 4.524758815765381,
      "kl": 0.0115966796875,
      "learning_rate": 9.998489376708072e-07,
      "loss": 0.0012,
      "reward": 3.5,
      "reward_std": 0.48709315061569214,
      "rewards/gpt4o_acoustic_reward": 0.699999988079071,
      "rewards/gpt4o_semantic_reward": 2.799999952316284,
      "step": 61
    },
    {
      "completion_length": 128.25,
      "epoch": 0.00669474138861894,
      "grad_norm": 4.649504661560059,
      "kl": 0.00640869140625,
      "learning_rate": 9.99842835072449e-07,
      "loss": 0.0006,
      "reward": 3.8500003814697266,
      "reward_std": 0.5542749762535095,
      "rewards/gpt4o_acoustic_reward": 0.699999988079071,
      "rewards/gpt4o_semantic_reward": 3.1500000953674316,
      "step": 62
    },
    {
      "completion_length": 182.25,
      "epoch": 0.006802721088435374,
      "grad_norm": 4.80541467666626,
      "kl": 0.0078125,
      "learning_rate": 9.99836611656133e-07,
      "loss": 0.0008,
      "reward": 4.012500286102295,
      "reward_std": 0.721532940864563,
      "rewards/gpt4o_acoustic_reward": 0.7125000357627869,
      "rewards/gpt4o_semantic_reward": 3.3000001907348633,
      "step": 63
    },
    {
      "completion_length": 103.5,
      "epoch": 0.006910700788251809,
      "grad_norm": 4.415143966674805,
      "kl": 0.01043701171875,
      "learning_rate": 9.998302674233629e-07,
      "loss": 0.001,
      "reward": 3.5749998092651367,
      "reward_std": 0.48531410098075867,
      "rewards/gpt4o_acoustic_reward": 0.7749999761581421,
      "rewards/gpt4o_semantic_reward": 2.799999952316284,
      "step": 64
    },
    {
      "completion_length": 155.0,
      "epoch": 0.007018680488068243,
      "grad_norm": 4.847844123840332,
      "kl": 0.00689697265625,
      "learning_rate": 9.998238023756725e-07,
      "loss": 0.0007,
      "reward": 3.950000047683716,
      "reward_std": 0.6172438263893127,
      "rewards/gpt4o_acoustic_reward": 0.7000000476837158,
      "rewards/gpt4o_semantic_reward": 3.25,
      "step": 65
    },
    {
      "completion_length": 110.5,
      "epoch": 0.007126660187884677,
      "grad_norm": 5.6347808837890625,
      "kl": 0.01129150390625,
      "learning_rate": 9.998172165146247e-07,
      "loss": 0.0011,
      "reward": 3.6125001907348633,
      "reward_std": 0.5318603515625,
      "rewards/gpt4o_acoustic_reward": 0.8125,
      "rewards/gpt4o_semantic_reward": 2.799999952316284,
      "step": 66
    },
    {
      "completion_length": 113.0,
      "epoch": 0.0072346398877011124,
      "grad_norm": 4.443930625915527,
      "kl": 0.0118408203125,
      "learning_rate": 9.99810509841811e-07,
      "loss": 0.0012,
      "reward": 3.75,
      "reward_std": 0.5992990732192993,
      "rewards/gpt4o_acoustic_reward": 0.7999999523162842,
      "rewards/gpt4o_semantic_reward": 2.9499998092651367,
      "step": 67
    },
    {
      "completion_length": 56.5,
      "epoch": 0.007342619587517547,
      "grad_norm": 5.055539608001709,
      "kl": 0.0093994140625,
      "learning_rate": 9.998036823588527e-07,
      "loss": 0.0009,
      "reward": 3.674999952316284,
      "reward_std": 0.40850570797920227,
      "rewards/gpt4o_acoustic_reward": 0.7750000357627869,
      "rewards/gpt4o_semantic_reward": 2.9000000953674316,
      "step": 68
    },
    {
      "completion_length": 165.75,
      "epoch": 0.007450599287333981,
      "grad_norm": 4.068113803863525,
      "kl": 0.0087890625,
      "learning_rate": 9.997967340673998e-07,
      "loss": 0.0009,
      "reward": 4.099999904632568,
      "reward_std": 0.4692710340023041,
      "rewards/gpt4o_acoustic_reward": 0.800000011920929,
      "rewards/gpt4o_semantic_reward": 3.299999713897705,
      "step": 69
    },
    {
      "completion_length": 163.25,
      "epoch": 0.007558578987150416,
      "grad_norm": 4.917808532714844,
      "kl": 0.0106201171875,
      "learning_rate": 9.997896649691323e-07,
      "loss": 0.0011,
      "reward": 3.6750001907348633,
      "reward_std": 0.3706974983215332,
      "rewards/gpt4o_acoustic_reward": 0.7749999761581421,
      "rewards/gpt4o_semantic_reward": 2.9000000953674316,
      "step": 70
    },
    {
      "completion_length": 244.5,
      "epoch": 0.00766655868696685,
      "grad_norm": 4.127923965454102,
      "kl": 0.007232666015625,
      "learning_rate": 9.997824750657583e-07,
      "loss": 0.0007,
      "reward": 4.012500286102295,
      "reward_std": 0.571098268032074,
      "rewards/gpt4o_acoustic_reward": 0.7624999284744263,
      "rewards/gpt4o_semantic_reward": 3.25,
      "step": 71
    },
    {
      "completion_length": 95.25,
      "epoch": 0.007774538386783284,
      "grad_norm": 4.875095367431641,
      "kl": 0.0091552734375,
      "learning_rate": 9.997751643590162e-07,
      "loss": 0.0009,
      "reward": 4.025000095367432,
      "reward_std": 0.4940524399280548,
      "rewards/gpt4o_acoustic_reward": 0.7749999761581421,
      "rewards/gpt4o_semantic_reward": 3.25,
      "step": 72
    },
    {
      "completion_length": 118.0,
      "epoch": 0.007882518086599719,
      "grad_norm": 5.769279479980469,
      "kl": 0.01031494140625,
      "learning_rate": 9.997677328506728e-07,
      "loss": 0.001,
      "reward": 3.950000047683716,
      "reward_std": 0.5956119298934937,
      "rewards/gpt4o_acoustic_reward": 0.800000011920929,
      "rewards/gpt4o_semantic_reward": 3.1499998569488525,
      "step": 73
    },
    {
      "completion_length": 68.25,
      "epoch": 0.007990497786416154,
      "grad_norm": 4.916319847106934,
      "kl": 0.00921630859375,
      "learning_rate": 9.997601805425245e-07,
      "loss": 0.0009,
      "reward": 4.212500095367432,
      "reward_std": 0.4324294924736023,
      "rewards/gpt4o_acoustic_reward": 0.762499988079071,
      "rewards/gpt4o_semantic_reward": 3.450000047683716,
      "step": 74
    },
    {
      "completion_length": 131.75,
      "epoch": 0.008098477486232589,
      "grad_norm": 3.795929193496704,
      "kl": 0.007049560546875,
      "learning_rate": 9.997525074363966e-07,
      "loss": 0.0007,
      "reward": 4.099999904632568,
      "reward_std": 0.4008368253707886,
      "rewards/gpt4o_acoustic_reward": 0.75,
      "rewards/gpt4o_semantic_reward": 3.3499999046325684,
      "step": 75
    },
    {
      "completion_length": 265.75,
      "epoch": 0.008206457186049022,
      "grad_norm": 4.116283416748047,
      "kl": 0.01019287109375,
      "learning_rate": 9.997447135341439e-07,
      "loss": 0.001,
      "reward": 4.150000095367432,
      "reward_std": 0.3457174301147461,
      "rewards/gpt4o_acoustic_reward": 0.7000000476837158,
      "rewards/gpt4o_semantic_reward": 3.4499998092651367,
      "step": 76
    },
    {
      "completion_length": 163.75,
      "epoch": 0.008314436885865457,
      "grad_norm": 4.394094944000244,
      "kl": 0.01019287109375,
      "learning_rate": 9.997367988376502e-07,
      "loss": 0.001,
      "reward": 4.1875,
      "reward_std": 0.5921262502670288,
      "rewards/gpt4o_acoustic_reward": 0.7874999642372131,
      "rewards/gpt4o_semantic_reward": 3.4000000953674316,
      "step": 77
    },
    {
      "completion_length": 128.0,
      "epoch": 0.008422416585681892,
      "grad_norm": 4.5872931480407715,
      "kl": 0.0078125,
      "learning_rate": 9.997287633488287e-07,
      "loss": 0.0008,
      "reward": 4.625,
      "reward_std": 0.28030532598495483,
      "rewards/gpt4o_acoustic_reward": 0.8249999284744263,
      "rewards/gpt4o_semantic_reward": 3.799999952316284,
      "step": 78
    },
    {
      "completion_length": 216.75,
      "epoch": 0.008530396285498325,
      "grad_norm": 4.629324913024902,
      "kl": 0.0101318359375,
      "learning_rate": 9.997206070696217e-07,
      "loss": 0.001,
      "reward": 3.362499952316284,
      "reward_std": 0.37205901741981506,
      "rewards/gpt4o_acoustic_reward": 0.7125000357627869,
      "rewards/gpt4o_semantic_reward": 2.6500000953674316,
      "step": 79
    },
    {
      "completion_length": 90.5,
      "epoch": 0.00863837598531476,
      "grad_norm": 5.419586658477783,
      "kl": 0.009033203125,
      "learning_rate": 9.997123300020005e-07,
      "loss": 0.0009,
      "reward": 4.525000095367432,
      "reward_std": 0.48073428869247437,
      "rewards/gpt4o_acoustic_reward": 0.7749999761581421,
      "rewards/gpt4o_semantic_reward": 3.75,
      "step": 80
    },
    {
      "completion_length": 75.25,
      "epoch": 0.008746355685131196,
      "grad_norm": 4.748573303222656,
      "kl": 0.00946044921875,
      "learning_rate": 9.997039321479659e-07,
      "loss": 0.0009,
      "reward": 4.487500190734863,
      "reward_std": 0.2590702176094055,
      "rewards/gpt4o_acoustic_reward": 0.7874999642372131,
      "rewards/gpt4o_semantic_reward": 3.6999998092651367,
      "step": 81
    },
    {
      "completion_length": 66.5,
      "epoch": 0.00885433538494763,
      "grad_norm": 5.085847854614258,
      "kl": 0.01220703125,
      "learning_rate": 9.996954135095478e-07,
      "loss": 0.0012,
      "reward": 4.025000095367432,
      "reward_std": 0.5771558880805969,
      "rewards/gpt4o_acoustic_reward": 0.7749999761581421,
      "rewards/gpt4o_semantic_reward": 3.249999761581421,
      "step": 82
    },
    {
      "completion_length": 124.75,
      "epoch": 0.008962315084764064,
      "grad_norm": 4.828271389007568,
      "kl": 0.00994873046875,
      "learning_rate": 9.99686774088805e-07,
      "loss": 0.001,
      "reward": 4.324999809265137,
      "reward_std": 0.45119568705558777,
      "rewards/gpt4o_acoustic_reward": 0.7749999761581421,
      "rewards/gpt4o_semantic_reward": 3.549999713897705,
      "step": 83
    },
    {
      "completion_length": 118.0,
      "epoch": 0.009070294784580499,
      "grad_norm": 4.358130931854248,
      "kl": 0.0093994140625,
      "learning_rate": 9.996780138878262e-07,
      "loss": 0.0009,
      "reward": 3.5999999046325684,
      "reward_std": 0.49348530173301697,
      "rewards/gpt4o_acoustic_reward": 0.7000000476837158,
      "rewards/gpt4o_semantic_reward": 2.8999998569488525,
      "step": 84
    },
    {
      "completion_length": 164.75,
      "epoch": 0.009178274484396934,
      "grad_norm": 4.527029037475586,
      "kl": 0.01123046875,
      "learning_rate": 9.996691329087285e-07,
      "loss": 0.0011,
      "reward": 4.012499809265137,
      "reward_std": 0.6691046357154846,
      "rewards/gpt4o_acoustic_reward": 0.762499988079071,
      "rewards/gpt4o_semantic_reward": 3.25,
      "step": 85
    },
    {
      "completion_length": 122.75,
      "epoch": 0.009286254184213368,
      "grad_norm": 5.344223976135254,
      "kl": 0.01171875,
      "learning_rate": 9.996601311536586e-07,
      "loss": 0.0012,
      "reward": 4.4375,
      "reward_std": 0.4996621012687683,
      "rewards/gpt4o_acoustic_reward": 0.7875000238418579,
      "rewards/gpt4o_semantic_reward": 3.6500000953674316,
      "step": 86
    },
    {
      "completion_length": 316.0,
      "epoch": 0.009394233884029803,
      "grad_norm": 4.248979568481445,
      "kl": 0.00927734375,
      "learning_rate": 9.996510086247921e-07,
      "loss": 0.0009,
      "reward": 3.7750000953674316,
      "reward_std": 0.5600669384002686,
      "rewards/gpt4o_acoustic_reward": 0.7250000238418579,
      "rewards/gpt4o_semantic_reward": 3.049999952316284,
      "step": 87
    },
    {
      "completion_length": 170.0,
      "epoch": 0.009502213583846238,
      "grad_norm": 4.813039779663086,
      "kl": 0.01385498046875,
      "learning_rate": 9.996417653243345e-07,
      "loss": 0.0014,
      "reward": 3.9375,
      "reward_std": 0.5633493661880493,
      "rewards/gpt4o_acoustic_reward": 0.7874999642372131,
      "rewards/gpt4o_semantic_reward": 3.1500000953674316,
      "step": 88
    },
    {
      "completion_length": 177.0,
      "epoch": 0.009610193283662671,
      "grad_norm": 5.789018630981445,
      "kl": 0.0106201171875,
      "learning_rate": 9.996324012545199e-07,
      "loss": 0.0011,
      "reward": 3.3375000953674316,
      "reward_std": 0.6463441848754883,
      "rewards/gpt4o_acoustic_reward": 0.6875,
      "rewards/gpt4o_semantic_reward": 2.6500000953674316,
      "step": 89
    },
    {
      "completion_length": 63.0,
      "epoch": 0.009718172983479106,
      "grad_norm": 4.934427261352539,
      "kl": 0.011962890625,
      "learning_rate": 9.996229164176112e-07,
      "loss": 0.0012,
      "reward": 4.474999904632568,
      "reward_std": 0.4358587861061096,
      "rewards/gpt4o_acoustic_reward": 0.824999988079071,
      "rewards/gpt4o_semantic_reward": 3.6499998569488525,
      "step": 90
    },
    {
      "completion_length": 61.0,
      "epoch": 0.009826152683295541,
      "grad_norm": 4.437107563018799,
      "kl": 0.0079345703125,
      "learning_rate": 9.996133108159017e-07,
      "loss": 0.0008,
      "reward": 4.274999618530273,
      "reward_std": 0.6249052882194519,
      "rewards/gpt4o_acoustic_reward": 0.7749999761581421,
      "rewards/gpt4o_semantic_reward": 3.5,
      "step": 91
    },
    {
      "completion_length": 127.0,
      "epoch": 0.009934132383111974,
      "grad_norm": 4.14124059677124,
      "kl": 0.0123291015625,
      "learning_rate": 9.996035844517127e-07,
      "loss": 0.0012,
      "reward": 4.27500057220459,
      "reward_std": 0.5244817733764648,
      "rewards/gpt4o_acoustic_reward": 0.7749999761581421,
      "rewards/gpt4o_semantic_reward": 3.5,
      "step": 92
    },
    {
      "completion_length": 96.0,
      "epoch": 0.01004211208292841,
      "grad_norm": 4.307178020477295,
      "kl": 0.0098876953125,
      "learning_rate": 9.995937373273956e-07,
      "loss": 0.001,
      "reward": 3.700000047683716,
      "reward_std": 0.6521656513214111,
      "rewards/gpt4o_acoustic_reward": 0.699999988079071,
      "rewards/gpt4o_semantic_reward": 3.0,
      "step": 93
    },
    {
      "completion_length": 106.75,
      "epoch": 0.010150091782744845,
      "grad_norm": 4.448451995849609,
      "kl": 0.0150146484375,
      "learning_rate": 9.995837694453301e-07,
      "loss": 0.0015,
      "reward": 3.612499952316284,
      "reward_std": 0.62925785779953,
      "rewards/gpt4o_acoustic_reward": 0.762499988079071,
      "rewards/gpt4o_semantic_reward": 2.8499999046325684,
      "step": 94
    },
    {
      "completion_length": 216.25,
      "epoch": 0.010258071482561278,
      "grad_norm": 4.755711555480957,
      "kl": 0.0113525390625,
      "learning_rate": 9.99573680807926e-07,
      "loss": 0.0011,
      "reward": 3.9875001907348633,
      "reward_std": 0.5578835010528564,
      "rewards/gpt4o_acoustic_reward": 0.737500011920929,
      "rewards/gpt4o_semantic_reward": 3.25,
      "step": 95
    },
    {
      "completion_length": 81.5,
      "epoch": 0.010366051182377713,
      "grad_norm": 5.312046527862549,
      "kl": 0.014892578125,
      "learning_rate": 9.995634714176214e-07,
      "loss": 0.0015,
      "reward": 4.125,
      "reward_std": 0.3556162118911743,
      "rewards/gpt4o_acoustic_reward": 0.7749999761581421,
      "rewards/gpt4o_semantic_reward": 3.3499999046325684,
      "step": 96
    },
    {
      "completion_length": 130.25,
      "epoch": 0.010474030882194148,
      "grad_norm": 4.927104473114014,
      "kl": 0.0096435546875,
      "learning_rate": 9.995531412768844e-07,
      "loss": 0.001,
      "reward": 3.674999952316284,
      "reward_std": 0.49086692929267883,
      "rewards/gpt4o_acoustic_reward": 0.7750000357627869,
      "rewards/gpt4o_semantic_reward": 2.9000000953674316,
      "step": 97
    },
    {
      "completion_length": 181.5,
      "epoch": 0.010582010582010581,
      "grad_norm": 4.6555562019348145,
      "kl": 0.01220703125,
      "learning_rate": 9.995426903882117e-07,
      "loss": 0.0012,
      "reward": 2.6125001907348633,
      "reward_std": 0.6184185743331909,
      "rewards/gpt4o_acoustic_reward": 0.6625000238418579,
      "rewards/gpt4o_semantic_reward": 1.9500000476837158,
      "step": 98
    },
    {
      "completion_length": 107.25,
      "epoch": 0.010689990281827016,
      "grad_norm": 6.272807598114014,
      "kl": 0.011474609375,
      "learning_rate": 9.995321187541295e-07,
      "loss": 0.0011,
      "reward": 4.212500095367432,
      "reward_std": 0.6722015142440796,
      "rewards/gpt4o_acoustic_reward": 0.762499988079071,
      "rewards/gpt4o_semantic_reward": 3.4499998092651367,
      "step": 99
    },
    {
      "completion_length": 298.0,
      "epoch": 0.010797969981643452,
      "grad_norm": 5.597227096557617,
      "kl": 0.008544921875,
      "learning_rate": 9.99521426377193e-07,
      "loss": 0.0009,
      "reward": 4.0625,
      "reward_std": 0.5641930103302002,
      "rewards/gpt4o_acoustic_reward": 0.7125000357627869,
      "rewards/gpt4o_semantic_reward": 3.3499999046325684,
      "step": 100
    },
    {
      "completion_length": 87.0,
      "epoch": 0.010905949681459885,
      "grad_norm": 4.724720001220703,
      "kl": 0.0130615234375,
      "learning_rate": 9.995106132599868e-07,
      "loss": 0.0013,
      "reward": 4.0625,
      "reward_std": 0.41856837272644043,
      "rewards/gpt4o_acoustic_reward": 0.8124999403953552,
      "rewards/gpt4o_semantic_reward": 3.25,
      "step": 101
    },
    {
      "completion_length": 115.5,
      "epoch": 0.01101392938127632,
      "grad_norm": 4.891082763671875,
      "kl": 0.01300048828125,
      "learning_rate": 9.994996794051246e-07,
      "loss": 0.0013,
      "reward": 4.074999809265137,
      "reward_std": 0.5545976758003235,
      "rewards/gpt4o_acoustic_reward": 0.7250000238418579,
      "rewards/gpt4o_semantic_reward": 3.3499999046325684,
      "step": 102
    },
    {
      "completion_length": 88.25,
      "epoch": 0.011121909081092755,
      "grad_norm": 5.216233253479004,
      "kl": 0.0274658203125,
      "learning_rate": 9.994886248152488e-07,
      "loss": 0.0027,
      "reward": 4.112500190734863,
      "reward_std": 0.518368124961853,
      "rewards/gpt4o_acoustic_reward": 0.762499988079071,
      "rewards/gpt4o_semantic_reward": 3.3499999046325684,
      "step": 103
    },
    {
      "completion_length": 174.75,
      "epoch": 0.011229888780909188,
      "grad_norm": 5.821533203125,
      "kl": 0.0203857421875,
      "learning_rate": 9.99477449493032e-07,
      "loss": 0.002,
      "reward": 3.7750000953674316,
      "reward_std": 0.46149128675460815,
      "rewards/gpt4o_acoustic_reward": 0.7750000357627869,
      "rewards/gpt4o_semantic_reward": 3.0,
      "step": 104
    },
    {
      "completion_length": 182.75,
      "epoch": 0.011337868480725623,
      "grad_norm": 3.947430372238159,
      "kl": 0.00994873046875,
      "learning_rate": 9.994661534411753e-07,
      "loss": 0.001,
      "reward": 3.875,
      "reward_std": 0.7980617880821228,
      "rewards/gpt4o_acoustic_reward": 0.7749999761581421,
      "rewards/gpt4o_semantic_reward": 3.1000001430511475,
      "step": 105
    },
    {
      "completion_length": 138.75,
      "epoch": 0.011445848180542059,
      "grad_norm": 4.00054931640625,
      "kl": 0.01318359375,
      "learning_rate": 9.994547366624088e-07,
      "loss": 0.0013,
      "reward": 3.7874999046325684,
      "reward_std": 0.23983532190322876,
      "rewards/gpt4o_acoustic_reward": 0.7874999642372131,
      "rewards/gpt4o_semantic_reward": 3.0,
      "step": 106
    },
    {
      "completion_length": 154.25,
      "epoch": 0.011553827880358492,
      "grad_norm": 4.049806118011475,
      "kl": 0.00860595703125,
      "learning_rate": 9.994431991594923e-07,
      "loss": 0.0009,
      "reward": 3.8125,
      "reward_std": 0.6929358839988708,
      "rewards/gpt4o_acoustic_reward": 0.762499988079071,
      "rewards/gpt4o_semantic_reward": 3.049999952316284,
      "step": 107
    },
    {
      "completion_length": 206.25,
      "epoch": 0.011661807580174927,
      "grad_norm": 6.010659217834473,
      "kl": 0.0203857421875,
      "learning_rate": 9.994315409352143e-07,
      "loss": 0.002,
      "reward": 4.262499809265137,
      "reward_std": 0.3880295753479004,
      "rewards/gpt4o_acoustic_reward": 0.762499988079071,
      "rewards/gpt4o_semantic_reward": 3.5,
      "step": 108
    },
    {
      "completion_length": 104.5,
      "epoch": 0.011769787279991362,
      "grad_norm": 6.280111789703369,
      "kl": 0.01318359375,
      "learning_rate": 9.994197619923932e-07,
      "loss": 0.0013,
      "reward": 3.9125001430511475,
      "reward_std": 0.5632771253585815,
      "rewards/gpt4o_acoustic_reward": 0.762499988079071,
      "rewards/gpt4o_semantic_reward": 3.1500000953674316,
      "step": 109
    },
    {
      "completion_length": 197.25,
      "epoch": 0.011877766979807795,
      "grad_norm": 3.786480665206909,
      "kl": 0.01300048828125,
      "learning_rate": 9.994078623338757e-07,
      "loss": 0.0013,
      "reward": 3.737499952316284,
      "reward_std": 0.6773794889450073,
      "rewards/gpt4o_acoustic_reward": 0.737500011920929,
      "rewards/gpt4o_semantic_reward": 3.0,
      "step": 110
    },
    {
      "completion_length": 209.25,
      "epoch": 0.01198574667962423,
      "grad_norm": 6.013534069061279,
      "kl": 0.06884765625,
      "learning_rate": 9.993958419625381e-07,
      "loss": 0.0069,
      "reward": 3.9125001430511475,
      "reward_std": 0.3996429741382599,
      "rewards/gpt4o_acoustic_reward": 0.762499988079071,
      "rewards/gpt4o_semantic_reward": 3.1499998569488525,
      "step": 111
    },
    {
      "completion_length": 200.5,
      "epoch": 0.012093726379440665,
      "grad_norm": 7.105154514312744,
      "kl": 0.0101318359375,
      "learning_rate": 9.993837008812862e-07,
      "loss": 0.001,
      "reward": 4.199999809265137,
      "reward_std": 0.2738218903541565,
      "rewards/gpt4o_acoustic_reward": 0.800000011920929,
      "rewards/gpt4o_semantic_reward": 3.4000000953674316,
      "step": 112
    },
    {
      "completion_length": 108.75,
      "epoch": 0.012201706079257099,
      "grad_norm": 4.474868297576904,
      "kl": 0.0113525390625,
      "learning_rate": 9.993714390930544e-07,
      "loss": 0.0011,
      "reward": 4.137499809265137,
      "reward_std": 0.43214571475982666,
      "rewards/gpt4o_acoustic_reward": 0.6875,
      "rewards/gpt4o_semantic_reward": 3.4499998092651367,
      "step": 113
    },
    {
      "completion_length": 63.5,
      "epoch": 0.012309685779073534,
      "grad_norm": 7.540884494781494,
      "kl": 0.0284423828125,
      "learning_rate": 9.993590566008062e-07,
      "loss": 0.0029,
      "reward": 4.087500095367432,
      "reward_std": 0.3996867537498474,
      "rewards/gpt4o_acoustic_reward": 0.7875000238418579,
      "rewards/gpt4o_semantic_reward": 3.299999952316284,
      "step": 114
    },
    {
      "completion_length": 231.0,
      "epoch": 0.012417665478889969,
      "grad_norm": 3.7791225910186768,
      "kl": 0.01336669921875,
      "learning_rate": 9.993465534075354e-07,
      "loss": 0.0013,
      "reward": 3.9000000953674316,
      "reward_std": 0.6481374502182007,
      "rewards/gpt4o_acoustic_reward": 0.7000000476837158,
      "rewards/gpt4o_semantic_reward": 3.200000047683716,
      "step": 115
    },
    {
      "completion_length": 199.25,
      "epoch": 0.012525645178706404,
      "grad_norm": 3.596402883529663,
      "kl": 0.01287841796875,
      "learning_rate": 9.993339295162635e-07,
      "loss": 0.0013,
      "reward": 3.9000000953674316,
      "reward_std": 0.7463849186897278,
      "rewards/gpt4o_acoustic_reward": 0.699999988079071,
      "rewards/gpt4o_semantic_reward": 3.200000047683716,
      "step": 116
    },
    {
      "completion_length": 134.75,
      "epoch": 0.012633624878522837,
      "grad_norm": 5.444986820220947,
      "kl": 0.055908203125,
      "learning_rate": 9.99321184930042e-07,
      "loss": 0.0056,
      "reward": 2.625,
      "reward_std": 0.2501000165939331,
      "rewards/gpt4o_acoustic_reward": 0.5,
      "rewards/gpt4o_semantic_reward": 2.125,
      "step": 117
    },
    {
      "completion_length": 100.5,
      "epoch": 0.012741604578339272,
      "grad_norm": 5.103532314300537,
      "kl": 0.05419921875,
      "learning_rate": 9.993083196519516e-07,
      "loss": 0.0054,
      "reward": 4.087500095367432,
      "reward_std": 0.6929433345794678,
      "rewards/gpt4o_acoustic_reward": 0.7875000238418579,
      "rewards/gpt4o_semantic_reward": 3.299999952316284,
      "step": 118
    },
    {
      "completion_length": 145.75,
      "epoch": 0.012849584278155708,
      "grad_norm": 6.422066688537598,
      "kl": 0.0164794921875,
      "learning_rate": 9.99295333685102e-07,
      "loss": 0.0016,
      "reward": 4.287499904632568,
      "reward_std": 0.46733689308166504,
      "rewards/gpt4o_acoustic_reward": 0.7874999642372131,
      "rewards/gpt4o_semantic_reward": 3.5,
      "step": 119
    },
    {
      "completion_length": 28.75,
      "epoch": 0.01295756397797214,
      "grad_norm": 5.2015886306762695,
      "kl": 0.1357421875,
      "learning_rate": 9.992822270326319e-07,
      "loss": 0.0135,
      "reward": 4.012499809265137,
      "reward_std": 0.9554036855697632,
      "rewards/gpt4o_acoustic_reward": 0.8125,
      "rewards/gpt4o_semantic_reward": 3.200000047683716,
      "step": 120
    },
    {
      "completion_length": 153.25,
      "epoch": 0.013065543677788576,
      "grad_norm": 5.840940475463867,
      "kl": 0.015869140625,
      "learning_rate": 9.99268999697709e-07,
      "loss": 0.0016,
      "reward": 4.112500190734863,
      "reward_std": 0.50977623462677,
      "rewards/gpt4o_acoustic_reward": 0.8125,
      "rewards/gpt4o_semantic_reward": 3.299999952316284,
      "step": 121
    },
    {
      "completion_length": 77.25,
      "epoch": 0.013173523377605011,
      "grad_norm": 4.669823169708252,
      "kl": 0.09814453125,
      "learning_rate": 9.992556516835313e-07,
      "loss": 0.0098,
      "reward": 3.4250001907348633,
      "reward_std": 0.7450007200241089,
      "rewards/gpt4o_acoustic_reward": 0.7749999761581421,
      "rewards/gpt4o_semantic_reward": 2.6500000953674316,
      "step": 122
    },
    {
      "completion_length": 103.5,
      "epoch": 0.013281503077421444,
      "grad_norm": 4.867919921875,
      "kl": 0.0201416015625,
      "learning_rate": 9.992421829933245e-07,
      "loss": 0.002,
      "reward": 3.2750000953674316,
      "reward_std": 0.5558943152427673,
      "rewards/gpt4o_acoustic_reward": 0.7250000238418579,
      "rewards/gpt4o_semantic_reward": 2.5500001907348633,
      "step": 123
    },
    {
      "completion_length": 94.25,
      "epoch": 0.01338948277723788,
      "grad_norm": 5.765076637268066,
      "kl": 0.0198974609375,
      "learning_rate": 9.992285936303443e-07,
      "loss": 0.002,
      "reward": 3.637500047683716,
      "reward_std": 0.8602118492126465,
      "rewards/gpt4o_acoustic_reward": 0.737500011920929,
      "rewards/gpt4o_semantic_reward": 2.9000000953674316,
      "step": 124
    },
    {
      "completion_length": 102.25,
      "epoch": 0.013497462477054314,
      "grad_norm": 6.424996852874756,
      "kl": 0.1044921875,
      "learning_rate": 9.992148835978757e-07,
      "loss": 0.0105,
      "reward": 4.125,
      "reward_std": 0.45705562829971313,
      "rewards/gpt4o_acoustic_reward": 0.8250000476837158,
      "rewards/gpt4o_semantic_reward": 3.299999952316284,
      "step": 125
    },
    {
      "completion_length": 150.75,
      "epoch": 0.013605442176870748,
      "grad_norm": 5.46433162689209,
      "kl": 0.07470703125,
      "learning_rate": 9.992010528992323e-07,
      "loss": 0.0075,
      "reward": 4.075000286102295,
      "reward_std": 0.3387722969055176,
      "rewards/gpt4o_acoustic_reward": 0.824999988079071,
      "rewards/gpt4o_semantic_reward": 3.25,
      "step": 126
    },
    {
      "completion_length": 200.75,
      "epoch": 0.013713421876687183,
      "grad_norm": 5.124598026275635,
      "kl": 0.0257568359375,
      "learning_rate": 9.991871015377573e-07,
      "loss": 0.0026,
      "reward": 3.8499999046325684,
      "reward_std": 0.6472960710525513,
      "rewards/gpt4o_acoustic_reward": 0.7000000476837158,
      "rewards/gpt4o_semantic_reward": 3.1500000953674316,
      "step": 127
    },
    {
      "completion_length": 311.75,
      "epoch": 0.013821401576503618,
      "grad_norm": 4.458286285400391,
      "kl": 0.01495361328125,
      "learning_rate": 9.991730295168229e-07,
      "loss": 0.0015,
      "reward": 4.387499809265137,
      "reward_std": 0.3887307941913605,
      "rewards/gpt4o_acoustic_reward": 0.7874999642372131,
      "rewards/gpt4o_semantic_reward": 3.5999999046325684,
      "step": 128
    },
    {
      "completion_length": 152.5,
      "epoch": 0.013929381276320051,
      "grad_norm": 4.544926166534424,
      "kl": 0.019775390625,
      "learning_rate": 9.991588368398303e-07,
      "loss": 0.002,
      "reward": 4.112500190734863,
      "reward_std": 0.5667690634727478,
      "rewards/gpt4o_acoustic_reward": 0.7625000476837158,
      "rewards/gpt4o_semantic_reward": 3.3499999046325684,
      "step": 129
    },
    {
      "completion_length": 202.5,
      "epoch": 0.014037360976136486,
      "grad_norm": 5.294361591339111,
      "kl": 0.027587890625,
      "learning_rate": 9.9914452351021e-07,
      "loss": 0.0028,
      "reward": 3.799999952316284,
      "reward_std": 0.42199504375457764,
      "rewards/gpt4o_acoustic_reward": 0.75,
      "rewards/gpt4o_semantic_reward": 3.049999952316284,
      "step": 130
    }
  ],
  "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
}