{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.22038567493112948,
  "eval_steps": 500,
  "global_step": 400,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "completion_length": 320.4375,
      "epoch": 0.0005509641873278236,
      "grad_norm": 0.17644217610359192,
      "kl": 0.0,
      "learning_rate": 0.0,
      "loss": -0.0,
      "reward": 4.869942247867584,
      "reward_std": 0.6764223501086235,
      "rewards/ngram_similarity_reward_func": 0.44751429930329323,
      "rewards/reasoning_quality_reward_func": 0.7849281281232834,
      "rewards/soft_format_reward_func": 0.21875,
      "rewards/sql_execution_reward_func": 2.965624988079071,
      "rewards/xmlcount_reward_func": 0.453125,
      "step": 1
    },
    {
      "completion_length": 299.21875,
      "epoch": 0.0011019283746556473,
      "grad_norm": 0.18495096266269684,
      "kl": 0.0,
      "learning_rate": 3.2e-07,
      "loss": 0.0,
      "reward": 5.52769136428833,
      "reward_std": 0.9518288820981979,
      "rewards/ngram_similarity_reward_func": 0.5377328470349312,
      "rewards/reasoning_quality_reward_func": 0.8191249966621399,
      "rewards/soft_format_reward_func": 0.203125,
      "rewards/sql_execution_reward_func": 3.5302082896232605,
      "rewards/xmlcount_reward_func": 0.4375,
      "step": 2
    },
    {
      "completion_length": 257.5,
      "epoch": 0.001652892561983471,
      "grad_norm": 0.19438059628009796,
      "kl": 0.0001392033045704011,
      "learning_rate": 6.4e-07,
      "loss": 0.0,
      "reward": 4.724464774131775,
      "reward_std": 0.8582430779933929,
      "rewards/ngram_similarity_reward_func": 0.4983517415821552,
      "rewards/reasoning_quality_reward_func": 0.7249968498945236,
      "rewards/soft_format_reward_func": 0.2265625,
      "rewards/sql_execution_reward_func": 2.8214285373687744,
      "rewards/xmlcount_reward_func": 0.453125,
      "step": 3
    },
    {
      "completion_length": 268.0,
      "epoch": 0.0022038567493112946,
      "grad_norm": 0.21523556113243103,
      "kl": 0.00014187309170665685,
      "learning_rate": 9.600000000000001e-07,
      "loss": 0.0,
      "reward": 5.745545268058777,
      "reward_std": 0.6914755180478096,
      "rewards/ngram_similarity_reward_func": 0.6702348068356514,
      "rewards/reasoning_quality_reward_func": 0.7814406454563141,
      "rewards/soft_format_reward_func": 0.1953125,
      "rewards/sql_execution_reward_func": 3.692307710647583,
      "rewards/xmlcount_reward_func": 0.40625,
      "step": 4
    },
    {
      "completion_length": 264.34375,
      "epoch": 0.0027548209366391185,
      "grad_norm": 0.20252738893032074,
      "kl": 0.0001558213734824676,
      "learning_rate": 1.28e-06,
      "loss": 0.0,
      "reward": 5.285050392150879,
      "reward_std": 0.9685250148177147,
      "rewards/ngram_similarity_reward_func": 0.5374472141265869,
      "rewards/reasoning_quality_reward_func": 0.7851031273603439,
      "rewards/soft_format_reward_func": 0.2109375,
      "rewards/sql_execution_reward_func": 3.3218750059604645,
      "rewards/xmlcount_reward_func": 0.4296875,
      "step": 5
    },
    {
      "completion_length": 296.90625,
      "epoch": 0.003305785123966942,
      "grad_norm": 0.18719881772994995,
      "kl": 0.00012105280075047631,
      "learning_rate": 1.6000000000000001e-06,
      "loss": 0.0,
      "reward": 5.8343517780303955,
      "reward_std": 0.7047934383153915,
      "rewards/ngram_similarity_reward_func": 0.7123507112264633,
      "rewards/reasoning_quality_reward_func": 0.8147093504667282,
      "rewards/soft_format_reward_func": 0.2109375,
      "rewards/sql_execution_reward_func": 3.666666626930237,
      "rewards/xmlcount_reward_func": 0.4296875,
      "step": 6
    },
    {
      "completion_length": 331.15625,
      "epoch": 0.003856749311294766,
      "grad_norm": 0.18682371079921722,
      "kl": 0.00012806565064238384,
      "learning_rate": 1.9200000000000003e-06,
      "loss": 0.0,
      "reward": 3.9372432827949524,
      "reward_std": 1.097667172551155,
      "rewards/ngram_similarity_reward_func": 0.44374800473451614,
      "rewards/reasoning_quality_reward_func": 0.7700343802571297,
      "rewards/soft_format_reward_func": 0.21875,
      "rewards/sql_execution_reward_func": 2.0398670434951782,
      "rewards/xmlcount_reward_func": 0.46484375,
      "step": 7
    },
    {
      "completion_length": 199.125,
      "epoch": 0.004407713498622589,
      "grad_norm": 0.232826367020607,
      "kl": 0.00014002827083459124,
      "learning_rate": 2.24e-06,
      "loss": 0.0,
      "reward": 5.875955939292908,
      "reward_std": 0.48954153060913086,
      "rewards/ngram_similarity_reward_func": 0.7089997082948685,
      "rewards/reasoning_quality_reward_func": 0.7411750257015228,
      "rewards/soft_format_reward_func": 0.1953125,
      "rewards/sql_execution_reward_func": 3.8203125,
      "rewards/xmlcount_reward_func": 0.41015625,
      "step": 8
    },
    {
      "completion_length": 259.5,
      "epoch": 0.0049586776859504135,
      "grad_norm": 0.21485431492328644,
      "kl": 0.00015729602819192223,
      "learning_rate": 2.56e-06,
      "loss": 0.0,
      "reward": 5.035609304904938,
      "reward_std": 0.7152721881866455,
      "rewards/ngram_similarity_reward_func": 0.6686828285455704,
      "rewards/reasoning_quality_reward_func": 0.6545812487602234,
      "rewards/soft_format_reward_func": 0.2109375,
      "rewards/sql_execution_reward_func": 3.071720242500305,
      "rewards/xmlcount_reward_func": 0.4296875,
      "step": 9
    },
    {
      "completion_length": 344.5625,
      "epoch": 0.005509641873278237,
      "grad_norm": 0.1722656637430191,
      "kl": 0.0001398367057845462,
      "learning_rate": 2.88e-06,
      "loss": 0.0,
      "reward": 4.781343579292297,
      "reward_std": 0.6409496404230595,
      "rewards/ngram_similarity_reward_func": 0.45625800639390945,
      "rewards/reasoning_quality_reward_func": 0.8253593742847443,
      "rewards/soft_format_reward_func": 0.2109375,
      "rewards/sql_execution_reward_func": 2.859101355075836,
      "rewards/xmlcount_reward_func": 0.4296875,
      "step": 10
    },
    {
      "completion_length": 205.5625,
      "epoch": 0.006060606060606061,
      "grad_norm": 0.22961853444576263,
      "kl": 0.00011612329399213195,
      "learning_rate": 3.2000000000000003e-06,
      "loss": 0.0,
      "reward": 5.289070129394531,
      "reward_std": 0.6560682579874992,
      "rewards/ngram_similarity_reward_func": 0.7160617113113403,
      "rewards/reasoning_quality_reward_func": 0.6841843873262405,
      "rewards/soft_format_reward_func": 0.2265625,
      "rewards/sql_execution_reward_func": 3.216949224472046,
      "rewards/xmlcount_reward_func": 0.4453125,
      "step": 11
    },
    {
      "completion_length": 249.15625,
      "epoch": 0.006611570247933884,
      "grad_norm": 0.23559518158435822,
      "kl": 0.00021559363449341618,
      "learning_rate": 3.52e-06,
      "loss": 0.0,
      "reward": 4.037402093410492,
      "reward_std": 1.2214516550302505,
      "rewards/ngram_similarity_reward_func": 0.2888917699456215,
      "rewards/reasoning_quality_reward_func": 0.7279375046491623,
      "rewards/soft_format_reward_func": 0.234375,
      "rewards/sql_execution_reward_func": 2.3291666507720947,
      "rewards/xmlcount_reward_func": 0.45703125,
      "step": 12
    },
    {
      "completion_length": 370.875,
      "epoch": 0.007162534435261708,
      "grad_norm": 0.15924976766109467,
      "kl": 0.0001305371188209392,
      "learning_rate": 3.8400000000000005e-06,
      "loss": 0.0,
      "reward": 4.453676164150238,
      "reward_std": 1.1237886846065521,
      "rewards/ngram_similarity_reward_func": 0.43848427198827267,
      "rewards/reasoning_quality_reward_func": 0.8854843825101852,
      "rewards/soft_format_reward_func": 0.1953125,
      "rewards/sql_execution_reward_func": 2.5242388248443604,
      "rewards/xmlcount_reward_func": 0.41015625,
      "step": 13
    },
    {
      "completion_length": 242.78125,
      "epoch": 0.007713498622589532,
      "grad_norm": 0.19664712250232697,
      "kl": 0.00015550732859992422,
      "learning_rate": 4.16e-06,
      "loss": 0.0,
      "reward": 5.34897255897522,
      "reward_std": 0.936906673014164,
      "rewards/ngram_similarity_reward_func": 0.4933701157569885,
      "rewards/reasoning_quality_reward_func": 0.7865875214338303,
      "rewards/soft_format_reward_func": 0.2265625,
      "rewards/sql_execution_reward_func": 3.397140085697174,
      "rewards/xmlcount_reward_func": 0.4453125,
      "step": 14
    },
    {
      "completion_length": 275.1875,
      "epoch": 0.008264462809917356,
      "grad_norm": 0.2056989222764969,
      "kl": 0.00014385476606548764,
      "learning_rate": 4.48e-06,
      "loss": 0.0,
      "reward": 4.982160329818726,
      "reward_std": 1.2392726242542267,
      "rewards/ngram_similarity_reward_func": 0.5587007477879524,
      "rewards/reasoning_quality_reward_func": 0.7623156160116196,
      "rewards/soft_format_reward_func": 0.2109375,
      "rewards/sql_execution_reward_func": 3.0283313989639282,
      "rewards/xmlcount_reward_func": 0.421875,
      "step": 15
    },
    {
      "completion_length": 236.71875,
      "epoch": 0.008815426997245178,
      "grad_norm": 0.237776979804039,
      "kl": 0.0001320975770795485,
      "learning_rate": 4.800000000000001e-06,
      "loss": 0.0,
      "reward": 4.581169128417969,
      "reward_std": 0.7548360303044319,
      "rewards/ngram_similarity_reward_func": 0.33604093454778194,
      "rewards/reasoning_quality_reward_func": 0.6466906294226646,
      "rewards/soft_format_reward_func": 0.2109375,
      "rewards/sql_execution_reward_func": 2.949999988079071,
      "rewards/xmlcount_reward_func": 0.4375,
      "step": 16
    },
    {
      "completion_length": 272.8125,
      "epoch": 0.009366391184573003,
      "grad_norm": 0.21105562150478363,
      "kl": 0.00015690055806771852,
      "learning_rate": 5.12e-06,
      "loss": 0.0,
      "reward": 4.165083706378937,
      "reward_std": 1.0532768182456493,
      "rewards/ngram_similarity_reward_func": 0.28024316392838955,
      "rewards/reasoning_quality_reward_func": 0.7582781314849854,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.4000000059604645,
      "rewards/xmlcount_reward_func": 0.484375,
      "step": 17
    },
    {
      "completion_length": 293.0,
      "epoch": 0.009917355371900827,
      "grad_norm": 0.20462216436862946,
      "kl": 0.0001445984344172757,
      "learning_rate": 5.4400000000000004e-06,
      "loss": 0.0,
      "reward": 4.429144859313965,
      "reward_std": 0.9585997201502323,
      "rewards/ngram_similarity_reward_func": 0.5073511302471161,
      "rewards/reasoning_quality_reward_func": 0.74210624396801,
      "rewards/soft_format_reward_func": 0.203125,
      "rewards/sql_execution_reward_func": 2.546875,
      "rewards/xmlcount_reward_func": 0.4296875,
      "step": 18
    },
    {
      "completion_length": 282.9375,
      "epoch": 0.01046831955922865,
      "grad_norm": 0.22315078973770142,
      "kl": 0.00015355706455011386,
      "learning_rate": 5.76e-06,
      "loss": 0.0,
      "reward": 3.7978206276893616,
      "reward_std": 1.1455887258052826,
      "rewards/ngram_similarity_reward_func": 0.44794050231575966,
      "rewards/reasoning_quality_reward_func": 0.6727968528866768,
      "rewards/soft_format_reward_func": 0.2265625,
      "rewards/sql_execution_reward_func": 2.02083339728415,
      "rewards/xmlcount_reward_func": 0.4296875,
      "step": 19
    },
    {
      "completion_length": 337.34375,
      "epoch": 0.011019283746556474,
      "grad_norm": 0.1831493079662323,
      "kl": 0.00014004043623572215,
      "learning_rate": 6.08e-06,
      "loss": 0.0,
      "reward": 4.765153706073761,
      "reward_std": 0.7194762080907822,
      "rewards/ngram_similarity_reward_func": 0.4698445089161396,
      "rewards/reasoning_quality_reward_func": 0.8687468618154526,
      "rewards/soft_format_reward_func": 0.234375,
      "rewards/sql_execution_reward_func": 2.7156249284744263,
      "rewards/xmlcount_reward_func": 0.4765625,
      "step": 20
    },
    {
      "completion_length": 266.34375,
      "epoch": 0.011570247933884297,
      "grad_norm": 0.23525506258010864,
      "kl": 0.0001575565038365312,
      "learning_rate": 6.4000000000000006e-06,
      "loss": 0.0,
      "reward": 4.262574315071106,
      "reward_std": 0.8786862194538116,
      "rewards/ngram_similarity_reward_func": 0.38863373547792435,
      "rewards/reasoning_quality_reward_func": 0.6575343608856201,
      "rewards/soft_format_reward_func": 0.203125,
      "rewards/sql_execution_reward_func": 2.571874976158142,
      "rewards/xmlcount_reward_func": 0.44140625,
      "step": 21
    },
    {
      "completion_length": 296.6875,
      "epoch": 0.012121212121212121,
      "grad_norm": 0.20155483484268188,
      "kl": 0.00013743148701905739,
      "learning_rate": 6.720000000000001e-06,
      "loss": 0.0,
      "reward": 3.4447723627090454,
      "reward_std": 1.341652363538742,
      "rewards/ngram_similarity_reward_func": 0.44517041742801666,
      "rewards/reasoning_quality_reward_func": 0.6897062510251999,
      "rewards/soft_format_reward_func": 0.203125,
      "rewards/sql_execution_reward_func": 1.6692708134651184,
      "rewards/xmlcount_reward_func": 0.4375,
      "step": 22
    },
    {
      "completion_length": 272.90625,
      "epoch": 0.012672176308539946,
      "grad_norm": 0.19946636259555817,
      "kl": 0.00012667695591517258,
      "learning_rate": 7.04e-06,
      "loss": 0.0,
      "reward": 5.500538945198059,
      "reward_std": 0.453230120241642,
      "rewards/ngram_similarity_reward_func": 0.4872107431292534,
      "rewards/reasoning_quality_reward_func": 0.7336406111717224,
      "rewards/soft_format_reward_func": 0.234375,
      "rewards/sql_execution_reward_func": 3.568750023841858,
      "rewards/xmlcount_reward_func": 0.4765625,
      "step": 23
    },
    {
      "completion_length": 297.5625,
      "epoch": 0.013223140495867768,
      "grad_norm": 0.16073261201381683,
      "kl": 0.00012373961362754926,
      "learning_rate": 7.360000000000001e-06,
      "loss": 0.0,
      "reward": 4.424063980579376,
      "reward_std": 0.8705550953745842,
      "rewards/ngram_similarity_reward_func": 0.4677232801914215,
      "rewards/reasoning_quality_reward_func": 0.7610281556844711,
      "rewards/soft_format_reward_func": 0.203125,
      "rewards/sql_execution_reward_func": 2.5781249701976776,
      "rewards/xmlcount_reward_func": 0.4140625,
      "step": 24
    },
    {
      "completion_length": 275.375,
      "epoch": 0.013774104683195593,
      "grad_norm": 0.20881594717502594,
      "kl": 0.00016447595407953486,
      "learning_rate": 7.680000000000001e-06,
      "loss": 0.0,
      "reward": 5.02682363986969,
      "reward_std": 0.6213728822767735,
      "rewards/ngram_similarity_reward_func": 0.5744562447071075,
      "rewards/reasoning_quality_reward_func": 0.7325312495231628,
      "rewards/soft_format_reward_func": 0.1953125,
      "rewards/sql_execution_reward_func": 3.0987426042556763,
      "rewards/xmlcount_reward_func": 0.42578125,
      "step": 25
    },
    {
      "completion_length": 351.875,
      "epoch": 0.014325068870523415,
      "grad_norm": 0.20804105699062347,
      "kl": 0.00012079035695933271,
      "learning_rate": 8.000000000000001e-06,
      "loss": 0.0,
      "reward": 3.953906536102295,
      "reward_std": 1.147509217262268,
      "rewards/ngram_similarity_reward_func": 0.2653359156101942,
      "rewards/reasoning_quality_reward_func": 0.7235343903303146,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.23066109418869,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 26
    },
    {
      "completion_length": 270.65625,
      "epoch": 0.01487603305785124,
      "grad_norm": 0.21302638947963715,
      "kl": 0.00016405481801484711,
      "learning_rate": 8.32e-06,
      "loss": 0.0,
      "reward": 5.348920106887817,
      "reward_std": 0.8308786526322365,
      "rewards/ngram_similarity_reward_func": 0.6120669096708298,
      "rewards/reasoning_quality_reward_func": 0.6852906197309494,
      "rewards/soft_format_reward_func": 0.1953125,
      "rewards/sql_execution_reward_func": 3.449999988079071,
      "rewards/xmlcount_reward_func": 0.40625,
      "step": 27
    },
    {
      "completion_length": 320.78125,
      "epoch": 0.015426997245179064,
      "grad_norm": 0.19438889622688293,
      "kl": 0.00012707046153082047,
      "learning_rate": 8.64e-06,
      "loss": 0.0,
      "reward": 5.530077338218689,
      "reward_std": 0.9891994670033455,
      "rewards/ngram_similarity_reward_func": 0.6032397300004959,
      "rewards/reasoning_quality_reward_func": 0.7643374800682068,
      "rewards/soft_format_reward_func": 0.203125,
      "rewards/sql_execution_reward_func": 3.521875023841858,
      "rewards/xmlcount_reward_func": 0.4375,
      "step": 28
    },
    {
      "completion_length": 287.09375,
      "epoch": 0.015977961432506887,
      "grad_norm": 0.18215666711330414,
      "kl": 0.0001701385626802221,
      "learning_rate": 8.96e-06,
      "loss": 0.0,
      "reward": 4.465075075626373,
      "reward_std": 0.7123237662017345,
      "rewards/ngram_similarity_reward_func": 0.4843626581132412,
      "rewards/reasoning_quality_reward_func": 0.7838374972343445,
      "rewards/soft_format_reward_func": 0.21875,
      "rewards/sql_execution_reward_func": 2.517187535762787,
      "rewards/xmlcount_reward_func": 0.4609375,
      "step": 29
    },
    {
      "completion_length": 334.5,
      "epoch": 0.01652892561983471,
      "grad_norm": 0.19178064167499542,
      "kl": 0.00016806645726319402,
      "learning_rate": 9.280000000000001e-06,
      "loss": 0.0,
      "reward": 4.016714096069336,
      "reward_std": 0.8908075541257858,
      "rewards/ngram_similarity_reward_func": 0.4669796824455261,
      "rewards/reasoning_quality_reward_func": 0.8512969017028809,
      "rewards/soft_format_reward_func": 0.234375,
      "rewards/sql_execution_reward_func": 1.987500036135316,
      "rewards/xmlcount_reward_func": 0.4765625,
      "step": 30
    },
    {
      "completion_length": 291.21875,
      "epoch": 0.017079889807162536,
      "grad_norm": 0.203475221991539,
      "kl": 0.0001841876764956396,
      "learning_rate": 9.600000000000001e-06,
      "loss": 0.0,
      "reward": 3.7459930181503296,
      "reward_std": 0.9910546541213989,
      "rewards/ngram_similarity_reward_func": 0.4136398918926716,
      "rewards/reasoning_quality_reward_func": 0.7011031210422516,
      "rewards/soft_format_reward_func": 0.1796875,
      "rewards/sql_execution_reward_func": 2.0609374940395355,
      "rewards/xmlcount_reward_func": 0.390625,
      "step": 31
    },
    {
      "completion_length": 287.375,
      "epoch": 0.017630853994490357,
      "grad_norm": 0.1784040927886963,
      "kl": 0.0001485680950281676,
      "learning_rate": 9.920000000000002e-06,
      "loss": 0.0,
      "reward": 4.851322293281555,
      "reward_std": 1.543946847319603,
      "rewards/ngram_similarity_reward_func": 0.3204057589173317,
      "rewards/reasoning_quality_reward_func": 0.6876875013113022,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 3.116666615009308,
      "rewards/xmlcount_reward_func": 0.484375,
      "step": 32
    },
    {
      "completion_length": 307.78125,
      "epoch": 0.01818181818181818,
      "grad_norm": 0.1809881329536438,
      "kl": 0.0001775631244527176,
      "learning_rate": 1.024e-05,
      "loss": 0.0,
      "reward": 4.381715297698975,
      "reward_std": 1.5795200169086456,
      "rewards/ngram_similarity_reward_func": 0.36260536313056946,
      "rewards/reasoning_quality_reward_func": 0.841553121805191,
      "rewards/soft_format_reward_func": 0.2265625,
      "rewards/sql_execution_reward_func": 2.490056812763214,
      "rewards/xmlcount_reward_func": 0.4609375,
      "step": 33
    },
    {
      "completion_length": 283.4375,
      "epoch": 0.018732782369146005,
      "grad_norm": 0.20421954989433289,
      "kl": 0.00020579069678205997,
      "learning_rate": 1.056e-05,
      "loss": 0.0,
      "reward": 4.6210795640945435,
      "reward_std": 0.7572419792413712,
      "rewards/ngram_similarity_reward_func": 0.3803170919418335,
      "rewards/reasoning_quality_reward_func": 0.8321687579154968,
      "rewards/soft_format_reward_func": 0.21875,
      "rewards/sql_execution_reward_func": 2.7249999940395355,
      "rewards/xmlcount_reward_func": 0.46484375,
      "step": 34
    },
    {
      "completion_length": 262.4375,
      "epoch": 0.01928374655647383,
      "grad_norm": 0.19971519708633423,
      "kl": 0.00024234944066847675,
      "learning_rate": 1.0880000000000001e-05,
      "loss": 0.0,
      "reward": 4.683843433856964,
      "reward_std": 0.7523762509226799,
      "rewards/ngram_similarity_reward_func": 0.5604191273450851,
      "rewards/reasoning_quality_reward_func": 0.8279343843460083,
      "rewards/soft_format_reward_func": 0.2265625,
      "rewards/sql_execution_reward_func": 2.61580228805542,
      "rewards/xmlcount_reward_func": 0.453125,
      "step": 35
    },
    {
      "completion_length": 357.1875,
      "epoch": 0.019834710743801654,
      "grad_norm": 0.15856172144412994,
      "kl": 0.00022058681133785285,
      "learning_rate": 1.1200000000000001e-05,
      "loss": 0.0,
      "reward": 3.9157042503356934,
      "reward_std": 0.9705793857574463,
      "rewards/ngram_similarity_reward_func": 0.4067963473498821,
      "rewards/reasoning_quality_reward_func": 0.8093781173229218,
      "rewards/soft_format_reward_func": 0.21875,
      "rewards/sql_execution_reward_func": 2.015935957431793,
      "rewards/xmlcount_reward_func": 0.46484375,
      "step": 36
    },
    {
      "completion_length": 370.78125,
      "epoch": 0.020385674931129475,
      "grad_norm": 0.17060106992721558,
      "kl": 0.00019636120123323053,
      "learning_rate": 1.152e-05,
      "loss": 0.0,
      "reward": 3.4376307129859924,
      "reward_std": 0.7870336920022964,
      "rewards/ngram_similarity_reward_func": 0.1624524835497141,
      "rewards/reasoning_quality_reward_func": 0.7595531493425369,
      "rewards/soft_format_reward_func": 0.21875,
      "rewards/sql_execution_reward_func": 1.8437500186264515,
      "rewards/xmlcount_reward_func": 0.453125,
      "step": 37
    },
    {
      "completion_length": 224.90625,
      "epoch": 0.0209366391184573,
      "grad_norm": 0.22568200528621674,
      "kl": 0.00021711869703722186,
      "learning_rate": 1.184e-05,
      "loss": 0.0,
      "reward": 4.367967009544373,
      "reward_std": 0.3832078240811825,
      "rewards/ngram_similarity_reward_func": 0.525226466357708,
      "rewards/reasoning_quality_reward_func": 0.6568031311035156,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.4593749940395355,
      "rewards/xmlcount_reward_func": 0.484375,
      "step": 38
    },
    {
      "completion_length": 307.78125,
      "epoch": 0.021487603305785124,
      "grad_norm": 0.18054841458797455,
      "kl": 0.0002979119963129051,
      "learning_rate": 1.216e-05,
      "loss": 0.0,
      "reward": 5.857931137084961,
      "reward_std": 0.6435607187449932,
      "rewards/ngram_similarity_reward_func": 0.5632529482245445,
      "rewards/reasoning_quality_reward_func": 0.7946781069040298,
      "rewards/soft_format_reward_func": 0.21875,
      "rewards/sql_execution_reward_func": 3.828125,
      "rewards/xmlcount_reward_func": 0.453125,
      "step": 39
    },
    {
      "completion_length": 272.71875,
      "epoch": 0.02203856749311295,
      "grad_norm": 0.2162315398454666,
      "kl": 0.0002977305048261769,
      "learning_rate": 1.248e-05,
      "loss": 0.0,
      "reward": 5.128432512283325,
      "reward_std": 0.481912475079298,
      "rewards/ngram_similarity_reward_func": 0.6281126625835896,
      "rewards/reasoning_quality_reward_func": 0.749278113245964,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 3.0166666358709335,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 40
    },
    {
      "completion_length": 305.78125,
      "epoch": 0.022589531680440773,
      "grad_norm": 0.19125565886497498,
      "kl": 0.0005761704087490216,
      "learning_rate": 1.2800000000000001e-05,
      "loss": 0.0,
      "reward": 4.746596932411194,
      "reward_std": 0.8742257878184319,
      "rewards/ngram_similarity_reward_func": 0.4397520497441292,
      "rewards/reasoning_quality_reward_func": 0.8023906350135803,
      "rewards/soft_format_reward_func": 0.2109375,
      "rewards/sql_execution_reward_func": 2.848204493522644,
      "rewards/xmlcount_reward_func": 0.4453125,
      "step": 41
    },
    {
      "completion_length": 303.4375,
      "epoch": 0.023140495867768594,
      "grad_norm": 0.26427727937698364,
      "kl": 0.0003486615059955511,
      "learning_rate": 1.3120000000000001e-05,
      "loss": 0.0,
      "reward": 4.8488258719444275,
      "reward_std": 1.1479568406939507,
      "rewards/ngram_similarity_reward_func": 0.5970666334033012,
      "rewards/reasoning_quality_reward_func": 0.7001812756061554,
      "rewards/soft_format_reward_func": 0.234375,
      "rewards/sql_execution_reward_func": 2.852359354496002,
      "rewards/xmlcount_reward_func": 0.46484375,
      "step": 42
    },
    {
      "completion_length": 239.90625,
      "epoch": 0.023691460055096418,
      "grad_norm": 0.22181741893291473,
      "kl": 0.0005031100954511203,
      "learning_rate": 1.3440000000000002e-05,
      "loss": 0.0,
      "reward": 4.874095380306244,
      "reward_std": 0.522947933524847,
      "rewards/ngram_similarity_reward_func": 0.6259480491280556,
      "rewards/reasoning_quality_reward_func": 0.6830906271934509,
      "rewards/soft_format_reward_func": 0.234375,
      "rewards/sql_execution_reward_func": 2.869744300842285,
      "rewards/xmlcount_reward_func": 0.4609375,
      "step": 43
    },
    {
      "completion_length": 342.6875,
      "epoch": 0.024242424242424242,
      "grad_norm": 0.19234292209148407,
      "kl": 0.0005275120565784164,
      "learning_rate": 1.376e-05,
      "loss": 0.0,
      "reward": 5.084302663803101,
      "reward_std": 0.5131422616541386,
      "rewards/ngram_similarity_reward_func": 0.5575088039040565,
      "rewards/reasoning_quality_reward_func": 0.8306999951601028,
      "rewards/soft_format_reward_func": 0.21875,
      "rewards/sql_execution_reward_func": 3.012499988079071,
      "rewards/xmlcount_reward_func": 0.46484375,
      "step": 44
    },
    {
      "completion_length": 332.15625,
      "epoch": 0.024793388429752067,
      "grad_norm": 0.19910266995429993,
      "kl": 0.0005306876337272115,
      "learning_rate": 1.408e-05,
      "loss": 0.0,
      "reward": 4.27899956703186,
      "reward_std": 1.0507808476686478,
      "rewards/ngram_similarity_reward_func": 0.36208710074424744,
      "rewards/reasoning_quality_reward_func": 0.7419124841690063,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.4406249821186066,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 45
    },
    {
      "completion_length": 337.65625,
      "epoch": 0.02534435261707989,
      "grad_norm": 0.19201728701591492,
      "kl": 0.0004212798521621153,
      "learning_rate": 1.4400000000000001e-05,
      "loss": 0.0,
      "reward": 4.576432943344116,
      "reward_std": 0.4770899601280689,
      "rewards/ngram_similarity_reward_func": 0.43657680973410606,
      "rewards/reasoning_quality_reward_func": 0.7371218800544739,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.6527343839406967,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 46
    },
    {
      "completion_length": 307.34375,
      "epoch": 0.025895316804407712,
      "grad_norm": 0.1831081062555313,
      "kl": 0.0005304741134750657,
      "learning_rate": 1.4720000000000001e-05,
      "loss": 0.0,
      "reward": 4.887179791927338,
      "reward_std": 1.0689307004213333,
      "rewards/ngram_similarity_reward_func": 0.5851547867059708,
      "rewards/reasoning_quality_reward_func": 0.8176499903202057,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.734375,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 47
    },
    {
      "completion_length": 271.875,
      "epoch": 0.026446280991735537,
      "grad_norm": 0.20844386518001556,
      "kl": 0.0006423939048545435,
      "learning_rate": 1.5040000000000002e-05,
      "loss": 0.0,
      "reward": 5.149140238761902,
      "reward_std": 0.467240609228611,
      "rewards/ngram_similarity_reward_func": 0.7046560496091843,
      "rewards/reasoning_quality_reward_func": 0.8413593918085098,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.853125000372529,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 48
    },
    {
      "completion_length": 253.8125,
      "epoch": 0.02699724517906336,
      "grad_norm": 0.2597540020942688,
      "kl": 0.0006101946491980925,
      "learning_rate": 1.5360000000000002e-05,
      "loss": 0.0,
      "reward": 5.648941874504089,
      "reward_std": 0.7485599722713232,
      "rewards/ngram_similarity_reward_func": 0.5820389166474342,
      "rewards/reasoning_quality_reward_func": 0.7419031411409378,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.574999988079071,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 49
    },
    {
      "completion_length": 309.1875,
      "epoch": 0.027548209366391185,
      "grad_norm": 0.19654162228107452,
      "kl": 0.001052336061547976,
      "learning_rate": 1.5680000000000002e-05,
      "loss": 0.0,
      "reward": 4.872635841369629,
      "reward_std": 0.6185624897480011,
      "rewards/ngram_similarity_reward_func": 0.5218040272593498,
      "rewards/reasoning_quality_reward_func": 0.8336374908685684,
      "rewards/soft_format_reward_func": 0.234375,
      "rewards/sql_execution_reward_func": 2.8023504614830017,
      "rewards/xmlcount_reward_func": 0.48046875,
      "step": 50
    },
    {
      "completion_length": 249.96875,
      "epoch": 0.02809917355371901,
      "grad_norm": 0.21819108724594116,
      "kl": 0.0008287025848403573,
      "learning_rate": 1.6000000000000003e-05,
      "loss": 0.0,
      "reward": 4.764900207519531,
      "reward_std": 0.904246237128973,
      "rewards/ngram_similarity_reward_func": 0.5974992588162422,
      "rewards/reasoning_quality_reward_func": 0.5569843873381615,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.8604166507720947,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 51
    },
    {
      "completion_length": 324.28125,
      "epoch": 0.02865013774104683,
      "grad_norm": 0.19466304779052734,
      "kl": 0.0011398963833926246,
      "learning_rate": 1.632e-05,
      "loss": 0.0,
      "reward": 5.226037383079529,
      "reward_std": 1.1548624336719513,
      "rewards/ngram_similarity_reward_func": 0.6494702398777008,
      "rewards/reasoning_quality_reward_func": 0.8090062439441681,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 3.0292798280715942,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 52
    },
    {
      "completion_length": 317.09375,
      "epoch": 0.029201101928374655,
      "grad_norm": 0.1966371387243271,
      "kl": 0.0016600479721091688,
      "learning_rate": 1.664e-05,
      "loss": 0.0001,
      "reward": 4.67767733335495,
      "reward_std": 0.6141744330525398,
      "rewards/ngram_similarity_reward_func": 0.33328988775610924,
      "rewards/reasoning_quality_reward_func": 0.8397000133991241,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.762499988079071,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 53
    },
    {
      "completion_length": 272.09375,
      "epoch": 0.02975206611570248,
      "grad_norm": 0.2513268291950226,
      "kl": 0.001468202390242368,
      "learning_rate": 1.696e-05,
      "loss": 0.0001,
      "reward": 4.569363176822662,
      "reward_std": 1.2691677808761597,
      "rewards/ngram_similarity_reward_func": 0.31854774057865143,
      "rewards/reasoning_quality_reward_func": 0.7408156394958496,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.759999990463257,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 54
    },
    {
      "completion_length": 314.875,
      "epoch": 0.030303030303030304,
      "grad_norm": 0.18981914222240448,
      "kl": 0.0009886874904623255,
      "learning_rate": 1.728e-05,
      "loss": 0.0,
      "reward": 6.429818868637085,
      "reward_std": 0.2436790969222784,
      "rewards/ngram_similarity_reward_func": 0.8572064787149429,
      "rewards/reasoning_quality_reward_func": 0.8226125091314316,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 4.0,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 55
    },
    {
      "completion_length": 340.5625,
      "epoch": 0.03085399449035813,
      "grad_norm": 0.21622487902641296,
      "kl": 0.001694115751888603,
      "learning_rate": 1.76e-05,
      "loss": 0.0001,
      "reward": 5.124483823776245,
      "reward_std": 0.8758301958441734,
      "rewards/ngram_similarity_reward_func": 0.46372122317552567,
      "rewards/reasoning_quality_reward_func": 0.8779499679803848,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.0406249463558197,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 56
    },
    {
      "completion_length": 323.21875,
      "epoch": 0.03140495867768595,
      "grad_norm": 0.21447351574897766,
      "kl": 0.001411169650964439,
      "learning_rate": 1.792e-05,
      "loss": 0.0001,
      "reward": 5.151256680488586,
      "reward_std": 0.8374371267855167,
      "rewards/ngram_similarity_reward_func": 0.4718443304300308,
      "rewards/reasoning_quality_reward_func": 0.8137874901294708,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.115625023841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 57
    },
    {
      "completion_length": 298.5625,
      "epoch": 0.031955922865013774,
      "grad_norm": 0.23229965567588806,
      "kl": 0.0021541172463912517,
      "learning_rate": 1.824e-05,
      "loss": 0.0001,
      "reward": 4.667271137237549,
      "reward_std": 0.8938897773623466,
      "rewards/ngram_similarity_reward_func": 0.5010946206748486,
      "rewards/reasoning_quality_reward_func": 0.7047874927520752,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.719201445579529,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 58
    },
    {
      "completion_length": 329.03125,
      "epoch": 0.032506887052341595,
      "grad_norm": 0.17888040840625763,
      "kl": 0.00264827988576144,
      "learning_rate": 1.8560000000000002e-05,
      "loss": 0.0001,
      "reward": 5.337520718574524,
      "reward_std": 0.9201234001666307,
      "rewards/ngram_similarity_reward_func": 0.749447762966156,
      "rewards/reasoning_quality_reward_func": 0.8555187433958054,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 3.005991816520691,
      "rewards/xmlcount_reward_func": 0.484375,
      "step": 59
    },
    {
      "completion_length": 319.21875,
      "epoch": 0.03305785123966942,
      "grad_norm": 0.2067958414554596,
      "kl": 0.00219221122097224,
      "learning_rate": 1.8880000000000002e-05,
      "loss": 0.0001,
      "reward": 4.817923903465271,
      "reward_std": 0.566604495048523,
      "rewards/ngram_similarity_reward_func": 0.48209819197654724,
      "rewards/reasoning_quality_reward_func": 0.862500011920929,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.7350446581840515,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 60
    },
    {
      "completion_length": 340.125,
      "epoch": 0.03360881542699724,
      "grad_norm": 0.21346579492092133,
      "kl": 0.0018046289042104036,
      "learning_rate": 1.9200000000000003e-05,
      "loss": 0.0001,
      "reward": 3.4821184873580933,
      "reward_std": 0.46908816881477833,
      "rewards/ngram_similarity_reward_func": 0.29987743869423866,
      "rewards/reasoning_quality_reward_func": 0.7490937560796738,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 1.6948660612106323,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 61
    },
    {
      "completion_length": 288.125,
      "epoch": 0.03415977961432507,
      "grad_norm": 0.22564105689525604,
      "kl": 0.002974803384859115,
      "learning_rate": 1.9520000000000003e-05,
      "loss": 0.0001,
      "reward": 4.813517332077026,
      "reward_std": 0.8355115279555321,
      "rewards/ngram_similarity_reward_func": 0.521810919046402,
      "rewards/reasoning_quality_reward_func": 0.6931937485933304,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.8641376346349716,
      "rewards/xmlcount_reward_func": 0.484375,
      "step": 62
    },
    {
      "completion_length": 298.5,
      "epoch": 0.03471074380165289,
      "grad_norm": 0.26067525148391724,
      "kl": 0.003967338881921023,
      "learning_rate": 1.9840000000000003e-05,
      "loss": 0.0002,
      "reward": 5.493682265281677,
      "reward_std": 0.5841312408447266,
      "rewards/ngram_similarity_reward_func": 0.6349512040615082,
      "rewards/reasoning_quality_reward_func": 0.7571687400341034,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 3.375,
      "rewards/xmlcount_reward_func": 0.484375,
      "step": 63
    },
    {
      "completion_length": 342.46875,
      "epoch": 0.03526170798898071,
      "grad_norm": 0.20481635630130768,
      "kl": 0.0031379713909700513,
      "learning_rate": 2.016e-05,
      "loss": 0.0001,
      "reward": 4.404810309410095,
      "reward_std": 0.7213887944817543,
      "rewards/ngram_similarity_reward_func": 0.5470792278647423,
      "rewards/reasoning_quality_reward_func": 0.7108562588691711,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.3968749940395355,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 64
    },
    {
      "completion_length": 308.09375,
      "epoch": 0.03581267217630854,
      "grad_norm": 0.1998625546693802,
      "kl": 0.002738240815233439,
      "learning_rate": 2.048e-05,
      "loss": 0.0001,
      "reward": 5.0117292404174805,
      "reward_std": 0.6765128672122955,
      "rewards/ngram_similarity_reward_func": 0.5754932910203934,
      "rewards/reasoning_quality_reward_func": 0.6852968633174896,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.000939190387726,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 65
    },
    {
      "completion_length": 299.15625,
      "epoch": 0.03636363636363636,
      "grad_norm": 0.22531455755233765,
      "kl": 0.002967089123558253,
      "learning_rate": 2.08e-05,
      "loss": 0.0001,
      "reward": 5.279067158699036,
      "reward_std": 0.7429635189473629,
      "rewards/ngram_similarity_reward_func": 0.7201411053538322,
      "rewards/reasoning_quality_reward_func": 0.8082750141620636,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.0084635615348816,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 66
    },
    {
      "completion_length": 331.96875,
      "epoch": 0.03691460055096419,
      "grad_norm": 0.2227819859981537,
      "kl": 0.003261425590608269,
      "learning_rate": 2.112e-05,
      "loss": 0.0001,
      "reward": 3.727479577064514,
      "reward_std": 0.9151296503841877,
      "rewards/ngram_similarity_reward_func": 0.5651393309235573,
      "rewards/reasoning_quality_reward_func": 0.7591843605041504,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 1.6531558632850647,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 67
    },
    {
      "completion_length": 405.5,
      "epoch": 0.03746556473829201,
      "grad_norm": 0.15338148176670074,
      "kl": 0.0035817845491692424,
      "learning_rate": 2.144e-05,
      "loss": 0.0001,
      "reward": 4.8438286781311035,
      "reward_std": 0.8857397325336933,
      "rewards/ngram_similarity_reward_func": 0.44733596593141556,
      "rewards/reasoning_quality_reward_func": 0.8080812692642212,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.846223920583725,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 68
    },
    {
      "completion_length": 378.9375,
      "epoch": 0.03801652892561983,
      "grad_norm": 0.2340143769979477,
      "kl": 0.004578751279041171,
      "learning_rate": 2.1760000000000002e-05,
      "loss": 0.0002,
      "reward": 5.061207890510559,
      "reward_std": 0.4648609720170498,
      "rewards/ngram_similarity_reward_func": 0.46709219366312027,
      "rewards/reasoning_quality_reward_func": 0.7191156446933746,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.125,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 69
    },
    {
      "completion_length": 333.0,
      "epoch": 0.03856749311294766,
      "grad_norm": 0.24051377177238464,
      "kl": 0.005106541211716831,
      "learning_rate": 2.2080000000000002e-05,
      "loss": 0.0002,
      "reward": 4.814022898674011,
      "reward_std": 0.6757130995392799,
      "rewards/ngram_similarity_reward_func": 0.5506042391061783,
      "rewards/reasoning_quality_reward_func": 0.6946687549352646,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.8187499940395355,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 70
    },
    {
      "completion_length": 354.625,
      "epoch": 0.03911845730027548,
      "grad_norm": 0.19152939319610596,
      "kl": 0.004864028189331293,
      "learning_rate": 2.2400000000000002e-05,
      "loss": 0.0002,
      "reward": 5.63380229473114,
      "reward_std": 0.28776115737855434,
      "rewards/ngram_similarity_reward_func": 0.5850102305412292,
      "rewards/reasoning_quality_reward_func": 0.8612155914306641,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.437576413154602,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 71
    },
    {
      "completion_length": 356.84375,
      "epoch": 0.03966942148760331,
      "grad_norm": 0.2026541531085968,
      "kl": 0.005835927091538906,
      "learning_rate": 2.272e-05,
      "loss": 0.0002,
      "reward": 4.283694386482239,
      "reward_std": 0.9084374904632568,
      "rewards/ngram_similarity_reward_func": 0.6248131990432739,
      "rewards/reasoning_quality_reward_func": 0.8549750000238419,
      "rewards/soft_format_reward_func": 0.234375,
      "rewards/sql_execution_reward_func": 2.0812499411404133,
      "rewards/xmlcount_reward_func": 0.48828125,
      "step": 72
    },
    {
      "completion_length": 390.125,
      "epoch": 0.04022038567493113,
      "grad_norm": 0.17204055190086365,
      "kl": 0.005907174781896174,
      "learning_rate": 2.304e-05,
      "loss": 0.0002,
      "reward": 5.143143951892853,
      "reward_std": 0.992171498015523,
      "rewards/ngram_similarity_reward_func": 0.5721568688750267,
      "rewards/reasoning_quality_reward_func": 0.9220718592405319,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.8989152908325195,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 73
    },
    {
      "completion_length": 348.25,
      "epoch": 0.04077134986225895,
      "grad_norm": 0.21762435138225555,
      "kl": 0.007789914496243,
      "learning_rate": 2.336e-05,
      "loss": 0.0003,
      "reward": 5.3299864530563354,
      "reward_std": 0.6468756832182407,
      "rewards/ngram_similarity_reward_func": 0.6819701194763184,
      "rewards/reasoning_quality_reward_func": 0.632903128862381,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.2651134729385376,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 74
    },
    {
      "completion_length": 438.8125,
      "epoch": 0.04132231404958678,
      "grad_norm": 0.17274680733680725,
      "kl": 0.007537530153058469,
      "learning_rate": 2.368e-05,
      "loss": 0.0003,
      "reward": 4.766891419887543,
      "reward_std": 1.051968291401863,
      "rewards/ngram_similarity_reward_func": 0.5098857581615448,
      "rewards/reasoning_quality_reward_func": 0.9213312268257141,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.585674434900284,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 75
    },
    {
      "completion_length": 376.46875,
      "epoch": 0.0418732782369146,
      "grad_norm": 0.180837482213974,
      "kl": 0.008852535858750343,
      "learning_rate": 2.4e-05,
      "loss": 0.0004,
      "reward": 5.6589401960372925,
      "reward_std": 1.0766965709626675,
      "rewards/ngram_similarity_reward_func": 0.80103699862957,
      "rewards/reasoning_quality_reward_func": 0.8266531229019165,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.28125,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 76
    },
    {
      "completion_length": 471.0,
      "epoch": 0.04242424242424243,
      "grad_norm": 0.16251911222934723,
      "kl": 0.00922732800245285,
      "learning_rate": 2.432e-05,
      "loss": 0.0004,
      "reward": 5.394707798957825,
      "reward_std": 0.4166533127427101,
      "rewards/ngram_similarity_reward_func": 0.5753259584307671,
      "rewards/reasoning_quality_reward_func": 0.8919156342744827,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.1774663478136063,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 77
    },
    {
      "completion_length": 373.875,
      "epoch": 0.04297520661157025,
      "grad_norm": 0.2256554365158081,
      "kl": 0.014685155590996146,
      "learning_rate": 2.464e-05,
      "loss": 0.0006,
      "reward": 4.077665746212006,
      "reward_std": 0.8099695518612862,
      "rewards/ngram_similarity_reward_func": 0.4379873350262642,
      "rewards/reasoning_quality_reward_func": 0.6830812394618988,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.20659720338881,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 78
    },
    {
      "completion_length": 333.0625,
      "epoch": 0.04352617079889807,
      "grad_norm": 0.2100255936384201,
      "kl": 0.014230422209948301,
      "learning_rate": 2.496e-05,
      "loss": 0.0006,
      "reward": 5.660989880561829,
      "reward_std": 1.0408196579664946,
      "rewards/ngram_similarity_reward_func": 0.630623109638691,
      "rewards/reasoning_quality_reward_func": 0.8808874785900116,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.3994791507720947,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 79
    },
    {
      "completion_length": 366.9375,
      "epoch": 0.0440771349862259,
      "grad_norm": 0.21354793012142181,
      "kl": 0.012891383143141866,
      "learning_rate": 2.5280000000000002e-05,
      "loss": 0.0005,
      "reward": 4.84545624256134,
      "reward_std": 0.8674880415201187,
      "rewards/ngram_similarity_reward_func": 0.7397499680519104,
      "rewards/reasoning_quality_reward_func": 0.883831262588501,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.471875049173832,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 80
    },
    {
      "completion_length": 383.1875,
      "epoch": 0.04462809917355372,
      "grad_norm": 0.19274823367595673,
      "kl": 0.0135972173884511,
      "learning_rate": 2.5600000000000002e-05,
      "loss": 0.0005,
      "reward": 5.688392877578735,
      "reward_std": 0.9918231293559074,
      "rewards/ngram_similarity_reward_func": 0.663111686706543,
      "rewards/reasoning_quality_reward_func": 0.8143437653779984,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.4609375,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 81
    },
    {
      "completion_length": 352.4375,
      "epoch": 0.045179063360881545,
      "grad_norm": 0.21800082921981812,
      "kl": 0.014336698222905397,
      "learning_rate": 2.5920000000000003e-05,
      "loss": 0.0006,
      "reward": 4.995338678359985,
      "reward_std": 1.0035785101354122,
      "rewards/ngram_similarity_reward_func": 0.6592198982834816,
      "rewards/reasoning_quality_reward_func": 0.804778128862381,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.781340718269348,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 82
    },
    {
      "completion_length": 383.40625,
      "epoch": 0.045730027548209366,
      "grad_norm": 0.19984574615955353,
      "kl": 0.010919353924691677,
      "learning_rate": 2.6240000000000003e-05,
      "loss": 0.0004,
      "reward": 5.569505095481873,
      "reward_std": 1.1113542802631855,
      "rewards/ngram_similarity_reward_func": 0.5657332241535187,
      "rewards/reasoning_quality_reward_func": 0.8553343713283539,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.3984374403953552,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 83
    },
    {
      "completion_length": 350.46875,
      "epoch": 0.04628099173553719,
      "grad_norm": 0.212202787399292,
      "kl": 0.014418089529499412,
      "learning_rate": 2.6560000000000003e-05,
      "loss": 0.0006,
      "reward": 5.024170517921448,
      "reward_std": 1.0424827486276627,
      "rewards/ngram_similarity_reward_func": 0.561284177005291,
      "rewards/reasoning_quality_reward_func": 0.9319968670606613,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.7808894515037537,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 84
    },
    {
      "completion_length": 385.34375,
      "epoch": 0.046831955922865015,
      "grad_norm": 0.21742691099643707,
      "kl": 0.012930582743138075,
      "learning_rate": 2.6880000000000004e-05,
      "loss": 0.0005,
      "reward": 5.104225754737854,
      "reward_std": 0.8797705173492432,
      "rewards/ngram_similarity_reward_func": 0.7052654698491096,
      "rewards/reasoning_quality_reward_func": 0.8242656290531158,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.8325071334838867,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 85
    },
    {
      "completion_length": 354.46875,
      "epoch": 0.047382920110192836,
      "grad_norm": 0.1901436597108841,
      "kl": 0.013906869105994701,
      "learning_rate": 2.7200000000000004e-05,
      "loss": 0.0006,
      "reward": 4.581935405731201,
      "reward_std": 0.7342670653015375,
      "rewards/ngram_similarity_reward_func": 0.7266011983156204,
      "rewards/reasoning_quality_reward_func": 0.8053343743085861,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.299999989569187,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 86
    },
    {
      "completion_length": 398.84375,
      "epoch": 0.047933884297520664,
      "grad_norm": 0.2085219770669937,
      "kl": 0.013961485121399164,
      "learning_rate": 2.752e-05,
      "loss": 0.0006,
      "reward": 4.450812578201294,
      "reward_std": 0.900760143995285,
      "rewards/ngram_similarity_reward_func": 0.5180609971284866,
      "rewards/reasoning_quality_reward_func": 0.8349218815565109,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.3595485985279083,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 87
    },
    {
      "completion_length": 403.65625,
      "epoch": 0.048484848484848485,
      "grad_norm": 0.2017008662223816,
      "kl": 0.014511758927255869,
      "learning_rate": 2.784e-05,
      "loss": 0.0006,
      "reward": 4.741463780403137,
      "reward_std": 1.2182637378573418,
      "rewards/ngram_similarity_reward_func": 0.5396230518817902,
      "rewards/reasoning_quality_reward_func": 0.8330906182527542,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.6187499463558197,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 88
    },
    {
      "completion_length": 378.15625,
      "epoch": 0.049035812672176306,
      "grad_norm": 0.1903093159198761,
      "kl": 0.015393777517601848,
      "learning_rate": 2.816e-05,
      "loss": 0.0006,
      "reward": 5.622711896896362,
      "reward_std": 0.33988741785287857,
      "rewards/ngram_similarity_reward_func": 0.6376026347279549,
      "rewards/reasoning_quality_reward_func": 0.8601093888282776,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.375,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 89
    },
    {
      "completion_length": 378.59375,
      "epoch": 0.049586776859504134,
      "grad_norm": 0.22570684552192688,
      "kl": 0.013180104549974203,
      "learning_rate": 2.8480000000000002e-05,
      "loss": 0.0005,
      "reward": 5.070401191711426,
      "reward_std": 1.1512226611375809,
      "rewards/ngram_similarity_reward_func": 0.3747144341468811,
      "rewards/reasoning_quality_reward_func": 0.8507468998432159,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.0949398279190063,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 90
    },
    {
      "completion_length": 302.5,
      "epoch": 0.050137741046831955,
      "grad_norm": 0.2454530894756317,
      "kl": 0.015146256890147924,
      "learning_rate": 2.8800000000000002e-05,
      "loss": 0.0006,
      "reward": 6.211691379547119,
      "reward_std": 0.3611980900168419,
      "rewards/ngram_similarity_reward_func": 0.6866343636065722,
      "rewards/reasoning_quality_reward_func": 0.8375000059604645,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.937557339668274,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 91
    },
    {
      "completion_length": 401.9375,
      "epoch": 0.05068870523415978,
      "grad_norm": 0.18618059158325195,
      "kl": 0.014557188609614968,
      "learning_rate": 2.9120000000000002e-05,
      "loss": 0.0006,
      "reward": 4.166280746459961,
      "reward_std": 0.3355471342802048,
      "rewards/ngram_similarity_reward_func": 0.4638994038105011,
      "rewards/reasoning_quality_reward_func": 0.7836312651634216,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.1687499824911356,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 92
    },
    {
      "completion_length": 389.28125,
      "epoch": 0.0512396694214876,
      "grad_norm": 0.20728613436222076,
      "kl": 0.016132957069203258,
      "learning_rate": 2.9440000000000003e-05,
      "loss": 0.0006,
      "reward": 4.4673961997032166,
      "reward_std": 0.6168640088289976,
      "rewards/ngram_similarity_reward_func": 0.32384203746914864,
      "rewards/reasoning_quality_reward_func": 0.7729812413454056,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.6205728948116302,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 93
    },
    {
      "completion_length": 400.59375,
      "epoch": 0.051790633608815424,
      "grad_norm": 0.2369987517595291,
      "kl": 0.014066802803426981,
      "learning_rate": 2.9760000000000003e-05,
      "loss": 0.0006,
      "reward": 4.345146119594574,
      "reward_std": 0.879574716091156,
      "rewards/ngram_similarity_reward_func": 0.46518055349588394,
      "rewards/reasoning_quality_reward_func": 0.7955906093120575,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.3343749940395355,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 94
    },
    {
      "completion_length": 404.78125,
      "epoch": 0.05234159779614325,
      "grad_norm": 0.20136411488056183,
      "kl": 0.01331414608284831,
      "learning_rate": 3.0080000000000003e-05,
      "loss": 0.0005,
      "reward": 5.412437796592712,
      "reward_std": 0.6525780213996768,
      "rewards/ngram_similarity_reward_func": 0.6002879440784454,
      "rewards/reasoning_quality_reward_func": 0.9183999747037888,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.143749952316284,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 95
    },
    {
      "completion_length": 484.125,
      "epoch": 0.05289256198347107,
      "grad_norm": 0.17763394117355347,
      "kl": 0.013403360731899738,
      "learning_rate": 3.0400000000000004e-05,
      "loss": 0.0005,
      "reward": 3.483354926109314,
      "reward_std": 0.8336529843509197,
      "rewards/ngram_similarity_reward_func": 0.5245249643921852,
      "rewards/reasoning_quality_reward_func": 0.8884250074625015,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 1.320405000820756,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 96
    },
    {
      "completion_length": 392.09375,
      "epoch": 0.0534435261707989,
      "grad_norm": 0.20346638560295105,
      "kl": 0.015865507069975138,
      "learning_rate": 3.0720000000000004e-05,
      "loss": 0.0006,
      "reward": 5.243882417678833,
      "reward_std": 0.398764681071043,
      "rewards/ngram_similarity_reward_func": 0.4689377136528492,
      "rewards/reasoning_quality_reward_func": 0.8582781255245209,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.166666626930237,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 97
    },
    {
      "completion_length": 401.84375,
      "epoch": 0.05399449035812672,
      "grad_norm": 0.199562206864357,
      "kl": 0.014394007856026292,
      "learning_rate": 3.104e-05,
      "loss": 0.0006,
      "reward": 5.247801661491394,
      "reward_std": 0.48932191729545593,
      "rewards/ngram_similarity_reward_func": 0.6490910351276398,
      "rewards/reasoning_quality_reward_func": 0.8779406100511551,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.970770038664341,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 98
    },
    {
      "completion_length": 400.0625,
      "epoch": 0.05454545454545454,
      "grad_norm": 0.2115149348974228,
      "kl": 0.01679670810699463,
      "learning_rate": 3.1360000000000005e-05,
      "loss": 0.0007,
      "reward": 5.41612982749939,
      "reward_std": 0.8755900040268898,
      "rewards/ngram_similarity_reward_func": 0.390383243560791,
      "rewards/reasoning_quality_reward_func": 0.894496887922287,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.381250023841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 99
    },
    {
      "completion_length": 434.4375,
      "epoch": 0.05509641873278237,
      "grad_norm": 0.17584332823753357,
      "kl": 0.013844694709405303,
      "learning_rate": 3.168e-05,
      "loss": 0.0006,
      "reward": 4.902067303657532,
      "reward_std": 0.9674029164016247,
      "rewards/ngram_similarity_reward_func": 0.5426275283098221,
      "rewards/reasoning_quality_reward_func": 0.8386062383651733,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.770833373069763,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 100
    },
    {
      "completion_length": 353.5,
      "epoch": 0.05564738292011019,
      "grad_norm": 0.22478629648685455,
      "kl": 0.01723932335153222,
      "learning_rate": 3.2000000000000005e-05,
      "loss": 0.0007,
      "reward": 4.4326289892196655,
      "reward_std": 0.7812964059412479,
      "rewards/ngram_similarity_reward_func": 0.3962555266916752,
      "rewards/reasoning_quality_reward_func": 0.7740812748670578,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.5122923254966736,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 101
    },
    {
      "completion_length": 414.90625,
      "epoch": 0.05619834710743802,
      "grad_norm": 0.18412569165229797,
      "kl": 0.01586929732002318,
      "learning_rate": 3.232e-05,
      "loss": 0.0006,
      "reward": 5.050408601760864,
      "reward_std": 0.8182419426739216,
      "rewards/ngram_similarity_reward_func": 0.5689711719751358,
      "rewards/reasoning_quality_reward_func": 0.8533124923706055,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.878125011920929,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 102
    },
    {
      "completion_length": 345.75,
      "epoch": 0.05674931129476584,
      "grad_norm": 0.22483907639980316,
      "kl": 0.015629392815753818,
      "learning_rate": 3.264e-05,
      "loss": 0.0006,
      "reward": 4.988196134567261,
      "reward_std": 0.6413777694106102,
      "rewards/ngram_similarity_reward_func": 0.3026648461818695,
      "rewards/reasoning_quality_reward_func": 0.8097499907016754,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 3.137499988079071,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 103
    },
    {
      "completion_length": 321.96875,
      "epoch": 0.05730027548209366,
      "grad_norm": 0.2463773488998413,
      "kl": 0.016125879483297467,
      "learning_rate": 3.296e-05,
      "loss": 0.0006,
      "reward": 5.780092716217041,
      "reward_std": 1.1997034065425396,
      "rewards/ngram_similarity_reward_func": 0.8225553929805756,
      "rewards/reasoning_quality_reward_func": 0.813787505030632,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.393750011920929,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 104
    },
    {
      "completion_length": 416.53125,
      "epoch": 0.05785123966942149,
      "grad_norm": 0.17782755196094513,
      "kl": 0.014560654060915112,
      "learning_rate": 3.328e-05,
      "loss": 0.0006,
      "reward": 5.53911292552948,
      "reward_std": 1.1121220849454403,
      "rewards/ngram_similarity_reward_func": 0.6285868287086487,
      "rewards/reasoning_quality_reward_func": 0.9138000011444092,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.2467262148857117,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 105
    },
    {
      "completion_length": 379.03125,
      "epoch": 0.05840220385674931,
      "grad_norm": 0.1791757345199585,
      "kl": 0.01170771149918437,
      "learning_rate": 3.3600000000000004e-05,
      "loss": 0.0005,
      "reward": 5.1766053438186646,
      "reward_std": 0.366503793746233,
      "rewards/ngram_similarity_reward_func": 0.859068438410759,
      "rewards/reasoning_quality_reward_func": 0.9066281318664551,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.6609087586402893,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 106
    },
    {
      "completion_length": 271.5625,
      "epoch": 0.05895316804407714,
      "grad_norm": 0.2364310324192047,
      "kl": 0.017134695779532194,
      "learning_rate": 3.392e-05,
      "loss": 0.0007,
      "reward": 5.159846603870392,
      "reward_std": 0.6249177735298872,
      "rewards/ngram_similarity_reward_func": 0.8409092575311661,
      "rewards/reasoning_quality_reward_func": 0.80018749833107,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.768750011920929,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 107
    },
    {
      "completion_length": 369.90625,
      "epoch": 0.05950413223140496,
      "grad_norm": 0.2089715600013733,
      "kl": 0.009871491696685553,
      "learning_rate": 3.4240000000000004e-05,
      "loss": 0.0004,
      "reward": 5.095018148422241,
      "reward_std": 0.8243512772023678,
      "rewards/ngram_similarity_reward_func": 0.6480785310268402,
      "rewards/reasoning_quality_reward_func": 0.8417312353849411,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.8552083373069763,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 108
    },
    {
      "completion_length": 342.09375,
      "epoch": 0.06005509641873278,
      "grad_norm": 0.19979025423526764,
      "kl": 0.011369143379852176,
      "learning_rate": 3.456e-05,
      "loss": 0.0005,
      "reward": 4.345081031322479,
      "reward_std": 0.897840328514576,
      "rewards/ngram_similarity_reward_func": 0.6619933396577835,
      "rewards/reasoning_quality_reward_func": 0.842462494969368,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.090625012293458,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 109
    },
    {
      "completion_length": 419.0625,
      "epoch": 0.06060606060606061,
      "grad_norm": 0.18090598285198212,
      "kl": 0.008968624519184232,
      "learning_rate": 3.4880000000000005e-05,
      "loss": 0.0004,
      "reward": 3.8538827002048492,
      "reward_std": 0.5639965068548918,
      "rewards/ngram_similarity_reward_func": 0.4821842089295387,
      "rewards/reasoning_quality_reward_func": 0.8049687594175339,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 1.8167298324406147,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 110
    },
    {
      "completion_length": 303.4375,
      "epoch": 0.06115702479338843,
      "grad_norm": 0.20876088738441467,
      "kl": 0.010701463092118502,
      "learning_rate": 3.52e-05,
      "loss": 0.0004,
      "reward": 6.340784788131714,
      "reward_std": 0.3915379513055086,
      "rewards/ngram_similarity_reward_func": 0.8014628738164902,
      "rewards/reasoning_quality_reward_func": 0.8518218845129013,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.9375,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 111
    },
    {
      "completion_length": 363.9375,
      "epoch": 0.06170798898071626,
      "grad_norm": 0.1850619614124298,
      "kl": 0.010584103525616229,
      "learning_rate": 3.5520000000000006e-05,
      "loss": 0.0004,
      "reward": 4.688681542873383,
      "reward_std": 0.8553222576156259,
      "rewards/ngram_similarity_reward_func": 0.36369094252586365,
      "rewards/reasoning_quality_reward_func": 0.7843656539916992,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.790624976158142,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 112
    },
    {
      "completion_length": 344.625,
      "epoch": 0.06225895316804408,
      "grad_norm": 0.23385602235794067,
      "kl": 0.020388634875416756,
      "learning_rate": 3.584e-05,
      "loss": 0.0008,
      "reward": 5.20521080493927,
      "reward_std": 0.6676309928297997,
      "rewards/ngram_similarity_reward_func": 0.7626158371567726,
      "rewards/reasoning_quality_reward_func": 0.8321781158447266,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.8604167103767395,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 113
    },
    {
      "completion_length": 351.375,
      "epoch": 0.0628099173553719,
      "grad_norm": 0.2119172066450119,
      "kl": 0.008771568071097136,
      "learning_rate": 3.6160000000000006e-05,
      "loss": 0.0004,
      "reward": 5.2659080028533936,
      "reward_std": 0.6435936316847801,
      "rewards/ngram_similarity_reward_func": 0.8063488081097603,
      "rewards/reasoning_quality_reward_func": 0.7595593929290771,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.949999988079071,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 114
    },
    {
      "completion_length": 307.28125,
      "epoch": 0.06336088154269973,
      "grad_norm": 0.20110149681568146,
      "kl": 0.00982317328453064,
      "learning_rate": 3.648e-05,
      "loss": 0.0004,
      "reward": 5.24526846408844,
      "reward_std": 0.5127543620765209,
      "rewards/ngram_similarity_reward_func": 0.5700779855251312,
      "rewards/reasoning_quality_reward_func": 0.8001906126737595,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.125,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 115
    },
    {
      "completion_length": 301.59375,
      "epoch": 0.06391184573002755,
      "grad_norm": 0.1948806792497635,
      "kl": 0.009930518572218716,
      "learning_rate": 3.680000000000001e-05,
      "loss": 0.0004,
      "reward": 6.027021288871765,
      "reward_std": 0.5728574581444263,
      "rewards/ngram_similarity_reward_func": 0.7108163684606552,
      "rewards/reasoning_quality_reward_func": 0.7347499877214432,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.8314549326896667,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 116
    },
    {
      "completion_length": 317.65625,
      "epoch": 0.06446280991735537,
      "grad_norm": 0.2220500111579895,
      "kl": 0.00799436867237091,
      "learning_rate": 3.7120000000000004e-05,
      "loss": 0.0003,
      "reward": 4.7908130288124084,
      "reward_std": 0.7186720389872789,
      "rewards/ngram_similarity_reward_func": 0.5876818932592869,
      "rewards/reasoning_quality_reward_func": 0.6968812718987465,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.7562499940395355,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 117
    },
    {
      "completion_length": 355.15625,
      "epoch": 0.06501377410468319,
      "grad_norm": 0.1883288323879242,
      "kl": 0.008259537047706544,
      "learning_rate": 3.744000000000001e-05,
      "loss": 0.0003,
      "reward": 4.540089786052704,
      "reward_std": 0.33911748975515366,
      "rewards/ngram_similarity_reward_func": 0.5239054560661316,
      "rewards/reasoning_quality_reward_func": 0.8036843687295914,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.462500000372529,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 118
    },
    {
      "completion_length": 325.71875,
      "epoch": 0.06556473829201102,
      "grad_norm": 0.22224414348602295,
      "kl": 0.007362338015809655,
      "learning_rate": 3.7760000000000004e-05,
      "loss": 0.0003,
      "reward": 5.972190499305725,
      "reward_std": 0.45987878926098347,
      "rewards/ngram_similarity_reward_func": 0.4636782184243202,
      "rewards/reasoning_quality_reward_func": 0.8796062618494034,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 3.890625,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 119
    },
    {
      "completion_length": 328.03125,
      "epoch": 0.06611570247933884,
      "grad_norm": 0.19121846556663513,
      "kl": 0.007113615749403834,
      "learning_rate": 3.808e-05,
      "loss": 0.0003,
      "reward": 5.125598132610321,
      "reward_std": 0.2027149423956871,
      "rewards/ngram_similarity_reward_func": 0.665495190769434,
      "rewards/reasoning_quality_reward_func": 0.7663531303405762,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.9437499940395355,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 120
    },
    {
      "completion_length": 262.46875,
      "epoch": 0.06666666666666667,
      "grad_norm": 0.2597827613353729,
      "kl": 0.007855100790038705,
      "learning_rate": 3.8400000000000005e-05,
      "loss": 0.0003,
      "reward": 4.136849880218506,
      "reward_std": 1.1940234303474426,
      "rewards/ngram_similarity_reward_func": 0.6112998872995377,
      "rewards/reasoning_quality_reward_func": 0.6818000078201294,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.093750024214387,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 121
    },
    {
      "completion_length": 319.4375,
      "epoch": 0.06721763085399449,
      "grad_norm": 0.24593015015125275,
      "kl": 0.007774659083224833,
      "learning_rate": 3.872e-05,
      "loss": 0.0003,
      "reward": 4.818424999713898,
      "reward_std": 0.49897436052560806,
      "rewards/ngram_similarity_reward_func": 0.5515125542879105,
      "rewards/reasoning_quality_reward_func": 0.7544124871492386,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.7625000178813934,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 122
    },
    {
      "completion_length": 381.8125,
      "epoch": 0.06776859504132231,
      "grad_norm": 0.16568975150585175,
      "kl": 0.006753749563358724,
      "learning_rate": 3.9040000000000006e-05,
      "loss": 0.0003,
      "reward": 4.689805388450623,
      "reward_std": 0.40961530432105064,
      "rewards/ngram_similarity_reward_func": 0.6910612061619759,
      "rewards/reasoning_quality_reward_func": 0.7318093627691269,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.5169349908828735,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 123
    },
    {
      "completion_length": 266.3125,
      "epoch": 0.06831955922865014,
      "grad_norm": 0.23362872004508972,
      "kl": 0.010426584631204605,
      "learning_rate": 3.936e-05,
      "loss": 0.0004,
      "reward": 5.068384110927582,
      "reward_std": 0.8161365427076817,
      "rewards/ngram_similarity_reward_func": 0.7751953825354576,
      "rewards/reasoning_quality_reward_func": 0.7123250067234039,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.850394994020462,
      "rewards/xmlcount_reward_func": 0.48828125,
      "step": 124
    },
    {
      "completion_length": 361.625,
      "epoch": 0.06887052341597796,
      "grad_norm": 0.19275341928005219,
      "kl": 0.008877517189830542,
      "learning_rate": 3.9680000000000006e-05,
      "loss": 0.0004,
      "reward": 5.248947739601135,
      "reward_std": 0.8990278542041779,
      "rewards/ngram_similarity_reward_func": 0.646971270442009,
      "rewards/reasoning_quality_reward_func": 0.8410031348466873,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.010973423719406,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 125
    },
    {
      "completion_length": 329.75,
      "epoch": 0.06942148760330578,
      "grad_norm": 0.20203961431980133,
      "kl": 0.008792976499535143,
      "learning_rate": 4e-05,
      "loss": 0.0004,
      "reward": 5.415657877922058,
      "reward_std": 0.6656006760895252,
      "rewards/ngram_similarity_reward_func": 0.7358945608139038,
      "rewards/reasoning_quality_reward_func": 0.8827375024557114,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 3.062650978565216,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 126
    },
    {
      "completion_length": 318.15625,
      "epoch": 0.0699724517906336,
      "grad_norm": 0.1906377226114273,
      "kl": 0.01101280003786087,
      "learning_rate": 3.999998425237866e-05,
      "loss": 0.0004,
      "reward": 4.9277477860450745,
      "reward_std": 0.5749133937060833,
      "rewards/ngram_similarity_reward_func": 0.5787328854203224,
      "rewards/reasoning_quality_reward_func": 0.8332812488079071,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.765733480453491,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 127
    },
    {
      "completion_length": 329.8125,
      "epoch": 0.07052341597796143,
      "grad_norm": 0.19732476770877838,
      "kl": 0.009841614286415279,
      "learning_rate": 3.999993700954217e-05,
      "loss": 0.0004,
      "reward": 3.5110082626342773,
      "reward_std": 0.3044098876416683,
      "rewards/ngram_similarity_reward_func": 0.45819390565156937,
      "rewards/reasoning_quality_reward_func": 0.7409937530755997,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 1.5774456206709146,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 128
    },
    {
      "completion_length": 288.0,
      "epoch": 0.07107438016528926,
      "grad_norm": 0.22847525775432587,
      "kl": 0.012037057429552078,
      "learning_rate": 3.999985827157321e-05,
      "loss": 0.0005,
      "reward": 5.634380459785461,
      "reward_std": 0.9955153241753578,
      "rewards/ngram_similarity_reward_func": 0.5509617477655411,
      "rewards/reasoning_quality_reward_func": 0.8290437757968903,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.5043750405311584,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 129
    },
    {
      "completion_length": 320.59375,
      "epoch": 0.07162534435261708,
      "grad_norm": 0.2519027888774872,
      "kl": 0.013253344572149217,
      "learning_rate": 3.999974803860954e-05,
      "loss": 0.0005,
      "reward": 4.560444593429565,
      "reward_std": 0.585046473890543,
      "rewards/ngram_similarity_reward_func": 0.5937070995569229,
      "rewards/reasoning_quality_reward_func": 0.8417375087738037,
      "rewards/soft_format_reward_func": 0.234375,
      "rewards/sql_execution_reward_func": 2.4062499944120646,
      "rewards/xmlcount_reward_func": 0.484375,
      "step": 130
    },
    {
      "completion_length": 378.59375,
      "epoch": 0.0721763085399449,
      "grad_norm": 0.19772933423519135,
      "kl": 0.01022614783141762,
      "learning_rate": 3.999960631084405e-05,
      "loss": 0.0004,
      "reward": 4.7231597900390625,
      "reward_std": 1.2054128795862198,
      "rewards/ngram_similarity_reward_func": 0.5730114951729774,
      "rewards/reasoning_quality_reward_func": 0.8266499936580658,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.589123249053955,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 131
    },
    {
      "completion_length": 357.6875,
      "epoch": 0.07272727272727272,
      "grad_norm": 0.18746080994606018,
      "kl": 0.011636733775958419,
      "learning_rate": 3.999943308852471e-05,
      "loss": 0.0005,
      "reward": 5.560112535953522,
      "reward_std": 0.4500330928713083,
      "rewards/ngram_similarity_reward_func": 0.6128779947757721,
      "rewards/reasoning_quality_reward_func": 0.8150843679904938,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.3821502327919006,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 132
    },
    {
      "completion_length": 380.875,
      "epoch": 0.07327823691460054,
      "grad_norm": 0.2277393341064453,
      "kl": 0.011123559204861522,
      "learning_rate": 3.999922837195462e-05,
      "loss": 0.0004,
      "reward": 4.1726327538490295,
      "reward_std": 1.1211602091789246,
      "rewards/ngram_similarity_reward_func": 0.5038543120026588,
      "rewards/reasoning_quality_reward_func": 0.8145249783992767,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.1159722805023193,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 133
    },
    {
      "completion_length": 348.28125,
      "epoch": 0.07382920110192838,
      "grad_norm": 0.22234097123146057,
      "kl": 0.013425002107396722,
      "learning_rate": 3.9998992161491994e-05,
      "loss": 0.0005,
      "reward": 5.440715551376343,
      "reward_std": 0.8107848465442657,
      "rewards/ngram_similarity_reward_func": 0.4736420288681984,
      "rewards/reasoning_quality_reward_func": 0.9014750123023987,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.3312233090400696,
      "rewards/xmlcount_reward_func": 0.484375,
      "step": 134
    },
    {
      "completion_length": 316.15625,
      "epoch": 0.0743801652892562,
      "grad_norm": 0.20705243945121765,
      "kl": 0.014697337057441473,
      "learning_rate": 3.9998724457550115e-05,
      "loss": 0.0006,
      "reward": 5.395981431007385,
      "reward_std": 0.4146622382104397,
      "rewards/ngram_similarity_reward_func": 0.4082877188920975,
      "rewards/reasoning_quality_reward_func": 0.8626937419176102,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.375,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 135
    },
    {
      "completion_length": 340.75,
      "epoch": 0.07493112947658402,
      "grad_norm": 0.20130571722984314,
      "kl": 0.014105458511039615,
      "learning_rate": 3.999842526059741e-05,
      "loss": 0.0006,
      "reward": 4.150662779808044,
      "reward_std": 0.9526075609028339,
      "rewards/ngram_similarity_reward_func": 0.3124133050441742,
      "rewards/reasoning_quality_reward_func": 0.7001812309026718,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.4115056842565536,
      "rewards/xmlcount_reward_func": 0.484375,
      "step": 136
    },
    {
      "completion_length": 395.375,
      "epoch": 0.07548209366391184,
      "grad_norm": 0.18294228613376617,
      "kl": 0.012597255408763885,
      "learning_rate": 3.9998094571157396e-05,
      "loss": 0.0005,
      "reward": 4.450990557670593,
      "reward_std": 1.3137284964323044,
      "rewards/ngram_similarity_reward_func": 0.4396781399846077,
      "rewards/reasoning_quality_reward_func": 0.8253750205039978,
      "rewards/soft_format_reward_func": 0.2265625,
      "rewards/sql_execution_reward_func": 2.474999964237213,
      "rewards/xmlcount_reward_func": 0.484375,
      "step": 137
    },
    {
      "completion_length": 333.84375,
      "epoch": 0.07603305785123966,
      "grad_norm": 0.23128224909305573,
      "kl": 0.012666527414694428,
      "learning_rate": 3.999773238980868e-05,
      "loss": 0.0005,
      "reward": 4.982425928115845,
      "reward_std": 0.6804494932293892,
      "rewards/ngram_similarity_reward_func": 0.5416133664548397,
      "rewards/reasoning_quality_reward_func": 0.8408125042915344,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.8656249940395355,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 138
    },
    {
      "completion_length": 330.5625,
      "epoch": 0.0765840220385675,
      "grad_norm": 0.22288239002227783,
      "kl": 0.01523360749706626,
      "learning_rate": 3.9997338717184987e-05,
      "loss": 0.0006,
      "reward": 5.102320075035095,
      "reward_std": 0.1288034450262785,
      "rewards/ngram_similarity_reward_func": 0.5263919830322266,
      "rewards/reasoning_quality_reward_func": 0.8259281069040298,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.0,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 139
    },
    {
      "completion_length": 304.03125,
      "epoch": 0.07713498622589532,
      "grad_norm": 0.24967846274375916,
      "kl": 0.01684622629545629,
      "learning_rate": 3.9996913553975145e-05,
      "loss": 0.0007,
      "reward": 4.576799035072327,
      "reward_std": 1.1620003134012222,
      "rewards/ngram_similarity_reward_func": 0.6101422533392906,
      "rewards/reasoning_quality_reward_func": 0.8382406085729599,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.394041359424591,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 140
    },
    {
      "completion_length": 315.125,
      "epoch": 0.07768595041322314,
      "grad_norm": 0.22092896699905396,
      "kl": 0.015367173589766026,
      "learning_rate": 3.9996456900923076e-05,
      "loss": 0.0006,
      "reward": 4.989863395690918,
      "reward_std": 0.23215042892843485,
      "rewards/ngram_similarity_reward_func": 0.5795633494853973,
      "rewards/reasoning_quality_reward_func": 0.7165500074625015,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.9437499940395355,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 141
    },
    {
      "completion_length": 271.1875,
      "epoch": 0.07823691460055096,
      "grad_norm": 0.24714848399162292,
      "kl": 0.02564011886715889,
      "learning_rate": 3.9995968758827795e-05,
      "loss": 0.001,
      "reward": 5.3810752630233765,
      "reward_std": 0.6415481753647327,
      "rewards/ngram_similarity_reward_func": 0.5185722336173058,
      "rewards/reasoning_quality_reward_func": 0.8093781471252441,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 3.318750023841858,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 142
    },
    {
      "completion_length": 323.53125,
      "epoch": 0.07878787878787878,
      "grad_norm": 0.22965122759342194,
      "kl": 0.022572920424863696,
      "learning_rate": 3.999544912854343e-05,
      "loss": 0.0009,
      "reward": 5.100914001464844,
      "reward_std": 0.9806921817362309,
      "rewards/ngram_similarity_reward_func": 0.5324078314006329,
      "rewards/reasoning_quality_reward_func": 0.766918733716011,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 3.0750250816345215,
      "rewards/xmlcount_reward_func": 0.484375,
      "step": 143
    },
    {
      "completion_length": 304.75,
      "epoch": 0.07933884297520662,
      "grad_norm": 0.23732416331768036,
      "kl": 0.015826057642698288,
      "learning_rate": 3.99948980109792e-05,
      "loss": 0.0006,
      "reward": 5.519599914550781,
      "reward_std": 0.4865953903645277,
      "rewards/ngram_similarity_reward_func": 0.6586082726716995,
      "rewards/reasoning_quality_reward_func": 0.8387843668460846,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.272207498550415,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 144
    },
    {
      "completion_length": 322.6875,
      "epoch": 0.07988980716253444,
      "grad_norm": 0.2301616668701172,
      "kl": 0.015833008335903287,
      "learning_rate": 3.99943154070994e-05,
      "loss": 0.0006,
      "reward": 5.664751410484314,
      "reward_std": 0.37940163537859917,
      "rewards/ngram_similarity_reward_func": 0.7232076525688171,
      "rewards/reasoning_quality_reward_func": 0.7540437281131744,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.4375,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 145
    },
    {
      "completion_length": 328.375,
      "epoch": 0.08044077134986226,
      "grad_norm": 0.20747892558574677,
      "kl": 0.019038383848965168,
      "learning_rate": 3.9993701317923455e-05,
      "loss": 0.0008,
      "reward": 6.029369235038757,
      "reward_std": 0.48181281238794327,
      "rewards/ngram_similarity_reward_func": 0.8008849322795868,
      "rewards/reasoning_quality_reward_func": 0.7909843921661377,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.6875,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 146
    },
    {
      "completion_length": 264.4375,
      "epoch": 0.08099173553719008,
      "grad_norm": 0.2791043817996979,
      "kl": 0.023179364623501897,
      "learning_rate": 3.999305574452583e-05,
      "loss": 0.0009,
      "reward": 5.317837357521057,
      "reward_std": 1.332659363746643,
      "rewards/ngram_similarity_reward_func": 0.7005831003189087,
      "rewards/reasoning_quality_reward_func": 0.7726156190037727,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.094638705253601,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 147
    },
    {
      "completion_length": 284.34375,
      "epoch": 0.0815426997245179,
      "grad_norm": 0.2564888894557953,
      "kl": 0.022093956358730793,
      "learning_rate": 3.9992378688036135e-05,
      "loss": 0.0009,
      "reward": 6.038819074630737,
      "reward_std": 0.5817711669951677,
      "rewards/ngram_similarity_reward_func": 0.7943221479654312,
      "rewards/reasoning_quality_reward_func": 0.8069968670606613,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.6875,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 148
    },
    {
      "completion_length": 329.0625,
      "epoch": 0.08209366391184574,
      "grad_norm": 0.27149972319602966,
      "kl": 0.013890980510041118,
      "learning_rate": 3.9991670149639026e-05,
      "loss": 0.0006,
      "reward": 4.029133200645447,
      "reward_std": 0.35773539170622826,
      "rewards/ngram_similarity_reward_func": 0.4883638247847557,
      "rewards/reasoning_quality_reward_func": 0.7056999802589417,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.085069417953491,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 149
    },
    {
      "completion_length": 315.8125,
      "epoch": 0.08264462809917356,
      "grad_norm": 0.22877110540866852,
      "kl": 0.017250674311071634,
      "learning_rate": 3.999093013057426e-05,
      "loss": 0.0007,
      "reward": 5.506930649280548,
      "reward_std": 0.9728036783635616,
      "rewards/ngram_similarity_reward_func": 0.631624698638916,
      "rewards/reasoning_quality_reward_func": 0.8579031229019165,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 3.279121458530426,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 150
    },
    {
      "completion_length": 309.84375,
      "epoch": 0.08319559228650138,
      "grad_norm": 0.238431915640831,
      "kl": 0.02233697008341551,
      "learning_rate": 3.999015863213667e-05,
      "loss": 0.0009,
      "reward": 4.859403848648071,
      "reward_std": 0.438848614692688,
      "rewards/ngram_similarity_reward_func": 0.5722757056355476,
      "rewards/reasoning_quality_reward_func": 0.8621281236410141,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.675000060349703,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 151
    },
    {
      "completion_length": 290.5,
      "epoch": 0.0837465564738292,
      "grad_norm": 0.2296566665172577,
      "kl": 0.020341533236205578,
      "learning_rate": 3.998935565567618e-05,
      "loss": 0.0008,
      "reward": 5.661309719085693,
      "reward_std": 0.30164369381964207,
      "rewards/ngram_similarity_reward_func": 0.45216045528650284,
      "rewards/reasoning_quality_reward_func": 0.8523968905210495,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 3.6184712648391724,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 152
    },
    {
      "completion_length": 386.59375,
      "epoch": 0.08429752066115702,
      "grad_norm": 0.20847496390342712,
      "kl": 0.01867723953910172,
      "learning_rate": 3.9988521202597795e-05,
      "loss": 0.0007,
      "reward": 4.674509406089783,
      "reward_std": 0.6103712171316147,
      "rewards/ngram_similarity_reward_func": 0.5504591688513756,
      "rewards/reasoning_quality_reward_func": 0.8562499582767487,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.5256127268075943,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 153
    },
    {
      "completion_length": 334.0,
      "epoch": 0.08484848484848485,
      "grad_norm": 0.2299894243478775,
      "kl": 0.018067997880280018,
      "learning_rate": 3.998765527436158e-05,
      "loss": 0.0007,
      "reward": 4.06777960062027,
      "reward_std": 0.5151195954531431,
      "rewards/ngram_similarity_reward_func": 0.5639265179634094,
      "rewards/reasoning_quality_reward_func": 0.8226031064987183,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 1.931250024586916,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 154
    },
    {
      "completion_length": 289.3125,
      "epoch": 0.08539944903581267,
      "grad_norm": 0.24813717603683472,
      "kl": 0.02420641388744116,
      "learning_rate": 3.9986757872482666e-05,
      "loss": 0.001,
      "reward": 5.503636837005615,
      "reward_std": 0.9658561106771231,
      "rewards/ngram_similarity_reward_func": 0.6948243826627731,
      "rewards/reasoning_quality_reward_func": 0.677562490105629,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.381250023841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 155
    },
    {
      "completion_length": 360.59375,
      "epoch": 0.0859504132231405,
      "grad_norm": 0.2219952791929245,
      "kl": 0.020170791540294886,
      "learning_rate": 3.99858289985313e-05,
      "loss": 0.0008,
      "reward": 5.577654242515564,
      "reward_std": 0.561808068305254,
      "rewards/ngram_similarity_reward_func": 0.6386853456497192,
      "rewards/reasoning_quality_reward_func": 0.8014687597751617,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.387499988079071,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 156
    },
    {
      "completion_length": 343.4375,
      "epoch": 0.08650137741046832,
      "grad_norm": 0.2123832255601883,
      "kl": 0.016611569561064243,
      "learning_rate": 3.998486865413273e-05,
      "loss": 0.0007,
      "reward": 4.503101110458374,
      "reward_std": 0.3235671203583479,
      "rewards/ngram_similarity_reward_func": 0.32479164749383926,
      "rewards/reasoning_quality_reward_func": 0.7783093750476837,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.650000000372529,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 157
    },
    {
      "completion_length": 335.875,
      "epoch": 0.08705234159779614,
      "grad_norm": 0.2268906980752945,
      "kl": 0.021912760101258755,
      "learning_rate": 3.9983876840967334e-05,
      "loss": 0.0009,
      "reward": 4.452318429946899,
      "reward_std": 0.15844441950321198,
      "rewards/ngram_similarity_reward_func": 0.741477757692337,
      "rewards/reasoning_quality_reward_func": 0.9108406156301498,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.050000000745058,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 158
    },
    {
      "completion_length": 387.4375,
      "epoch": 0.08760330578512397,
      "grad_norm": 0.2001405656337738,
      "kl": 0.01731085404753685,
      "learning_rate": 3.99828535607705e-05,
      "loss": 0.0007,
      "reward": 4.3963674902915955,
      "reward_std": 0.8671909868717194,
      "rewards/ngram_similarity_reward_func": 0.4167655184864998,
      "rewards/reasoning_quality_reward_func": 0.8365906178951263,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.3930112421512604,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 159
    },
    {
      "completion_length": 333.25,
      "epoch": 0.0881542699724518,
      "grad_norm": 0.21372917294502258,
      "kl": 0.02031551953405142,
      "learning_rate": 3.998179881533272e-05,
      "loss": 0.0008,
      "reward": 5.088305115699768,
      "reward_std": 0.48543216846883297,
      "rewards/ngram_similarity_reward_func": 0.6934583783149719,
      "rewards/reasoning_quality_reward_func": 0.7854719012975693,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.859375024214387,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 160
    },
    {
      "completion_length": 347.5,
      "epoch": 0.08870523415977961,
      "grad_norm": 0.21207331120967865,
      "kl": 0.01876061293296516,
      "learning_rate": 3.998071260649951e-05,
      "loss": 0.0008,
      "reward": 5.182012319564819,
      "reward_std": 0.8623120933771133,
      "rewards/ngram_similarity_reward_func": 0.7045288980007172,
      "rewards/reasoning_quality_reward_func": 0.8479843586683273,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.879499137401581,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 161
    },
    {
      "completion_length": 396.21875,
      "epoch": 0.08925619834710743,
      "grad_norm": 0.19062969088554382,
      "kl": 0.019745307508856058,
      "learning_rate": 3.997959493617145e-05,
      "loss": 0.0008,
      "reward": 4.953526854515076,
      "reward_std": 1.0659264251589775,
      "rewards/ngram_similarity_reward_func": 0.5991052836179733,
      "rewards/reasoning_quality_reward_func": 0.9544218629598618,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.650000035762787,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 162
    },
    {
      "completion_length": 430.03125,
      "epoch": 0.08980716253443526,
      "grad_norm": 0.18317510187625885,
      "kl": 0.018107001669704914,
      "learning_rate": 3.9978445806304154e-05,
      "loss": 0.0007,
      "reward": 4.987079620361328,
      "reward_std": 1.479330651462078,
      "rewards/ngram_similarity_reward_func": 0.5653297081589699,
      "rewards/reasoning_quality_reward_func": 0.8459687680006027,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.837499976158142,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 163
    },
    {
      "completion_length": 364.21875,
      "epoch": 0.09035812672176309,
      "grad_norm": 0.18856443464756012,
      "kl": 0.018292442662641406,
      "learning_rate": 3.997726521890832e-05,
      "loss": 0.0007,
      "reward": 5.456216335296631,
      "reward_std": 0.4839339912869036,
      "rewards/ngram_similarity_reward_func": 0.549975611269474,
      "rewards/reasoning_quality_reward_func": 0.7906156331300735,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.365625023841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 164
    },
    {
      "completion_length": 263.0625,
      "epoch": 0.09090909090909091,
      "grad_norm": 0.235592320561409,
      "kl": 0.026188641786575317,
      "learning_rate": 3.997605317604964e-05,
      "loss": 0.001,
      "reward": 5.203069448471069,
      "reward_std": 0.21571800857782364,
      "rewards/ngram_similarity_reward_func": 0.6629913374781609,
      "rewards/reasoning_quality_reward_func": 0.7900780886411667,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.0,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 165
    },
    {
      "completion_length": 364.8125,
      "epoch": 0.09146005509641873,
      "grad_norm": 0.20042334496974945,
      "kl": 0.019627680303528905,
      "learning_rate": 3.997480967984889e-05,
      "loss": 0.0008,
      "reward": 5.227686882019043,
      "reward_std": 0.5553257688879967,
      "rewards/ngram_similarity_reward_func": 0.7056699246168137,
      "rewards/reasoning_quality_reward_func": 0.7935687005519867,
      "rewards/soft_format_reward_func": 0.234375,
      "rewards/sql_execution_reward_func": 3.001885771751404,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 166
    },
    {
      "completion_length": 384.6875,
      "epoch": 0.09201101928374655,
      "grad_norm": 0.1835215836763382,
      "kl": 0.018579522613435984,
      "learning_rate": 3.9973534732481844e-05,
      "loss": 0.0007,
      "reward": 4.3468785881996155,
      "reward_std": 1.2626140713691711,
      "rewards/ngram_similarity_reward_func": 0.6147419437766075,
      "rewards/reasoning_quality_reward_func": 0.7123031318187714,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.269833564758301,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 167
    },
    {
      "completion_length": 413.25,
      "epoch": 0.09256198347107437,
      "grad_norm": 0.22163331508636475,
      "kl": 0.020407014526426792,
      "learning_rate": 3.9972228336179326e-05,
      "loss": 0.0008,
      "reward": 4.436458110809326,
      "reward_std": 0.6687717800959945,
      "rewards/ngram_similarity_reward_func": 0.5802749246358871,
      "rewards/reasoning_quality_reward_func": 0.9345687329769135,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.1872394923120737,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 168
    },
    {
      "completion_length": 298.59375,
      "epoch": 0.09311294765840221,
      "grad_norm": 0.22880753874778748,
      "kl": 0.028387621510773897,
      "learning_rate": 3.99708904932272e-05,
      "loss": 0.0011,
      "reward": 4.087678909301758,
      "reward_std": 0.45572323724627495,
      "rewards/ngram_similarity_reward_func": 0.8158641308546066,
      "rewards/reasoning_quality_reward_func": 0.8272124975919724,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 1.694602245464921,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 169
    },
    {
      "completion_length": 399.46875,
      "epoch": 0.09366391184573003,
      "grad_norm": 0.19029830396175385,
      "kl": 0.01864568702876568,
      "learning_rate": 3.996952120596631e-05,
      "loss": 0.0007,
      "reward": 4.570008218288422,
      "reward_std": 1.062185663729906,
      "rewards/ngram_similarity_reward_func": 0.5074205696582794,
      "rewards/reasoning_quality_reward_func": 0.8485249876976013,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.464062422513962,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 170
    },
    {
      "completion_length": 337.8125,
      "epoch": 0.09421487603305785,
      "grad_norm": 0.20451989769935608,
      "kl": 0.021481969393789768,
      "learning_rate": 3.9968120476792565e-05,
      "loss": 0.0009,
      "reward": 5.498373866081238,
      "reward_std": 0.619726475328207,
      "rewards/ngram_similarity_reward_func": 0.5506386682391167,
      "rewards/reasoning_quality_reward_func": 0.7498187571763992,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.447916626930237,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 171
    },
    {
      "completion_length": 374.09375,
      "epoch": 0.09476584022038567,
      "grad_norm": 0.18137097358703613,
      "kl": 0.023001483641564846,
      "learning_rate": 3.9966688308156866e-05,
      "loss": 0.0009,
      "reward": 5.394274592399597,
      "reward_std": 1.0623726639896631,
      "rewards/ngram_similarity_reward_func": 0.7177901491522789,
      "rewards/reasoning_quality_reward_func": 0.8952343761920929,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.03125,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 172
    },
    {
      "completion_length": 478.6875,
      "epoch": 0.09531680440771349,
      "grad_norm": 0.16733182966709137,
      "kl": 0.020202928688377142,
      "learning_rate": 3.9965224702565136e-05,
      "loss": 0.0008,
      "reward": 4.732444882392883,
      "reward_std": 0.6037857383489609,
      "rewards/ngram_similarity_reward_func": 0.5060711354017258,
      "rewards/reasoning_quality_reward_func": 0.8433843404054642,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.632989525794983,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 173
    },
    {
      "completion_length": 411.6875,
      "epoch": 0.09586776859504133,
      "grad_norm": 0.1691085547208786,
      "kl": 0.018214168259873986,
      "learning_rate": 3.9963729662578286e-05,
      "loss": 0.0007,
      "reward": 4.569848656654358,
      "reward_std": 0.42689267359673977,
      "rewards/ngram_similarity_reward_func": 0.516717467457056,
      "rewards/reasoning_quality_reward_func": 0.9218812584877014,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.381250023841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 174
    },
    {
      "completion_length": 409.125,
      "epoch": 0.09641873278236915,
      "grad_norm": 0.19372420012950897,
      "kl": 0.018990315962582827,
      "learning_rate": 3.996220319081225e-05,
      "loss": 0.0008,
      "reward": 4.731012344360352,
      "reward_std": 0.6671605389565229,
      "rewards/ngram_similarity_reward_func": 0.5885710120201111,
      "rewards/reasoning_quality_reward_func": 0.8485312312841415,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.543910026550293,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 175
    },
    {
      "completion_length": 432.9375,
      "epoch": 0.09696969696969697,
      "grad_norm": 0.1991139054298401,
      "kl": 0.019880631007254124,
      "learning_rate": 3.996064528993794e-05,
      "loss": 0.0008,
      "reward": 4.397683918476105,
      "reward_std": 0.34171948954463005,
      "rewards/ngram_similarity_reward_func": 0.4442303776741028,
      "rewards/reasoning_quality_reward_func": 0.9097374677658081,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.2937160432338715,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 176
    },
    {
      "completion_length": 383.21875,
      "epoch": 0.09752066115702479,
      "grad_norm": 0.18429315090179443,
      "kl": 0.01815024996176362,
      "learning_rate": 3.995905596268129e-05,
      "loss": 0.0007,
      "reward": 5.426558613777161,
      "reward_std": 0.8819810226559639,
      "rewards/ngram_similarity_reward_func": 0.5017272680997849,
      "rewards/reasoning_quality_reward_func": 0.9060812294483185,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.268750011920929,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 177
    },
    {
      "completion_length": 353.3125,
      "epoch": 0.09807162534435261,
      "grad_norm": 0.20830138027668,
      "kl": 0.01908263936638832,
      "learning_rate": 3.995743521182319e-05,
      "loss": 0.0008,
      "reward": 4.340333938598633,
      "reward_std": 1.1430581733584404,
      "rewards/ngram_similarity_reward_func": 0.6442931517958641,
      "rewards/reasoning_quality_reward_func": 0.7766593545675278,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.169381409883499,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 178
    },
    {
      "completion_length": 343.46875,
      "epoch": 0.09862258953168045,
      "grad_norm": 0.21066419780254364,
      "kl": 0.01992220152169466,
      "learning_rate": 3.995578304019952e-05,
      "loss": 0.0008,
      "reward": 6.076923251152039,
      "reward_std": 0.4823677083477378,
      "rewards/ngram_similarity_reward_func": 0.7457924634218216,
      "rewards/reasoning_quality_reward_func": 0.8663656115531921,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.7147651314735413,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 179
    },
    {
      "completion_length": 325.90625,
      "epoch": 0.09917355371900827,
      "grad_norm": 0.19968515634536743,
      "kl": 0.020209768321365118,
      "learning_rate": 3.995409945070115e-05,
      "loss": 0.0008,
      "reward": 5.791889667510986,
      "reward_std": 0.6311796046793461,
      "rewards/ngram_similarity_reward_func": 0.6983146965503693,
      "rewards/reasoning_quality_reward_func": 0.8998249769210815,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.443750023841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 180
    },
    {
      "completion_length": 416.625,
      "epoch": 0.09972451790633609,
      "grad_norm": 0.1794973462820053,
      "kl": 0.022402073722332716,
      "learning_rate": 3.995238444627392e-05,
      "loss": 0.0009,
      "reward": 5.705436944961548,
      "reward_std": 1.1184853315353394,
      "rewards/ngram_similarity_reward_func": 0.696608766913414,
      "rewards/reasoning_quality_reward_func": 0.9400781244039536,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.318750023841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 181
    },
    {
      "completion_length": 342.90625,
      "epoch": 0.10027548209366391,
      "grad_norm": 0.23436102271080017,
      "kl": 0.02327228942885995,
      "learning_rate": 3.9950638029918634e-05,
      "loss": 0.0009,
      "reward": 5.666057586669922,
      "reward_std": 1.0826963391155005,
      "rewards/ngram_similarity_reward_func": 0.6460169702768326,
      "rewards/reasoning_quality_reward_func": 0.8794156312942505,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.390625,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 182
    },
    {
      "completion_length": 400.8125,
      "epoch": 0.10082644628099173,
      "grad_norm": 0.19203157722949982,
      "kl": 0.02062929724343121,
      "learning_rate": 3.994886020469105e-05,
      "loss": 0.0008,
      "reward": 6.246921420097351,
      "reward_std": 0.5135991498827934,
      "rewards/ngram_similarity_reward_func": 0.7910341769456863,
      "rewards/reasoning_quality_reward_func": 0.877762496471405,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.828125,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 183
    },
    {
      "completion_length": 353.5,
      "epoch": 0.10137741046831956,
      "grad_norm": 0.21837492287158966,
      "kl": 0.02511785039678216,
      "learning_rate": 3.9947050973701906e-05,
      "loss": 0.001,
      "reward": 4.722505569458008,
      "reward_std": 1.244483396410942,
      "rewards/ngram_similarity_reward_func": 0.556546151638031,
      "rewards/reasoning_quality_reward_func": 0.9590156376361847,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.456943780183792,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 184
    },
    {
      "completion_length": 318.5,
      "epoch": 0.10192837465564739,
      "grad_norm": 0.2269653081893921,
      "kl": 0.028365733567625284,
      "learning_rate": 3.994521034011687e-05,
      "loss": 0.0011,
      "reward": 5.827512860298157,
      "reward_std": 0.5985339358448982,
      "rewards/ngram_similarity_reward_func": 0.9015379846096039,
      "rewards/reasoning_quality_reward_func": 0.9220687299966812,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.25390625,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 185
    },
    {
      "completion_length": 393.5625,
      "epoch": 0.1024793388429752,
      "grad_norm": 0.18627303838729858,
      "kl": 0.022229368798434734,
      "learning_rate": 3.994333830715656e-05,
      "loss": 0.0009,
      "reward": 5.174306750297546,
      "reward_std": 0.12488455418497324,
      "rewards/ngram_similarity_reward_func": 0.544351615011692,
      "rewards/reasoning_quality_reward_func": 0.8740843832492828,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.005870819091797,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 186
    },
    {
      "completion_length": 330.0,
      "epoch": 0.10303030303030303,
      "grad_norm": 0.20830100774765015,
      "kl": 0.018538800533860922,
      "learning_rate": 3.994143487809655e-05,
      "loss": 0.0007,
      "reward": 4.05972746014595,
      "reward_std": 0.08067170064896345,
      "rewards/ngram_similarity_reward_func": 0.428286824375391,
      "rewards/reasoning_quality_reward_func": 0.8345656543970108,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.046875,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 187
    },
    {
      "completion_length": 297.25,
      "epoch": 0.10358126721763085,
      "grad_norm": 0.2860408425331116,
      "kl": 0.02084790775552392,
      "learning_rate": 3.993950005626733e-05,
      "loss": 0.0008,
      "reward": 5.500156283378601,
      "reward_std": 1.422216221690178,
      "rewards/ngram_similarity_reward_func": 0.6882125735282898,
      "rewards/reasoning_quality_reward_func": 0.8994437158107758,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.162500023841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 188
    },
    {
      "completion_length": 388.875,
      "epoch": 0.10413223140495868,
      "grad_norm": 0.19558590650558472,
      "kl": 0.01771295559592545,
      "learning_rate": 3.9937533845054324e-05,
      "loss": 0.0007,
      "reward": 4.036859631538391,
      "reward_std": 0.613134354352951,
      "rewards/ngram_similarity_reward_func": 0.4997222572565079,
      "rewards/reasoning_quality_reward_func": 0.8683875054121017,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 1.918750012293458,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 189
    },
    {
      "completion_length": 364.78125,
      "epoch": 0.1046831955922865,
      "grad_norm": 0.19008566439151764,
      "kl": 0.019377468386664987,
      "learning_rate": 3.993553624789788e-05,
      "loss": 0.0008,
      "reward": 5.542128086090088,
      "reward_std": 0.9688821043819189,
      "rewards/ngram_similarity_reward_func": 0.6578373312950134,
      "rewards/reasoning_quality_reward_func": 0.870228111743927,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.264062523841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 190
    },
    {
      "completion_length": 396.0625,
      "epoch": 0.10523415977961433,
      "grad_norm": 0.16607074439525604,
      "kl": 0.01804225263185799,
      "learning_rate": 3.993350726829329e-05,
      "loss": 0.0007,
      "reward": 5.158402681350708,
      "reward_std": 0.45148452930152416,
      "rewards/ngram_similarity_reward_func": 0.6183214709162712,
      "rewards/reasoning_quality_reward_func": 0.8900812566280365,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.900000000372529,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 191
    },
    {
      "completion_length": 316.59375,
      "epoch": 0.10578512396694215,
      "grad_norm": 0.256091833114624,
      "kl": 0.023836666252464056,
      "learning_rate": 3.9931446909790704e-05,
      "loss": 0.001,
      "reward": 4.693995952606201,
      "reward_std": 1.0115011408925056,
      "rewards/ngram_similarity_reward_func": 0.5454741790890694,
      "rewards/reasoning_quality_reward_func": 0.7672718614339828,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.6312499940395355,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 192
    },
    {
      "completion_length": 284.1875,
      "epoch": 0.10633608815426997,
      "grad_norm": 0.2528315484523773,
      "kl": 0.019845283590257168,
      "learning_rate": 3.992935517599521e-05,
      "loss": 0.0008,
      "reward": 4.947393178939819,
      "reward_std": 1.1741669736802578,
      "rewards/ngram_similarity_reward_func": 0.5156897157430649,
      "rewards/reasoning_quality_reward_func": 0.7132375091314316,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.9684659242630005,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 193
    },
    {
      "completion_length": 285.6875,
      "epoch": 0.1068870523415978,
      "grad_norm": 0.27771133184432983,
      "kl": 0.02378112031146884,
      "learning_rate": 3.9927232070566796e-05,
      "loss": 0.001,
      "reward": 5.101982116699219,
      "reward_std": 0.5751842856407166,
      "rewards/ngram_similarity_reward_func": 0.6297433972358704,
      "rewards/reasoning_quality_reward_func": 0.7566218674182892,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.9656169414520264,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 194
    },
    {
      "completion_length": 294.8125,
      "epoch": 0.10743801652892562,
      "grad_norm": 0.24460813403129578,
      "kl": 0.020680315559729934,
      "learning_rate": 3.992507759722034e-05,
      "loss": 0.0008,
      "reward": 5.134785413742065,
      "reward_std": 1.0555476024746895,
      "rewards/ngram_similarity_reward_func": 0.5507791340351105,
      "rewards/reasoning_quality_reward_func": 0.8152562528848648,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.018750011920929,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 195
    },
    {
      "completion_length": 703.5625,
      "epoch": 0.10798898071625344,
      "grad_norm": 2991890176.0,
      "kl": 0.020488561131060123,
      "learning_rate": 3.992289175972559e-05,
      "loss": 0.0008,
      "reward": 3.9725103974342346,
      "reward_std": 0.41914119105786085,
      "rewards/ngram_similarity_reward_func": 0.5232791267335415,
      "rewards/reasoning_quality_reward_func": 0.8187624961137772,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 1.900000000372529,
      "rewards/xmlcount_reward_func": 0.48828125,
      "step": 196
    },
    {
      "completion_length": 335.03125,
      "epoch": 0.10853994490358126,
      "grad_norm": 0.2154967337846756,
      "kl": 0.023860695539042354,
      "learning_rate": 3.992067456190719e-05,
      "loss": 0.001,
      "reward": 5.6779327392578125,
      "reward_std": 0.9344974812120199,
      "rewards/ngram_similarity_reward_func": 0.7124889642000198,
      "rewards/reasoning_quality_reward_func": 0.8748187273740768,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.340624988079071,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 197
    },
    {
      "completion_length": 258.84375,
      "epoch": 0.10909090909090909,
      "grad_norm": 0.2338087409734726,
      "kl": 0.024218559730798006,
      "learning_rate": 3.991842600764465e-05,
      "loss": 0.001,
      "reward": 4.8456215262413025,
      "reward_std": 0.7369251511991024,
      "rewards/ngram_similarity_reward_func": 0.7502120733261108,
      "rewards/reasoning_quality_reward_func": 0.786034345626831,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.559375012293458,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 198
    },
    {
      "completion_length": 319.8125,
      "epoch": 0.10964187327823692,
      "grad_norm": 0.1923948973417282,
      "kl": 0.02447389904409647,
      "learning_rate": 3.9916146100872335e-05,
      "loss": 0.001,
      "reward": 5.350589036941528,
      "reward_std": 1.022205639630556,
      "rewards/ngram_similarity_reward_func": 0.6956609636545181,
      "rewards/reasoning_quality_reward_func": 0.8775843381881714,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.03125,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 199
    },
    {
      "completion_length": 330.5,
      "epoch": 0.11019283746556474,
      "grad_norm": 0.21280765533447266,
      "kl": 0.02328745974227786,
      "learning_rate": 3.9913834845579497e-05,
      "loss": 0.0009,
      "reward": 6.1320801973342896,
      "reward_std": 0.6563425548374653,
      "rewards/ngram_similarity_reward_func": 0.7054241448640823,
      "rewards/reasoning_quality_reward_func": 0.8579062521457672,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.818750023841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 200
    },
    {
      "completion_length": 405.34375,
      "epoch": 0.11074380165289256,
      "grad_norm": 0.182103231549263,
      "kl": 0.03351978538557887,
      "learning_rate": 3.9911492245810205e-05,
      "loss": 0.0013,
      "reward": 4.166096389293671,
      "reward_std": 1.1394103169441223,
      "rewards/ngram_similarity_reward_func": 0.40636465325951576,
      "rewards/reasoning_quality_reward_func": 0.8845656365156174,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.1407911479473114,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 201
    },
    {
      "completion_length": 281.21875,
      "epoch": 0.11129476584022038,
      "grad_norm": 0.21188591420650482,
      "kl": 0.026487734634429216,
      "learning_rate": 3.990911830566341e-05,
      "loss": 0.0011,
      "reward": 5.1708256006240845,
      "reward_std": 0.23907891288399696,
      "rewards/ngram_similarity_reward_func": 0.5855287685990334,
      "rewards/reasoning_quality_reward_func": 0.8977968990802765,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.9375,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 202
    },
    {
      "completion_length": 353.59375,
      "epoch": 0.1118457300275482,
      "grad_norm": 0.2017420381307602,
      "kl": 0.02244984870776534,
      "learning_rate": 3.9906713029292857e-05,
      "loss": 0.0009,
      "reward": 5.772566795349121,
      "reward_std": 0.6675969585776329,
      "rewards/ngram_similarity_reward_func": 0.7385187670588493,
      "rewards/reasoning_quality_reward_func": 0.8501937389373779,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.4338542222976685,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 203
    },
    {
      "completion_length": 386.4375,
      "epoch": 0.11239669421487604,
      "grad_norm": 0.19382567703723907,
      "kl": 0.023544660303741693,
      "learning_rate": 3.9904276420907156e-05,
      "loss": 0.0009,
      "reward": 4.252572178840637,
      "reward_std": 0.9041159972548485,
      "rewards/ngram_similarity_reward_func": 0.5244164317846298,
      "rewards/reasoning_quality_reward_func": 0.9468812346458435,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.031274698674679,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 204
    },
    {
      "completion_length": 314.0,
      "epoch": 0.11294765840220386,
      "grad_norm": 0.2395251989364624,
      "kl": 0.02716870093718171,
      "learning_rate": 3.990180848476973e-05,
      "loss": 0.0011,
      "reward": 4.1917003989219666,
      "reward_std": 0.5034153386950493,
      "rewards/ngram_similarity_reward_func": 0.39592545852065086,
      "rewards/reasoning_quality_reward_func": 0.8332750052213669,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.2125000059604645,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 205
    },
    {
      "completion_length": 296.75,
      "epoch": 0.11349862258953168,
      "grad_norm": 0.21134355664253235,
      "kl": 0.02956994017586112,
      "learning_rate": 3.9899309225198804e-05,
      "loss": 0.0012,
      "reward": 5.693682789802551,
      "reward_std": 0.14962213300168514,
      "rewards/ngram_similarity_reward_func": 0.48320604115724564,
      "rewards/reasoning_quality_reward_func": 0.9472437500953674,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.5132330656051636,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 206
    },
    {
      "completion_length": 360.40625,
      "epoch": 0.1140495867768595,
      "grad_norm": 0.19918246567249298,
      "kl": 0.025009909877553582,
      "learning_rate": 3.9896778646567435e-05,
      "loss": 0.001,
      "reward": 5.201718509197235,
      "reward_std": 0.6198860108852386,
      "rewards/ngram_similarity_reward_func": 0.5952777564525604,
      "rewards/reasoning_quality_reward_func": 0.8564406335353851,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.0,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 207
    },
    {
      "completion_length": 300.5,
      "epoch": 0.11460055096418732,
      "grad_norm": 0.21714074909687042,
      "kl": 0.028932437300682068,
      "learning_rate": 3.989421675330346e-05,
      "loss": 0.0012,
      "reward": 5.289386034011841,
      "reward_std": 0.34446535632014275,
      "rewards/ngram_similarity_reward_func": 0.6295698061585426,
      "rewards/reasoning_quality_reward_func": 0.849631279706955,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.060185194015503,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 208
    },
    {
      "completion_length": 302.09375,
      "epoch": 0.11515151515151516,
      "grad_norm": 0.22969579696655273,
      "kl": 0.02834631549194455,
      "learning_rate": 3.989162354988951e-05,
      "loss": 0.0011,
      "reward": 6.066040635108948,
      "reward_std": 0.4212003592401743,
      "rewards/ngram_similarity_reward_func": 0.6426936239004135,
      "rewards/reasoning_quality_reward_func": 0.9233468472957611,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.75,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 209
    },
    {
      "completion_length": 282.9375,
      "epoch": 0.11570247933884298,
      "grad_norm": 0.27085578441619873,
      "kl": 0.03407869813963771,
      "learning_rate": 3.988899904086302e-05,
      "loss": 0.0014,
      "reward": 5.664801597595215,
      "reward_std": 0.6608479917049408,
      "rewards/ngram_similarity_reward_func": 0.4807889610528946,
      "rewards/reasoning_quality_reward_func": 0.8715125024318695,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.5625,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 210
    },
    {
      "completion_length": 302.4375,
      "epoch": 0.1162534435261708,
      "grad_norm": 0.23865915834903717,
      "kl": 0.028029571752995253,
      "learning_rate": 3.988634323081617e-05,
      "loss": 0.0011,
      "reward": 5.113776326179504,
      "reward_std": 0.6124768815934658,
      "rewards/ngram_similarity_reward_func": 0.588068351149559,
      "rewards/reasoning_quality_reward_func": 0.9353062510490417,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.840401768684387,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 211
    },
    {
      "completion_length": 404.5625,
      "epoch": 0.11680440771349862,
      "grad_norm": 0.19385962188243866,
      "kl": 0.023993458598852158,
      "learning_rate": 3.988365612439594e-05,
      "loss": 0.001,
      "reward": 4.831529200077057,
      "reward_std": 0.9868805557489395,
      "rewards/ngram_similarity_reward_func": 0.4991406351327896,
      "rewards/reasoning_quality_reward_func": 0.8356624841690063,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.7467261254787445,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 212
    },
    {
      "completion_length": 268.9375,
      "epoch": 0.11735537190082644,
      "grad_norm": 0.26125678420066833,
      "kl": 0.03615891141816974,
      "learning_rate": 3.988093772630405e-05,
      "loss": 0.0014,
      "reward": 4.792828559875488,
      "reward_std": 0.8394854664802551,
      "rewards/ngram_similarity_reward_func": 0.5779032781720161,
      "rewards/reasoning_quality_reward_func": 0.8452280759811401,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.6196970343589783,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 213
    },
    {
      "completion_length": 334.65625,
      "epoch": 0.11790633608815428,
      "grad_norm": 0.20092688500881195,
      "kl": 0.03267950704321265,
      "learning_rate": 3.987818804129698e-05,
      "loss": 0.0013,
      "reward": 4.568771660327911,
      "reward_std": 0.7863789275288582,
      "rewards/ngram_similarity_reward_func": 0.5678156092762947,
      "rewards/reasoning_quality_reward_func": 0.8088218569755554,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.442134201526642,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 214
    },
    {
      "completion_length": 318.25,
      "epoch": 0.1184573002754821,
      "grad_norm": 0.2342282086610794,
      "kl": 0.04227781528607011,
      "learning_rate": 3.9875407074185944e-05,
      "loss": 0.0017,
      "reward": 4.674242615699768,
      "reward_std": 0.40761171467602253,
      "rewards/ngram_similarity_reward_func": 0.5611892491579056,
      "rewards/reasoning_quality_reward_func": 0.8630531281232834,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.5,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 215
    },
    {
      "completion_length": 372.9375,
      "epoch": 0.11900826446280992,
      "grad_norm": 0.2003372460603714,
      "kl": 0.029141624458134174,
      "learning_rate": 3.98725948298369e-05,
      "loss": 0.0012,
      "reward": 3.812993586063385,
      "reward_std": 0.43411241471767426,
      "rewards/ngram_similarity_reward_func": 0.39938735216856003,
      "rewards/reasoning_quality_reward_func": 0.8104812502861023,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 1.868750000372529,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 216
    },
    {
      "completion_length": 396.25,
      "epoch": 0.11955922865013774,
      "grad_norm": 0.18980158865451813,
      "kl": 0.02535274112597108,
      "learning_rate": 3.986975131317055e-05,
      "loss": 0.001,
      "reward": 4.936676025390625,
      "reward_std": 0.364948783069849,
      "rewards/ngram_similarity_reward_func": 0.40192921832203865,
      "rewards/reasoning_quality_reward_func": 0.9534968733787537,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.831250011920929,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 217
    },
    {
      "completion_length": 375.1875,
      "epoch": 0.12011019283746556,
      "grad_norm": 0.20410524308681488,
      "kl": 0.028285119216889143,
      "learning_rate": 3.986687652916228e-05,
      "loss": 0.0011,
      "reward": 4.852037727832794,
      "reward_std": 0.6013146862387657,
      "rewards/ngram_similarity_reward_func": 0.5481557548046112,
      "rewards/reasoning_quality_reward_func": 0.9435781389474869,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.629835307598114,
      "rewards/xmlcount_reward_func": 0.48828125,
      "step": 218
    },
    {
      "completion_length": 313.8125,
      "epoch": 0.1206611570247934,
      "grad_norm": 0.2331204116344452,
      "kl": 0.029286117292940617,
      "learning_rate": 3.986397048284221e-05,
      "loss": 0.0012,
      "reward": 6.103626608848572,
      "reward_std": 0.7551473397761583,
      "rewards/ngram_similarity_reward_func": 0.9218818843364716,
      "rewards/reasoning_quality_reward_func": 0.9150781184434891,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.5166666507720947,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 219
    },
    {
      "completion_length": 453.65625,
      "epoch": 0.12121212121212122,
      "grad_norm": 0.17305000126361847,
      "kl": 0.024284148355945945,
      "learning_rate": 3.986103317929515e-05,
      "loss": 0.001,
      "reward": 4.042162179946899,
      "reward_std": 0.8696547485888004,
      "rewards/ngram_similarity_reward_func": 0.4331934526562691,
      "rewards/reasoning_quality_reward_func": 0.9441250115633011,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 1.9187500476837158,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 220
    },
    {
      "completion_length": 299.3125,
      "epoch": 0.12176308539944904,
      "grad_norm": 0.24970360100269318,
      "kl": 0.03603295097127557,
      "learning_rate": 3.985806462366061e-05,
      "loss": 0.0014,
      "reward": 4.9597339034080505,
      "reward_std": 0.6746684070676565,
      "rewards/ngram_similarity_reward_func": 0.5999251548200846,
      "rewards/reasoning_quality_reward_func": 0.9411812275648117,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.6686274707317352,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 221
    },
    {
      "completion_length": 329.71875,
      "epoch": 0.12231404958677686,
      "grad_norm": 0.22025638818740845,
      "kl": 0.029848780017346144,
      "learning_rate": 3.9855064821132766e-05,
      "loss": 0.0012,
      "reward": 4.7530147433280945,
      "reward_std": 0.8279001023620367,
      "rewards/ngram_similarity_reward_func": 0.6213285624980927,
      "rewards/reasoning_quality_reward_func": 0.935853123664856,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.4458332899957895,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 222
    },
    {
      "completion_length": 343.84375,
      "epoch": 0.12286501377410468,
      "grad_norm": 0.2058170586824417,
      "kl": 0.028606257401406765,
      "learning_rate": 3.985203377696048e-05,
      "loss": 0.0011,
      "reward": 3.611181676387787,
      "reward_std": 0.6118378676474094,
      "rewards/ngram_similarity_reward_func": 0.43057817965745926,
      "rewards/reasoning_quality_reward_func": 0.8422812223434448,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 1.5883223712444305,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 223
    },
    {
      "completion_length": 280.40625,
      "epoch": 0.12341597796143251,
      "grad_norm": 0.21927501261234283,
      "kl": 0.033682138193398714,
      "learning_rate": 3.984897149644729e-05,
      "loss": 0.0013,
      "reward": 5.1440348625183105,
      "reward_std": 0.20506118424236774,
      "rewards/ngram_similarity_reward_func": 0.5855693146586418,
      "rewards/reasoning_quality_reward_func": 0.8084656298160553,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.0,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 224
    },
    {
      "completion_length": 283.1875,
      "epoch": 0.12396694214876033,
      "grad_norm": 0.24156756699085236,
      "kl": 0.03955437010154128,
      "learning_rate": 3.984587798495138e-05,
      "loss": 0.0016,
      "reward": 4.03164005279541,
      "reward_std": 0.2968643084168434,
      "rewards/ngram_similarity_reward_func": 0.47685894183814526,
      "rewards/reasoning_quality_reward_func": 0.892281249165535,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 1.9124999884516,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 225
    },
    {
      "completion_length": 296.1875,
      "epoch": 0.12451790633608815,
      "grad_norm": 0.22788529098033905,
      "kl": 0.028762479312717915,
      "learning_rate": 3.9842753247885556e-05,
      "loss": 0.0012,
      "reward": 4.804973840713501,
      "reward_std": 0.3813077211380005,
      "rewards/ngram_similarity_reward_func": 0.6031543985009193,
      "rewards/reasoning_quality_reward_func": 0.8704156130552292,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.581403911113739,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 226
    },
    {
      "completion_length": 355.875,
      "epoch": 0.12506887052341598,
      "grad_norm": 0.21720479428768158,
      "kl": 0.036472429521381855,
      "learning_rate": 3.98395972907173e-05,
      "loss": 0.0015,
      "reward": 4.325909376144409,
      "reward_std": 1.0186720192432404,
      "rewards/ngram_similarity_reward_func": 0.36984385922551155,
      "rewards/reasoning_quality_reward_func": 0.8185656368732452,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.387499988079071,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 227
    },
    {
      "completion_length": 336.34375,
      "epoch": 0.1256198347107438,
      "grad_norm": 0.21111954748630524,
      "kl": 0.03481492307037115,
      "learning_rate": 3.98364101189687e-05,
      "loss": 0.0014,
      "reward": 5.871004104614258,
      "reward_std": 0.6391111630946398,
      "rewards/ngram_similarity_reward_func": 0.5640165954828262,
      "rewards/reasoning_quality_reward_func": 0.8694874793291092,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.6875,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 228
    },
    {
      "completion_length": 340.75,
      "epoch": 0.12617079889807162,
      "grad_norm": 0.2204820215702057,
      "kl": 0.031259714625775814,
      "learning_rate": 3.983319173821647e-05,
      "loss": 0.0012,
      "reward": 5.412296295166016,
      "reward_std": 0.8595960773527622,
      "rewards/ngram_similarity_reward_func": 0.7111901938915253,
      "rewards/reasoning_quality_reward_func": 0.8448562473058701,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.106250047683716,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 229
    },
    {
      "completion_length": 255.875,
      "epoch": 0.12672176308539945,
      "grad_norm": 0.25943106412887573,
      "kl": 0.04452124796807766,
      "learning_rate": 3.9829942154091906e-05,
      "loss": 0.0018,
      "reward": 6.061091780662537,
      "reward_std": 0.4641379490494728,
      "rewards/ngram_similarity_reward_func": 0.7555730193853378,
      "rewards/reasoning_quality_reward_func": 0.8680187463760376,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.6875,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 230
    },
    {
      "completion_length": 301.3125,
      "epoch": 0.12727272727272726,
      "grad_norm": 0.23102571070194244,
      "kl": 0.03599635045975447,
      "learning_rate": 3.9826661372280944e-05,
      "loss": 0.0014,
      "reward": 4.419783353805542,
      "reward_std": 0.8062819801270962,
      "rewards/ngram_similarity_reward_func": 0.5628854036331177,
      "rewards/reasoning_quality_reward_func": 0.8430249989032745,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.2873105704784393,
      "rewards/xmlcount_reward_func": 0.484375,
      "step": 231
    },
    {
      "completion_length": 273.75,
      "epoch": 0.1278236914600551,
      "grad_norm": 0.20261888206005096,
      "kl": 0.03829146595671773,
      "learning_rate": 3.9823349398524065e-05,
      "loss": 0.0015,
      "reward": 5.705955743789673,
      "reward_std": 0.09410573169589043,
      "rewards/ngram_similarity_reward_func": 0.53651512414217,
      "rewards/reasoning_quality_reward_func": 0.9233468621969223,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.5,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 232
    },
    {
      "completion_length": 252.0625,
      "epoch": 0.12837465564738293,
      "grad_norm": 0.2714664936065674,
      "kl": 0.038887111470103264,
      "learning_rate": 3.982000623861637e-05,
      "loss": 0.0016,
      "reward": 6.144020676612854,
      "reward_std": 0.4619959443807602,
      "rewards/ngram_similarity_reward_func": 0.8142364621162415,
      "rewards/reasoning_quality_reward_func": 0.8297843635082245,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.75,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 233
    },
    {
      "completion_length": 286.34375,
      "epoch": 0.12892561983471074,
      "grad_norm": 0.26958340406417847,
      "kl": 0.035695494152605534,
      "learning_rate": 3.981663189840749e-05,
      "loss": 0.0014,
      "reward": 5.006019473075867,
      "reward_std": 0.5233548749238253,
      "rewards/ngram_similarity_reward_func": 0.5396503284573555,
      "rewards/reasoning_quality_reward_func": 0.9669124782085419,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.7494564056396484,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 234
    },
    {
      "completion_length": 247.5625,
      "epoch": 0.12947658402203857,
      "grad_norm": 0.2858230769634247,
      "kl": 0.04263108689337969,
      "learning_rate": 3.981322638380162e-05,
      "loss": 0.0017,
      "reward": 5.012692451477051,
      "reward_std": 0.5616378858685493,
      "rewards/ngram_similarity_reward_func": 0.7794595509767532,
      "rewards/reasoning_quality_reward_func": 0.9307031184434891,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.552529752254486,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 235
    },
    {
      "completion_length": 304.65625,
      "epoch": 0.13002754820936638,
      "grad_norm": 0.23146970570087433,
      "kl": 0.03943014796823263,
      "learning_rate": 3.980978970075753e-05,
      "loss": 0.0016,
      "reward": 3.2968255281448364,
      "reward_std": 0.46545013785362244,
      "rewards/ngram_similarity_reward_func": 0.538728728890419,
      "rewards/reasoning_quality_reward_func": 0.9518468827009201,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 1.056250024586916,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 236
    },
    {
      "completion_length": 250.0,
      "epoch": 0.1305785123966942,
      "grad_norm": 0.28375205397605896,
      "kl": 0.048823241144418716,
      "learning_rate": 3.98063218552885e-05,
      "loss": 0.002,
      "reward": 6.333166718482971,
      "reward_std": 0.6670871237292886,
      "rewards/ngram_similarity_reward_func": 0.9430857449769974,
      "rewards/reasoning_quality_reward_func": 0.8213312327861786,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.818750023841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 237
    },
    {
      "completion_length": 288.28125,
      "epoch": 0.13112947658402205,
      "grad_norm": 0.2897850275039673,
      "kl": 0.05474950838834047,
      "learning_rate": 3.980282285346234e-05,
      "loss": 0.0022,
      "reward": 4.936635732650757,
      "reward_std": 0.9571874216198921,
      "rewards/ngram_similarity_reward_func": 0.5331421084702015,
      "rewards/reasoning_quality_reward_func": 0.8816187679767609,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.771875023841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 238
    },
    {
      "completion_length": 292.90625,
      "epoch": 0.13168044077134985,
      "grad_norm": 0.25066161155700684,
      "kl": 0.0505065331235528,
      "learning_rate": 3.979929270140138e-05,
      "loss": 0.002,
      "reward": 4.661517977714539,
      "reward_std": 0.617581807076931,
      "rewards/ngram_similarity_reward_func": 0.47327733039855957,
      "rewards/reasoning_quality_reward_func": 0.9132406264543533,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.525000000372529,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 239
    },
    {
      "completion_length": 293.1875,
      "epoch": 0.1322314049586777,
      "grad_norm": 0.25533682107925415,
      "kl": 0.035948424600064754,
      "learning_rate": 3.979573140528246e-05,
      "loss": 0.0014,
      "reward": 5.166618824005127,
      "reward_std": 0.30848325975239277,
      "rewards/ngram_similarity_reward_func": 0.6302158683538437,
      "rewards/reasoning_quality_reward_func": 0.8239031285047531,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.962500000372529,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 240
    },
    {
      "completion_length": 298.8125,
      "epoch": 0.1327823691460055,
      "grad_norm": 0.25314339995384216,
      "kl": 0.04567734710872173,
      "learning_rate": 3.9792138971336894e-05,
      "loss": 0.0018,
      "reward": 4.819394946098328,
      "reward_std": 0.539910311345011,
      "rewards/ngram_similarity_reward_func": 0.48395439982414246,
      "rewards/reasoning_quality_reward_func": 0.8862218409776688,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.7109375,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 241
    },
    {
      "completion_length": 291.90625,
      "epoch": 0.13333333333333333,
      "grad_norm": 0.2507646977901459,
      "kl": 0.04530429048463702,
      "learning_rate": 3.97885154058505e-05,
      "loss": 0.0018,
      "reward": 5.002235651016235,
      "reward_std": 1.1008880883455276,
      "rewards/ngram_similarity_reward_func": 0.3467079699039459,
      "rewards/reasoning_quality_reward_func": 0.8908156156539917,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.014712154865265,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 242
    },
    {
      "completion_length": 296.625,
      "epoch": 0.13388429752066117,
      "grad_norm": 0.22489966452121735,
      "kl": 0.04879872780293226,
      "learning_rate": 3.9784860715163554e-05,
      "loss": 0.002,
      "reward": 4.727094829082489,
      "reward_std": 0.5959070585668087,
      "rewards/ngram_similarity_reward_func": 0.6028279513120651,
      "rewards/reasoning_quality_reward_func": 0.8726093620061874,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.5133764147758484,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 243
    },
    {
      "completion_length": 295.21875,
      "epoch": 0.13443526170798897,
      "grad_norm": 0.24386678636074066,
      "kl": 0.051864149048924446,
      "learning_rate": 3.978117490567081e-05,
      "loss": 0.0021,
      "reward": 5.0689777135849,
      "reward_std": 1.3473074436187744,
      "rewards/ngram_similarity_reward_func": 0.5109423100948334,
      "rewards/reasoning_quality_reward_func": 0.8226187527179718,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.9854166209697723,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 244
    },
    {
      "completion_length": 302.53125,
      "epoch": 0.1349862258953168,
      "grad_norm": 0.23841246962547302,
      "kl": 0.046447524800896645,
      "learning_rate": 3.977745798382145e-05,
      "loss": 0.0019,
      "reward": 5.19568932056427,
      "reward_std": 0.6665928773581982,
      "rewards/ngram_similarity_reward_func": 0.5114956721663475,
      "rewards/reasoning_quality_reward_func": 0.8529437333345413,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.081250011920929,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 245
    },
    {
      "completion_length": 285.3125,
      "epoch": 0.13553719008264462,
      "grad_norm": 0.24394561350345612,
      "kl": 0.07110137771815062,
      "learning_rate": 3.977370995611911e-05,
      "loss": 0.0028,
      "reward": 5.317797780036926,
      "reward_std": 0.748687582090497,
      "rewards/ngram_similarity_reward_func": 0.761343315243721,
      "rewards/reasoning_quality_reward_func": 0.9205906242132187,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.885863959789276,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 246
    },
    {
      "completion_length": 240.6875,
      "epoch": 0.13608815426997245,
      "grad_norm": 0.5823207497596741,
      "kl": 0.05728767812252045,
      "learning_rate": 3.976993082912184e-05,
      "loss": 0.0023,
      "reward": 4.7882702350616455,
      "reward_std": 1.1691070161759853,
      "rewards/ngram_similarity_reward_func": 0.8106983751058578,
      "rewards/reasoning_quality_reward_func": 0.8306968808174133,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.4124999884516,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 247
    },
    {
      "completion_length": 383.125,
      "epoch": 0.13663911845730028,
      "grad_norm": 0.20902281999588013,
      "kl": 0.04080106969922781,
      "learning_rate": 3.976612060944212e-05,
      "loss": 0.0016,
      "reward": 3.9735687971115112,
      "reward_std": 0.9921468570828438,
      "rewards/ngram_similarity_reward_func": 0.553644448518753,
      "rewards/reasoning_quality_reward_func": 0.9062531292438507,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 1.763671338558197,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 248
    },
    {
      "completion_length": 325.3125,
      "epoch": 0.1371900826446281,
      "grad_norm": 0.20672467350959778,
      "kl": 0.048159994184970856,
      "learning_rate": 3.9762279303746834e-05,
      "loss": 0.0019,
      "reward": 5.602479934692383,
      "reward_std": 0.6213366352021694,
      "rewards/ngram_similarity_reward_func": 0.7582212835550308,
      "rewards/reasoning_quality_reward_func": 0.8691312670707703,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 3.2368461191654205,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 249
    },
    {
      "completion_length": 240.625,
      "epoch": 0.13774104683195593,
      "grad_norm": 0.23862190544605255,
      "kl": 0.043528757989406586,
      "learning_rate": 3.975840691875725e-05,
      "loss": 0.0017,
      "reward": 5.306386113166809,
      "reward_std": 0.40666226856410503,
      "rewards/ngram_similarity_reward_func": 0.582251526415348,
      "rewards/reasoning_quality_reward_func": 0.9582843631505966,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.0158501863479614,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 250
    },
    {
      "completion_length": 258.75,
      "epoch": 0.13829201101928373,
      "grad_norm": 0.242393359541893,
      "kl": 0.04345339676365256,
      "learning_rate": 3.975450346124901e-05,
      "loss": 0.0017,
      "reward": 5.407206416130066,
      "reward_std": 0.1907122479751706,
      "rewards/ngram_similarity_reward_func": 0.7364222556352615,
      "rewards/reasoning_quality_reward_func": 0.9207843691110611,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.0,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 251
    },
    {
      "completion_length": 375.8125,
      "epoch": 0.13884297520661157,
      "grad_norm": 0.20287199318408966,
      "kl": 0.04171336302533746,
      "learning_rate": 3.975056893805213e-05,
      "loss": 0.0017,
      "reward": 5.309088826179504,
      "reward_std": 0.8154166713356972,
      "rewards/ngram_similarity_reward_func": 0.57688919454813,
      "rewards/reasoning_quality_reward_func": 0.9614062458276749,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.0286057591438293,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 252
    },
    {
      "completion_length": 255.3125,
      "epoch": 0.1393939393939394,
      "grad_norm": 0.37277403473854065,
      "kl": 0.055895852856338024,
      "learning_rate": 3.9746603356050995e-05,
      "loss": 0.0022,
      "reward": 5.884204268455505,
      "reward_std": 0.7793909600004554,
      "rewards/ngram_similarity_reward_func": 0.8019774854183197,
      "rewards/reasoning_quality_reward_func": 0.8433874696493149,
      "rewards/soft_format_reward_func": 0.234375,
      "rewards/sql_execution_reward_func": 3.520089268684387,
      "rewards/xmlcount_reward_func": 0.484375,
      "step": 253
    },
    {
      "completion_length": 255.875,
      "epoch": 0.1399449035812672,
      "grad_norm": 0.29770925641059875,
      "kl": 0.0581395011395216,
      "learning_rate": 3.974260672218432e-05,
      "loss": 0.0023,
      "reward": 5.9299410581588745,
      "reward_std": 0.7718696547672153,
      "rewards/ngram_similarity_reward_func": 0.8424378931522369,
      "rewards/reasoning_quality_reward_func": 0.900003120303154,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.4375,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 254
    },
    {
      "completion_length": 257.96875,
      "epoch": 0.14049586776859505,
      "grad_norm": 0.26074984669685364,
      "kl": 0.05843827687203884,
      "learning_rate": 3.973857904344517e-05,
      "loss": 0.0023,
      "reward": 5.4922404289245605,
      "reward_std": 0.6049819709733129,
      "rewards/ngram_similarity_reward_func": 0.7895422875881195,
      "rewards/reasoning_quality_reward_func": 0.8641562163829803,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.088541626930237,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 255
    },
    {
      "completion_length": 324.96875,
      "epoch": 0.14104683195592285,
      "grad_norm": 0.19740520417690277,
      "kl": 0.03368695126846433,
      "learning_rate": 3.9734520326880896e-05,
      "loss": 0.0013,
      "reward": 4.873990416526794,
      "reward_std": 0.32823346462100744,
      "rewards/ngram_similarity_reward_func": 0.5533030405640602,
      "rewards/reasoning_quality_reward_func": 0.9691250026226044,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.609375024214387,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 256
    },
    {
      "completion_length": 244.65625,
      "epoch": 0.1415977961432507,
      "grad_norm": 0.2642117440700531,
      "kl": 0.05307223042473197,
      "learning_rate": 3.9730430579593194e-05,
      "loss": 0.0021,
      "reward": 5.556440949440002,
      "reward_std": 0.5802447013556957,
      "rewards/ngram_similarity_reward_func": 0.8942098617553711,
      "rewards/reasoning_quality_reward_func": 0.8356687277555466,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.0843749940395355,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 257
    },
    {
      "completion_length": 321.5625,
      "epoch": 0.14214876033057852,
      "grad_norm": 0.22471274435520172,
      "kl": 0.03925685305148363,
      "learning_rate": 3.972630980873805e-05,
      "loss": 0.0016,
      "reward": 4.501880168914795,
      "reward_std": 0.3446642216295004,
      "rewards/ngram_similarity_reward_func": 0.8485676646232605,
      "rewards/reasoning_quality_reward_func": 0.8158125132322311,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.087500000372529,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 258
    },
    {
      "completion_length": 277.3125,
      "epoch": 0.14269972451790633,
      "grad_norm": 0.27796733379364014,
      "kl": 0.04466933757066727,
      "learning_rate": 3.9722158021525705e-05,
      "loss": 0.0018,
      "reward": 5.723112463951111,
      "reward_std": 0.7802695594727993,
      "rewards/ngram_similarity_reward_func": 0.784197986125946,
      "rewards/reasoning_quality_reward_func": 0.9308843612670898,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.25803005695343,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 259
    },
    {
      "completion_length": 296.5,
      "epoch": 0.14325068870523416,
      "grad_norm": 0.2100607454776764,
      "kl": 0.04140225937590003,
      "learning_rate": 3.9717975225220694e-05,
      "loss": 0.0017,
      "reward": 5.683765530586243,
      "reward_std": 0.17780059948563576,
      "rewards/ngram_similarity_reward_func": 0.5844893828034401,
      "rewards/reasoning_quality_reward_func": 0.8843874931335449,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 3.480513632297516,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 260
    },
    {
      "completion_length": 286.9375,
      "epoch": 0.14380165289256197,
      "grad_norm": 0.22722919285297394,
      "kl": 0.04341763723641634,
      "learning_rate": 3.971376142714181e-05,
      "loss": 0.0017,
      "reward": 5.795986533164978,
      "reward_std": 0.7511387579143047,
      "rewards/ngram_similarity_reward_func": 0.6579175814986229,
      "rewards/reasoning_quality_reward_func": 0.8193187415599823,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.568750023841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 261
    },
    {
      "completion_length": 272.28125,
      "epoch": 0.1443526170798898,
      "grad_norm": 0.24568969011306763,
      "kl": 0.04710317309945822,
      "learning_rate": 3.970951663466208e-05,
      "loss": 0.0019,
      "reward": 5.506887316703796,
      "reward_std": 0.2679957514628768,
      "rewards/ngram_similarity_reward_func": 0.5386841148138046,
      "rewards/reasoning_quality_reward_func": 0.8682031035423279,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.350000023841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 262
    },
    {
      "completion_length": 305.375,
      "epoch": 0.14490358126721764,
      "grad_norm": 0.21139253675937653,
      "kl": 0.04605023190379143,
      "learning_rate": 3.9705240855208784e-05,
      "loss": 0.0018,
      "reward": 5.181397914886475,
      "reward_std": 1.057370364665985,
      "rewards/ngram_similarity_reward_func": 0.661843441426754,
      "rewards/reasoning_quality_reward_func": 0.888234406709671,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.8813201785087585,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 263
    },
    {
      "completion_length": 271.03125,
      "epoch": 0.14545454545454545,
      "grad_norm": 0.26097819209098816,
      "kl": 0.051187122240662575,
      "learning_rate": 3.970093409626339e-05,
      "loss": 0.002,
      "reward": 5.423590660095215,
      "reward_std": 1.1091685481369495,
      "rewards/ngram_similarity_reward_func": 0.6472969055175781,
      "rewards/reasoning_quality_reward_func": 0.8262937366962433,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.199999988079071,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 264
    },
    {
      "completion_length": 294.96875,
      "epoch": 0.14600550964187328,
      "grad_norm": 0.19815152883529663,
      "kl": 0.03947392990812659,
      "learning_rate": 3.969659636536159e-05,
      "loss": 0.0016,
      "reward": 6.399207711219788,
      "reward_std": 0.6201869952492416,
      "rewards/ngram_similarity_reward_func": 0.8455296456813812,
      "rewards/reasoning_quality_reward_func": 0.9849281013011932,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.818750023841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 265
    },
    {
      "completion_length": 244.65625,
      "epoch": 0.1465564738292011,
      "grad_norm": 0.28845566511154175,
      "kl": 0.05662275850772858,
      "learning_rate": 3.969222767009327e-05,
      "loss": 0.0023,
      "reward": 6.5214715003967285,
      "reward_std": 0.4162396639585495,
      "rewards/ngram_similarity_reward_func": 0.9302965402603149,
      "rewards/reasoning_quality_reward_func": 0.966174989938736,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.875,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 266
    },
    {
      "completion_length": 293.6875,
      "epoch": 0.14710743801652892,
      "grad_norm": 0.22893352806568146,
      "kl": 0.049870249815285206,
      "learning_rate": 3.968782801810249e-05,
      "loss": 0.002,
      "reward": 5.87822413444519,
      "reward_std": 0.23571899812668562,
      "rewards/ngram_similarity_reward_func": 0.7401157915592194,
      "rewards/reasoning_quality_reward_func": 0.9297750145196915,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.458333373069763,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 267
    },
    {
      "completion_length": 266.46875,
      "epoch": 0.14765840220385676,
      "grad_norm": 0.24100886285305023,
      "kl": 0.04448758251965046,
      "learning_rate": 3.9683397417087485e-05,
      "loss": 0.0018,
      "reward": 5.794814825057983,
      "reward_std": 0.9286904577165842,
      "rewards/ngram_similarity_reward_func": 0.6004566587507725,
      "rewards/reasoning_quality_reward_func": 0.8016500174999237,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.6427082419395447,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 268
    },
    {
      "completion_length": 270.0,
      "epoch": 0.14820936639118457,
      "grad_norm": 0.21846872568130493,
      "kl": 0.04746053647249937,
      "learning_rate": 3.9678935874800625e-05,
      "loss": 0.0019,
      "reward": 5.677929759025574,
      "reward_std": 0.5501862168312073,
      "rewards/ngram_similarity_reward_func": 0.7018141895532608,
      "rewards/reasoning_quality_reward_func": 0.9011155962944031,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.324999988079071,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 269
    },
    {
      "completion_length": 277.78125,
      "epoch": 0.1487603305785124,
      "grad_norm": 0.22686147689819336,
      "kl": 0.04434172995388508,
      "learning_rate": 3.967444339904844e-05,
      "loss": 0.0018,
      "reward": 5.99455201625824,
      "reward_std": 0.9304151609539986,
      "rewards/ngram_similarity_reward_func": 0.8580194413661957,
      "rewards/reasoning_quality_reward_func": 0.9534968584775925,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.433035731315613,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 270
    },
    {
      "completion_length": 303.5,
      "epoch": 0.1493112947658402,
      "grad_norm": 0.1960648149251938,
      "kl": 0.043685623444616795,
      "learning_rate": 3.966991999769157e-05,
      "loss": 0.0017,
      "reward": 4.973733127117157,
      "reward_std": 0.4642886146903038,
      "rewards/ngram_similarity_reward_func": 0.6768549829721451,
      "rewards/reasoning_quality_reward_func": 0.853128120303154,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.6937499940395355,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 271
    },
    {
      "completion_length": 233.34375,
      "epoch": 0.14986225895316804,
      "grad_norm": 0.24710723757743835,
      "kl": 0.04824818857014179,
      "learning_rate": 3.9665365678644775e-05,
      "loss": 0.0019,
      "reward": 5.288196682929993,
      "reward_std": 0.331680865958333,
      "rewards/ngram_similarity_reward_func": 0.7206526473164558,
      "rewards/reasoning_quality_reward_func": 0.8544187396764755,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.963125467300415,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 272
    },
    {
      "completion_length": 300.375,
      "epoch": 0.15041322314049588,
      "grad_norm": 0.22310598194599152,
      "kl": 0.04207586403936148,
      "learning_rate": 3.96607804498769e-05,
      "loss": 0.0017,
      "reward": 6.412749409675598,
      "reward_std": 0.13878355454653502,
      "rewards/ngram_similarity_reward_func": 0.7110868729650974,
      "rewards/reasoning_quality_reward_func": 0.9516624957323074,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 4.0,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 273
    },
    {
      "completion_length": 317.75,
      "epoch": 0.15096418732782368,
      "grad_norm": 0.21317726373672485,
      "kl": 0.0413897605612874,
      "learning_rate": 3.9656164319410896e-05,
      "loss": 0.0017,
      "reward": 5.604174494743347,
      "reward_std": 0.6962556056678295,
      "rewards/ngram_similarity_reward_func": 0.7298476845026016,
      "rewards/reasoning_quality_reward_func": 0.8636125177145004,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.260714292526245,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 274
    },
    {
      "completion_length": 288.46875,
      "epoch": 0.15151515151515152,
      "grad_norm": 0.22421520948410034,
      "kl": 0.050917964428663254,
      "learning_rate": 3.965151729532376e-05,
      "loss": 0.002,
      "reward": 5.4232847690582275,
      "reward_std": 0.5644791424274445,
      "rewards/ngram_similarity_reward_func": 0.6506596505641937,
      "rewards/reasoning_quality_reward_func": 0.8788749873638153,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.143750011920929,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 275
    },
    {
      "completion_length": 251.71875,
      "epoch": 0.15206611570247933,
      "grad_norm": 0.23075851798057556,
      "kl": 0.049781170673668385,
      "learning_rate": 3.964683938574657e-05,
      "loss": 0.002,
      "reward": 6.158035159111023,
      "reward_std": 0.6800675764679909,
      "rewards/ngram_similarity_reward_func": 0.7913953810930252,
      "rewards/reasoning_quality_reward_func": 0.9652531147003174,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.651386857032776,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 276
    },
    {
      "completion_length": 335.34375,
      "epoch": 0.15261707988980716,
      "grad_norm": 0.22718046605587006,
      "kl": 0.04499669745564461,
      "learning_rate": 3.9642130598864416e-05,
      "loss": 0.0018,
      "reward": 4.994721055030823,
      "reward_std": 1.4692881107330322,
      "rewards/ngram_similarity_reward_func": 0.5956012457609177,
      "rewards/reasoning_quality_reward_func": 0.9838281124830246,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.665291577577591,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 277
    },
    {
      "completion_length": 322.0,
      "epoch": 0.153168044077135,
      "grad_norm": 0.22079861164093018,
      "kl": 0.04652388393878937,
      "learning_rate": 3.963739094291645e-05,
      "loss": 0.0019,
      "reward": 5.424874842166901,
      "reward_std": 0.7950988709926605,
      "rewards/ngram_similarity_reward_func": 0.6247125938534737,
      "rewards/reasoning_quality_reward_func": 0.9047874808311462,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.1453746557235718,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 278
    },
    {
      "completion_length": 338.4375,
      "epoch": 0.1537190082644628,
      "grad_norm": 0.1912488043308258,
      "kl": 0.04448720021173358,
      "learning_rate": 3.963262042619579e-05,
      "loss": 0.0018,
      "reward": 4.985793948173523,
      "reward_std": 0.6385920755565166,
      "rewards/ngram_similarity_reward_func": 0.45386505872011185,
      "rewards/reasoning_quality_reward_func": 0.9066249877214432,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.87530380487442,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 279
    },
    {
      "completion_length": 311.9375,
      "epoch": 0.15426997245179064,
      "grad_norm": 0.23654381930828094,
      "kl": 0.042094375006854534,
      "learning_rate": 3.9627819057049606e-05,
      "loss": 0.0017,
      "reward": 5.476590871810913,
      "reward_std": 0.9432802386581898,
      "rewards/ngram_similarity_reward_func": 0.7778784781694412,
      "rewards/reasoning_quality_reward_func": 0.9674625098705292,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.981249988079071,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 280
    },
    {
      "completion_length": 312.84375,
      "epoch": 0.15482093663911844,
      "grad_norm": 0.19831141829490662,
      "kl": 0.04571097483858466,
      "learning_rate": 3.962298684387901e-05,
      "loss": 0.0018,
      "reward": 5.465214550495148,
      "reward_std": 0.37351537495851517,
      "rewards/ngram_similarity_reward_func": 0.8932832926511765,
      "rewards/reasoning_quality_reward_func": 0.8711499869823456,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.962500000372529,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 281
    },
    {
      "completion_length": 311.4375,
      "epoch": 0.15537190082644628,
      "grad_norm": 0.21597616374492645,
      "kl": 0.04106222651898861,
      "learning_rate": 3.9618123795139097e-05,
      "loss": 0.0016,
      "reward": 5.318217515945435,
      "reward_std": 0.4485152531415224,
      "rewards/ngram_similarity_reward_func": 0.7299813777208328,
      "rewards/reasoning_quality_reward_func": 0.8781312555074692,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.9601049665361643,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 282
    },
    {
      "completion_length": 346.75,
      "epoch": 0.15592286501377411,
      "grad_norm": 0.21119298040866852,
      "kl": 0.04303087433800101,
      "learning_rate": 3.961322991933893e-05,
      "loss": 0.0017,
      "reward": 4.398180544376373,
      "reward_std": 0.5019839182496071,
      "rewards/ngram_similarity_reward_func": 0.6567201986908913,
      "rewards/reasoning_quality_reward_func": 0.9292281419038773,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.0622321609407663,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 283
    },
    {
      "completion_length": 346.5625,
      "epoch": 0.15647382920110192,
      "grad_norm": 0.19119900465011597,
      "kl": 0.03984567429870367,
      "learning_rate": 3.960830522504148e-05,
      "loss": 0.0016,
      "reward": 5.276624083518982,
      "reward_std": 0.9634360345080495,
      "rewards/ngram_similarity_reward_func": 0.4931582808494568,
      "rewards/reasoning_quality_reward_func": 0.8990906029939651,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.134375035762787,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 284
    },
    {
      "completion_length": 345.96875,
      "epoch": 0.15702479338842976,
      "grad_norm": 0.19695895910263062,
      "kl": 0.04605107009410858,
      "learning_rate": 3.960334972086368e-05,
      "loss": 0.0018,
      "reward": 5.900192379951477,
      "reward_std": 0.33200081810355186,
      "rewards/ngram_similarity_reward_func": 0.7789903432130814,
      "rewards/reasoning_quality_reward_func": 0.9128687381744385,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.458333373069763,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 285
    },
    {
      "completion_length": 313.78125,
      "epoch": 0.15757575757575756,
      "grad_norm": 0.2306852638721466,
      "kl": 0.046183349564671516,
      "learning_rate": 3.959836341547633e-05,
      "loss": 0.0018,
      "reward": 5.900614619255066,
      "reward_std": 1.0413149818778038,
      "rewards/ngram_similarity_reward_func": 0.6455520987510681,
      "rewards/reasoning_quality_reward_func": 0.9128749966621399,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.592187523841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 286
    },
    {
      "completion_length": 402.09375,
      "epoch": 0.1581267217630854,
      "grad_norm": 0.19088539481163025,
      "kl": 0.0372261106967926,
      "learning_rate": 3.959334631760417e-05,
      "loss": 0.0015,
      "reward": 4.586619019508362,
      "reward_std": 0.6555525548756123,
      "rewards/ngram_similarity_reward_func": 0.49085547029972076,
      "rewards/reasoning_quality_reward_func": 0.9001906365156174,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.4455728828907013,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 287
    },
    {
      "completion_length": 344.0,
      "epoch": 0.15867768595041323,
      "grad_norm": 0.1954616755247116,
      "kl": 0.04827097523957491,
      "learning_rate": 3.958829843602578e-05,
      "loss": 0.0019,
      "reward": 5.24393105506897,
      "reward_std": 0.9986111298203468,
      "rewards/ngram_similarity_reward_func": 0.5283029600977898,
      "rewards/reasoning_quality_reward_func": 0.915628120303154,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.0500000417232513,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 288
    },
    {
      "completion_length": 341.6875,
      "epoch": 0.15922865013774104,
      "grad_norm": 0.1899774968624115,
      "kl": 0.048033496364951134,
      "learning_rate": 3.9583219779573627e-05,
      "loss": 0.0019,
      "reward": 5.464048862457275,
      "reward_std": 0.3537519481033087,
      "rewards/ngram_similarity_reward_func": 0.58940839022398,
      "rewards/reasoning_quality_reward_func": 0.9715156108140945,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.153124988079071,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 289
    },
    {
      "completion_length": 312.0,
      "epoch": 0.15977961432506887,
      "grad_norm": 0.2059154361486435,
      "kl": 0.04651415906846523,
      "learning_rate": 3.957811035713401e-05,
      "loss": 0.0019,
      "reward": 6.19173276424408,
      "reward_std": 0.34558434644714,
      "rewards/ngram_similarity_reward_func": 0.662133127450943,
      "rewards/reasoning_quality_reward_func": 0.904599979519844,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.875,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 290
    },
    {
      "completion_length": 323.90625,
      "epoch": 0.16033057851239668,
      "grad_norm": 0.23240487277507782,
      "kl": 0.04265825217589736,
      "learning_rate": 3.9572970177647075e-05,
      "loss": 0.0017,
      "reward": 4.652861475944519,
      "reward_std": 1.0630485489964485,
      "rewards/ngram_similarity_reward_func": 0.646807998418808,
      "rewards/reasoning_quality_reward_func": 0.8400812447071075,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.415972203016281,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 291
    },
    {
      "completion_length": 358.625,
      "epoch": 0.16088154269972452,
      "grad_norm": 0.1716865450143814,
      "kl": 0.03734135162085295,
      "learning_rate": 3.956779925010677e-05,
      "loss": 0.0015,
      "reward": 4.951283812522888,
      "reward_std": 0.7698026737198234,
      "rewards/ngram_similarity_reward_func": 0.5239454880356789,
      "rewards/reasoning_quality_reward_func": 0.9681999981403351,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.70913827419281,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 292
    },
    {
      "completion_length": 268.0,
      "epoch": 0.16143250688705235,
      "grad_norm": 0.2326546013355255,
      "kl": 0.05312045384198427,
      "learning_rate": 3.9562597583560853e-05,
      "loss": 0.0021,
      "reward": 6.010141849517822,
      "reward_std": 0.13594077294692397,
      "rewards/ngram_similarity_reward_func": 0.8132637590169907,
      "rewards/reasoning_quality_reward_func": 0.9468781054019928,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.5,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 293
    },
    {
      "completion_length": 348.375,
      "epoch": 0.16198347107438016,
      "grad_norm": 0.20909111201763153,
      "kl": 0.04334063595160842,
      "learning_rate": 3.955736518711088e-05,
      "loss": 0.0017,
      "reward": 4.645334839820862,
      "reward_std": 1.0997912641614676,
      "rewards/ngram_similarity_reward_func": 0.6751100122928619,
      "rewards/reasoning_quality_reward_func": 0.9170999974012375,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.303124964237213,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 294
    },
    {
      "completion_length": 334.28125,
      "epoch": 0.162534435261708,
      "grad_norm": 0.2078557014465332,
      "kl": 0.05154033098369837,
      "learning_rate": 3.955210206991214e-05,
      "loss": 0.0021,
      "reward": 4.414716958999634,
      "reward_std": 0.27334881015121937,
      "rewards/ngram_similarity_reward_func": 0.8088349848985672,
      "rewards/reasoning_quality_reward_func": 0.8915437608957291,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 1.964338207617402,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 295
    },
    {
      "completion_length": 291.34375,
      "epoch": 0.1630853994490358,
      "grad_norm": 0.2205725908279419,
      "kl": 0.05120929051190615,
      "learning_rate": 3.954680824117371e-05,
      "loss": 0.002,
      "reward": 4.995299935340881,
      "reward_std": 0.7768121734261513,
      "rewards/ngram_similarity_reward_func": 0.6539343222975731,
      "rewards/reasoning_quality_reward_func": 0.8788656145334244,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.712499976158142,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 296
    },
    {
      "completion_length": 360.8125,
      "epoch": 0.16363636363636364,
      "grad_norm": 0.2111254781484604,
      "kl": 0.05011467915028334,
      "learning_rate": 3.9541483710158394e-05,
      "loss": 0.002,
      "reward": 5.093964338302612,
      "reward_std": 0.6022808458656073,
      "rewards/ngram_similarity_reward_func": 0.5596413463354111,
      "rewards/reasoning_quality_reward_func": 0.8801562637090683,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.9041666388511658,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 297
    },
    {
      "completion_length": 356.21875,
      "epoch": 0.16418732782369147,
      "grad_norm": 0.18199336528778076,
      "kl": 0.039406790398061275,
      "learning_rate": 3.95361284861827e-05,
      "loss": 0.0016,
      "reward": 5.053895711898804,
      "reward_std": 0.707212783396244,
      "rewards/ngram_similarity_reward_func": 0.5054874122142792,
      "rewards/reasoning_quality_reward_func": 0.9025749713182449,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.8958332538604736,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 298
    },
    {
      "completion_length": 279.46875,
      "epoch": 0.16473829201101928,
      "grad_norm": 0.22217494249343872,
      "kl": 0.04744693264365196,
      "learning_rate": 3.953074257861687e-05,
      "loss": 0.0019,
      "reward": 4.5300610065460205,
      "reward_std": 0.8665875643491745,
      "rewards/ngram_similarity_reward_func": 0.5285483300685883,
      "rewards/reasoning_quality_reward_func": 0.9544124901294708,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.2971000969409943,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 299
    },
    {
      "completion_length": 358.625,
      "epoch": 0.1652892561983471,
      "grad_norm": 0.18872666358947754,
      "kl": 0.03953286446630955,
      "learning_rate": 3.9525325996884806e-05,
      "loss": 0.0016,
      "reward": 5.148352324962616,
      "reward_std": 0.8442191705107689,
      "rewards/ngram_similarity_reward_func": 0.572711817920208,
      "rewards/reasoning_quality_reward_func": 0.9073531180620193,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.9378186464309692,
      "rewards/xmlcount_reward_func": 0.48828125,
      "step": 300
    },
    {
      "completion_length": 231.0625,
      "epoch": 0.16584022038567492,
      "grad_norm": 15085.142578125,
      "kl": 1487.7954742237926,
      "learning_rate": 3.9519878750464104e-05,
      "loss": 59.5118,
      "reward": 4.705110490322113,
      "reward_std": 0.9745630919933319,
      "rewards/ngram_similarity_reward_func": 0.6344875395298004,
      "rewards/reasoning_quality_reward_func": 0.8397093415260315,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.4887260496616364,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 301
    },
    {
      "completion_length": 325.25,
      "epoch": 0.16639118457300275,
      "grad_norm": 0.19878940284252167,
      "kl": 0.045971596613526344,
      "learning_rate": 3.9514400848886e-05,
      "loss": 0.0018,
      "reward": 4.153869867324829,
      "reward_std": 0.46764261834323406,
      "rewards/ngram_similarity_reward_func": 0.5976075232028961,
      "rewards/reasoning_quality_reward_func": 0.8437624722719193,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 1.962500000372529,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 302
    },
    {
      "completion_length": 372.1875,
      "epoch": 0.1669421487603306,
      "grad_norm": 0.19307558238506317,
      "kl": 0.03322633821517229,
      "learning_rate": 3.950889230173538e-05,
      "loss": 0.0013,
      "reward": 5.475817263126373,
      "reward_std": 0.7113875262439251,
      "rewards/ngram_similarity_reward_func": 0.7144142091274261,
      "rewards/reasoning_quality_reward_func": 0.8395281285047531,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.171875,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 303
    },
    {
      "completion_length": 287.34375,
      "epoch": 0.1674931129476584,
      "grad_norm": 0.23243941366672516,
      "kl": 0.04701520316302776,
      "learning_rate": 3.9503353118650746e-05,
      "loss": 0.0019,
      "reward": 4.836592972278595,
      "reward_std": 0.8819735422730446,
      "rewards/ngram_similarity_reward_func": 0.6426493376493454,
      "rewards/reasoning_quality_reward_func": 0.7783187627792358,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.6656249463558197,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 304
    },
    {
      "completion_length": 255.875,
      "epoch": 0.16804407713498623,
      "grad_norm": 0.2246531844139099,
      "kl": 0.037704838905483484,
      "learning_rate": 3.949778330932419e-05,
      "loss": 0.0015,
      "reward": 5.717690587043762,
      "reward_std": 0.3982728123664856,
      "rewards/ngram_similarity_reward_func": 0.49516966938972473,
      "rewards/reasoning_quality_reward_func": 0.8650749623775482,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.607446014881134,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 305
    },
    {
      "completion_length": 325.4375,
      "epoch": 0.16859504132231404,
      "grad_norm": 0.18789063394069672,
      "kl": 0.029463017359375954,
      "learning_rate": 3.949218288350143e-05,
      "loss": 0.0012,
      "reward": 5.543826520442963,
      "reward_std": 0.33056612126529217,
      "rewards/ngram_similarity_reward_func": 0.5390391424298286,
      "rewards/reasoning_quality_reward_func": 0.8172874599695206,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.4375,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 306
    },
    {
      "completion_length": 296.15625,
      "epoch": 0.16914600550964187,
      "grad_norm": 0.16334256529808044,
      "kl": 0.03493391117081046,
      "learning_rate": 3.94865518509817e-05,
      "loss": 0.0014,
      "reward": 5.422693729400635,
      "reward_std": 0.44842586666345596,
      "rewards/ngram_similarity_reward_func": 0.762115553021431,
      "rewards/reasoning_quality_reward_func": 0.968390628695488,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.942187488079071,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 307
    },
    {
      "completion_length": 241.84375,
      "epoch": 0.1696969696969697,
      "grad_norm": 0.2291998565196991,
      "kl": 0.04088439978659153,
      "learning_rate": 3.9480890221617845e-05,
      "loss": 0.0016,
      "reward": 5.460954070091248,
      "reward_std": 0.5135178118944168,
      "rewards/ngram_similarity_reward_func": 0.7478948384523392,
      "rewards/reasoning_quality_reward_func": 0.8943093717098236,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.068750023841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 308
    },
    {
      "completion_length": 266.0625,
      "epoch": 0.17024793388429751,
      "grad_norm": 0.2349492460489273,
      "kl": 0.03218721831217408,
      "learning_rate": 3.947519800531621e-05,
      "loss": 0.0013,
      "reward": 5.274180769920349,
      "reward_std": 0.38493969291448593,
      "rewards/ngram_similarity_reward_func": 0.7311684787273407,
      "rewards/reasoning_quality_reward_func": 0.8555124998092651,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.9375,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 309
    },
    {
      "completion_length": 327.65625,
      "epoch": 0.17079889807162535,
      "grad_norm": 0.18817688524723053,
      "kl": 0.02942902222275734,
      "learning_rate": 3.9469475212036664e-05,
      "loss": 0.0012,
      "reward": 5.482720971107483,
      "reward_std": 0.7104192487895489,
      "rewards/ngram_similarity_reward_func": 0.5867645293474197,
      "rewards/reasoning_quality_reward_func": 0.9522062391042709,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.193750023841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 310
    },
    {
      "completion_length": 273.125,
      "epoch": 0.17134986225895316,
      "grad_norm": 0.2287682145833969,
      "kl": 0.033470026683062315,
      "learning_rate": 3.94637218517926e-05,
      "loss": 0.0013,
      "reward": 5.091960847377777,
      "reward_std": 0.3252540584653616,
      "rewards/ngram_similarity_reward_func": 0.5948922485113144,
      "rewards/reasoning_quality_reward_func": 0.6845687329769135,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.0625,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 311
    },
    {
      "completion_length": 245.84375,
      "epoch": 0.171900826446281,
      "grad_norm": 0.22232896089553833,
      "kl": 0.03752465033903718,
      "learning_rate": 3.9457937934650864e-05,
      "loss": 0.0015,
      "reward": 5.6537628173828125,
      "reward_std": 0.6166419759392738,
      "rewards/ngram_similarity_reward_func": 0.6175289452075958,
      "rewards/reasoning_quality_reward_func": 0.8865812420845032,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.3996527194976807,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 312
    },
    {
      "completion_length": 324.53125,
      "epoch": 0.17245179063360883,
      "grad_norm": 0.20553645491600037,
      "kl": 0.026154049206525087,
      "learning_rate": 3.9452123470731775e-05,
      "loss": 0.001,
      "reward": 5.003131866455078,
      "reward_std": 0.9055781671777368,
      "rewards/ngram_similarity_reward_func": 0.528158463537693,
      "rewards/reasoning_quality_reward_func": 0.9619562476873398,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.763017237186432,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 313
    },
    {
      "completion_length": 336.65625,
      "epoch": 0.17300275482093663,
      "grad_norm": 0.19151301681995392,
      "kl": 0.026151685509830713,
      "learning_rate": 3.9446278470209126e-05,
      "loss": 0.001,
      "reward": 5.278534412384033,
      "reward_std": 0.7870542462915182,
      "rewards/ngram_similarity_reward_func": 0.5619845911860466,
      "rewards/reasoning_quality_reward_func": 0.8321749866008759,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.134375035762787,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 314
    },
    {
      "completion_length": 289.15625,
      "epoch": 0.17355371900826447,
      "grad_norm": 0.21619422733783722,
      "kl": 0.03077174024656415,
      "learning_rate": 3.9440402943310115e-05,
      "loss": 0.0012,
      "reward": 6.1457953453063965,
      "reward_std": 0.5268538296222687,
      "rewards/ngram_similarity_reward_func": 0.8656275272369385,
      "rewards/reasoning_quality_reward_func": 0.9536781013011932,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.5764898657798767,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 315
    },
    {
      "completion_length": 284.03125,
      "epoch": 0.17410468319559227,
      "grad_norm": 0.20744605362415314,
      "kl": 0.03749189153313637,
      "learning_rate": 3.9434496900315364e-05,
      "loss": 0.0015,
      "reward": 6.346541047096252,
      "reward_std": 0.1635672696866095,
      "rewards/ngram_similarity_reward_func": 0.7461725249886513,
      "rewards/reasoning_quality_reward_func": 0.8503687381744385,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 4.0,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 316
    },
    {
      "completion_length": 231.46875,
      "epoch": 0.1746556473829201,
      "grad_norm": 0.2660014033317566,
      "kl": 0.046677826438099146,
      "learning_rate": 3.942856035155888e-05,
      "loss": 0.0019,
      "reward": 5.241629242897034,
      "reward_std": 0.9536032751202583,
      "rewards/ngram_similarity_reward_func": 0.6909109801054001,
      "rewards/reasoning_quality_reward_func": 0.7422749996185303,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.0584434270858765,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 317
    },
    {
      "completion_length": 273.5625,
      "epoch": 0.17520661157024794,
      "grad_norm": 0.2176179140806198,
      "kl": 0.024531573057174683,
      "learning_rate": 3.9422593307428066e-05,
      "loss": 0.001,
      "reward": 6.355871677398682,
      "reward_std": 0.30343178287148476,
      "rewards/ngram_similarity_reward_func": 0.8523779511451721,
      "rewards/reasoning_quality_reward_func": 0.8003687411546707,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.953125,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 318
    },
    {
      "completion_length": 255.90625,
      "epoch": 0.17575757575757575,
      "grad_norm": 0.18894927203655243,
      "kl": 0.030739556066691875,
      "learning_rate": 3.941659577836366e-05,
      "loss": 0.0012,
      "reward": 5.919140338897705,
      "reward_std": 0.818994514644146,
      "rewards/ngram_similarity_reward_func": 0.7786872237920761,
      "rewards/reasoning_quality_reward_func": 0.8185781389474869,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.571875035762787,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 319
    },
    {
      "completion_length": 305.59375,
      "epoch": 0.1763085399449036,
      "grad_norm": 0.21975234150886536,
      "kl": 0.025410700123757124,
      "learning_rate": 3.941056777485976e-05,
      "loss": 0.001,
      "reward": 5.507596731185913,
      "reward_std": 1.1633977964520454,
      "rewards/ngram_similarity_reward_func": 0.6910500153899193,
      "rewards/reasoning_quality_reward_func": 0.7977968752384186,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.2687500715255737,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 320
    },
    {
      "completion_length": 294.625,
      "epoch": 0.1768595041322314,
      "grad_norm": 0.20554637908935547,
      "kl": 0.0322769028134644,
      "learning_rate": 3.940450930746377e-05,
      "loss": 0.0013,
      "reward": 4.946973979473114,
      "reward_std": 0.5494002737104893,
      "rewards/ngram_similarity_reward_func": 0.6856970638036728,
      "rewards/reasoning_quality_reward_func": 0.8580906242132187,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.6531862616539,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 321
    },
    {
      "completion_length": 256.21875,
      "epoch": 0.17741046831955923,
      "grad_norm": 0.2145739644765854,
      "kl": 0.03180500864982605,
      "learning_rate": 3.9398420386776426e-05,
      "loss": 0.0013,
      "reward": 5.552285552024841,
      "reward_std": 0.8173827175050974,
      "rewards/ngram_similarity_reward_func": 0.5058534406125546,
      "rewards/reasoning_quality_reward_func": 0.7332812547683716,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.563151001930237,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 322
    },
    {
      "completion_length": 279.125,
      "epoch": 0.17796143250688706,
      "grad_norm": 0.1939256340265274,
      "kl": 0.02843255316838622,
      "learning_rate": 3.939230102345171e-05,
      "loss": 0.0011,
      "reward": 5.2210052609443665,
      "reward_std": 0.5881230421364307,
      "rewards/ngram_similarity_reward_func": 0.5912159308791161,
      "rewards/reasoning_quality_reward_func": 0.876474991440773,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.0033143758773804,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 323
    },
    {
      "completion_length": 274.46875,
      "epoch": 0.17851239669421487,
      "grad_norm": 0.22554560005664825,
      "kl": 0.03546756645664573,
      "learning_rate": 3.9386151228196895e-05,
      "loss": 0.0014,
      "reward": 5.559608161449432,
      "reward_std": 0.581847544759512,
      "rewards/ngram_similarity_reward_func": 0.716680034995079,
      "rewards/reasoning_quality_reward_func": 0.8007405996322632,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.292187511920929,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 324
    },
    {
      "completion_length": 295.1875,
      "epoch": 0.1790633608815427,
      "grad_norm": 0.18014301359653473,
      "kl": 0.022343548480421305,
      "learning_rate": 3.9379971011772504e-05,
      "loss": 0.0009,
      "reward": 5.441017985343933,
      "reward_std": 0.2723664790391922,
      "rewards/ngram_similarity_reward_func": 0.6954012662172318,
      "rewards/reasoning_quality_reward_func": 0.9148875027894974,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.080729126930237,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 325
    },
    {
      "completion_length": 293.46875,
      "epoch": 0.1796143250688705,
      "grad_norm": 0.2187529355287552,
      "kl": 0.028804882895201445,
      "learning_rate": 3.937376038499228e-05,
      "loss": 0.0012,
      "reward": 4.726682066917419,
      "reward_std": 0.5115012377500534,
      "rewards/ngram_similarity_reward_func": 0.650241494178772,
      "rewards/reasoning_quality_reward_func": 0.8306999951601028,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.495740629732609,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 326
    },
    {
      "completion_length": 275.75,
      "epoch": 0.18016528925619835,
      "grad_norm": 0.22510570287704468,
      "kl": 0.03640979016199708,
      "learning_rate": 3.936751935872318e-05,
      "loss": 0.0015,
      "reward": 5.372348427772522,
      "reward_std": 0.4674986694008112,
      "rewards/ngram_similarity_reward_func": 0.550095409154892,
      "rewards/reasoning_quality_reward_func": 0.9128781408071518,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.159375011920929,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 327
    },
    {
      "completion_length": 281.78125,
      "epoch": 0.18071625344352618,
      "grad_norm": 0.20479747653007507,
      "kl": 0.033773055765777826,
      "learning_rate": 3.936124794388534e-05,
      "loss": 0.0014,
      "reward": 5.1518741846084595,
      "reward_std": 0.560208105482161,
      "rewards/ngram_similarity_reward_func": 0.6775991767644882,
      "rewards/reasoning_quality_reward_func": 0.8898999840021133,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.834375023841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 328
    },
    {
      "completion_length": 285.625,
      "epoch": 0.181267217630854,
      "grad_norm": 0.22793985903263092,
      "kl": 0.03366591269150376,
      "learning_rate": 3.935494615145209e-05,
      "loss": 0.0013,
      "reward": 6.059327960014343,
      "reward_std": 0.7773629277944565,
      "rewards/ngram_similarity_reward_func": 0.7025894075632095,
      "rewards/reasoning_quality_reward_func": 0.8931968659162521,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.713541626930237,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 329
    },
    {
      "completion_length": 336.8125,
      "epoch": 0.18181818181818182,
      "grad_norm": 0.1901523768901825,
      "kl": 0.032234815414994955,
      "learning_rate": 3.93486139924499e-05,
      "loss": 0.0013,
      "reward": 5.0375999212265015,
      "reward_std": 0.6576861925423145,
      "rewards/ngram_similarity_reward_func": 0.42715924978256226,
      "rewards/reasoning_quality_reward_func": 0.8424718827009201,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.021875023841858,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 330
    },
    {
      "completion_length": 235.65625,
      "epoch": 0.18236914600550963,
      "grad_norm": 0.24399752914905548,
      "kl": 0.05670465435832739,
      "learning_rate": 3.934225147795837e-05,
      "loss": 0.0023,
      "reward": 4.981407284736633,
      "reward_std": 0.3371787779033184,
      "rewards/ngram_similarity_reward_func": 0.8251384645700455,
      "rewards/reasoning_quality_reward_func": 0.8187687247991562,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.587500000372529,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 331
    },
    {
      "completion_length": 292.875,
      "epoch": 0.18292011019283747,
      "grad_norm": 0.20769080519676208,
      "kl": 0.03431240702047944,
      "learning_rate": 3.933585861911022e-05,
      "loss": 0.0014,
      "reward": 5.94089937210083,
      "reward_std": 0.5120038166642189,
      "rewards/ngram_similarity_reward_func": 0.6616713255643845,
      "rewards/reasoning_quality_reward_func": 0.9630531370639801,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.566175103187561,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 332
    },
    {
      "completion_length": 317.96875,
      "epoch": 0.1834710743801653,
      "grad_norm": 0.19837544858455658,
      "kl": 0.04345704289153218,
      "learning_rate": 3.932943542709127e-05,
      "loss": 0.0017,
      "reward": 5.87508225440979,
      "reward_std": 0.728335517924279,
      "rewards/ngram_similarity_reward_func": 0.6758104041218758,
      "rewards/reasoning_quality_reward_func": 0.8773968815803528,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.571875035762787,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 333
    },
    {
      "completion_length": 321.21875,
      "epoch": 0.1840220385674931,
      "grad_norm": 0.19171608984470367,
      "kl": 0.03467270638793707,
      "learning_rate": 3.93229819131404e-05,
      "loss": 0.0014,
      "reward": 4.748794198036194,
      "reward_std": 0.31168190762400627,
      "rewards/ngram_similarity_reward_func": 0.6374318674206734,
      "rewards/reasoning_quality_reward_func": 0.8902687430381775,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.474999999627471,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 334
    },
    {
      "completion_length": 326.21875,
      "epoch": 0.18457300275482094,
      "grad_norm": 0.24707810580730438,
      "kl": 0.033593188505619764,
      "learning_rate": 3.9316498088549546e-05,
      "loss": 0.0013,
      "reward": 5.005239486694336,
      "reward_std": 0.6308664847165346,
      "rewards/ngram_similarity_reward_func": 0.5010050013661385,
      "rewards/reasoning_quality_reward_func": 0.9448593854904175,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.825000047683716,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 335
    },
    {
      "completion_length": 333.78125,
      "epoch": 0.18512396694214875,
      "grad_norm": 0.2006770819425583,
      "kl": 0.03541389387100935,
      "learning_rate": 3.930998396466371e-05,
      "loss": 0.0014,
      "reward": 4.869897305965424,
      "reward_std": 0.8772668093442917,
      "rewards/ngram_similarity_reward_func": 0.6786605939269066,
      "rewards/reasoning_quality_reward_func": 0.9045999944210052,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.536636769771576,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 336
    },
    {
      "completion_length": 334.75,
      "epoch": 0.18567493112947658,
      "grad_norm": 0.23299890756607056,
      "kl": 0.03776941588148475,
      "learning_rate": 3.930343955288087e-05,
      "loss": 0.0015,
      "reward": 5.045654296875,
      "reward_std": 0.8416954539716244,
      "rewards/ngram_similarity_reward_func": 0.48145127668976784,
      "rewards/reasoning_quality_reward_func": 0.8571718782186508,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.96875,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 337
    },
    {
      "completion_length": 321.15625,
      "epoch": 0.18622589531680442,
      "grad_norm": 0.17136535048484802,
      "kl": 0.03944785753265023,
      "learning_rate": 3.929686486465203e-05,
      "loss": 0.0016,
      "reward": 5.567359685897827,
      "reward_std": 0.4021573094651103,
      "rewards/ngram_similarity_reward_func": 0.5492004677653313,
      "rewards/reasoning_quality_reward_func": 0.891181230545044,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.3769782185554504,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 338
    },
    {
      "completion_length": 297.90625,
      "epoch": 0.18677685950413223,
      "grad_norm": 0.19861724972724915,
      "kl": 0.04017520043998957,
      "learning_rate": 3.9290259911481136e-05,
      "loss": 0.0016,
      "reward": 6.412063717842102,
      "reward_std": 0.15601356606930494,
      "rewards/ngram_similarity_reward_func": 0.6923949718475342,
      "rewards/reasoning_quality_reward_func": 0.9696687310934067,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 4.0,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 339
    },
    {
      "completion_length": 383.53125,
      "epoch": 0.18732782369146006,
      "grad_norm": 0.20008857548236847,
      "kl": 0.03661731444299221,
      "learning_rate": 3.928362470492513e-05,
      "loss": 0.0015,
      "reward": 5.050732851028442,
      "reward_std": 0.48043219558894634,
      "rewards/ngram_similarity_reward_func": 0.45102448761463165,
      "rewards/reasoning_quality_reward_func": 0.8288749903440475,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.0208332538604736,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 340
    },
    {
      "completion_length": 359.46875,
      "epoch": 0.18787878787878787,
      "grad_norm": 0.19358529150485992,
      "kl": 0.04704639548435807,
      "learning_rate": 3.9276959256593864e-05,
      "loss": 0.0019,
      "reward": 6.217095971107483,
      "reward_std": 0.7373478394001722,
      "rewards/ngram_similarity_reward_func": 0.7406114935874939,
      "rewards/reasoning_quality_reward_func": 0.9639844000339508,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.762500047683716,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 341
    },
    {
      "completion_length": 311.21875,
      "epoch": 0.1884297520661157,
      "grad_norm": 0.21851791441440582,
      "kl": 0.05122498236596584,
      "learning_rate": 3.9270263578150105e-05,
      "loss": 0.002,
      "reward": 3.6670703291893005,
      "reward_std": 0.5597222158685327,
      "rewards/ngram_similarity_reward_func": 0.2372889369726181,
      "rewards/reasoning_quality_reward_func": 0.967281237244606,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 1.712500000372529,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 342
    },
    {
      "completion_length": 385.34375,
      "epoch": 0.18898071625344354,
      "grad_norm": 0.17612223327159882,
      "kl": 0.03961751610040665,
      "learning_rate": 3.926353768130954e-05,
      "loss": 0.0016,
      "reward": 4.985505223274231,
      "reward_std": 1.204222321510315,
      "rewards/ngram_similarity_reward_func": 0.6430400907993317,
      "rewards/reasoning_quality_reward_func": 0.7996249794960022,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.79284006357193,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 343
    },
    {
      "completion_length": 358.875,
      "epoch": 0.18953168044077134,
      "grad_norm": 0.20758967101573944,
      "kl": 0.03950031287968159,
      "learning_rate": 3.9256781577840674e-05,
      "loss": 0.0016,
      "reward": 5.803722977638245,
      "reward_std": 0.9824237041175365,
      "rewards/ngram_similarity_reward_func": 0.7456263303756714,
      "rewards/reasoning_quality_reward_func": 0.9909937530755997,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.3171029686927795,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 344
    },
    {
      "completion_length": 320.6875,
      "epoch": 0.19008264462809918,
      "grad_norm": 0.21947449445724487,
      "kl": 0.05133596807718277,
      "learning_rate": 3.9249995279564944e-05,
      "loss": 0.0021,
      "reward": 6.299055457115173,
      "reward_std": 0.4854021407663822,
      "rewards/ngram_similarity_reward_func": 0.7720273956656456,
      "rewards/reasoning_quality_reward_func": 0.8957781046628952,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.881250023841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 345
    },
    {
      "completion_length": 340.78125,
      "epoch": 0.19063360881542699,
      "grad_norm": 0.21150778234004974,
      "kl": 0.05103140790015459,
      "learning_rate": 3.924317879835655e-05,
      "loss": 0.002,
      "reward": 5.144864082336426,
      "reward_std": 0.9677173532545567,
      "rewards/ngram_similarity_reward_func": 0.6048543602228165,
      "rewards/reasoning_quality_reward_func": 0.8404562622308731,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.9495535492897034,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 346
    },
    {
      "completion_length": 354.0625,
      "epoch": 0.19118457300275482,
      "grad_norm": 0.20196419954299927,
      "kl": 0.04508020356297493,
      "learning_rate": 3.923633214614254e-05,
      "loss": 0.0018,
      "reward": 4.779169261455536,
      "reward_std": 0.4141131155192852,
      "rewards/ngram_similarity_reward_func": 0.677888035774231,
      "rewards/reasoning_quality_reward_func": 0.9512812346220016,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.400000000372529,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 347
    },
    {
      "completion_length": 360.6875,
      "epoch": 0.19173553719008266,
      "grad_norm": 0.22140376269817352,
      "kl": 0.045372942462563515,
      "learning_rate": 3.922945533490275e-05,
      "loss": 0.0018,
      "reward": 4.834069013595581,
      "reward_std": 0.2656398843973875,
      "rewards/ngram_similarity_reward_func": 0.6186202839016914,
      "rewards/reasoning_quality_reward_func": 0.8582687526941299,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 2.618898820132017,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 348
    },
    {
      "completion_length": 373.21875,
      "epoch": 0.19228650137741046,
      "grad_norm": 0.17201413214206696,
      "kl": 0.04705087933689356,
      "learning_rate": 3.922254837666978e-05,
      "loss": 0.0019,
      "reward": 4.586026132106781,
      "reward_std": 0.6061949357390404,
      "rewards/ngram_similarity_reward_func": 0.49902907758951187,
      "rewards/reasoning_quality_reward_func": 0.9830968677997589,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.353900223970413,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 349
    },
    {
      "completion_length": 251.125,
      "epoch": 0.1928374655647383,
      "grad_norm": 0.2287546843290329,
      "kl": 0.07076171319931746,
      "learning_rate": 3.9215611283528986e-05,
      "loss": 0.0028,
      "reward": 6.265152931213379,
      "reward_std": 0.5510429758578539,
      "rewards/ngram_similarity_reward_func": 0.8625842481851578,
      "rewards/reasoning_quality_reward_func": 0.9025687277317047,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.75,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 350
    },
    {
      "completion_length": 363.375,
      "epoch": 0.1933884297520661,
      "grad_norm": 0.18959863483905792,
      "kl": 0.04837387427687645,
      "learning_rate": 3.920864406761844e-05,
      "loss": 0.0019,
      "reward": 4.867045998573303,
      "reward_std": 0.14322672551497817,
      "rewards/ngram_similarity_reward_func": 0.6028896346688271,
      "rewards/reasoning_quality_reward_func": 0.9891562461853027,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.525000000372529,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 351
    },
    {
      "completion_length": 399.28125,
      "epoch": 0.19393939393939394,
      "grad_norm": 0.18078306317329407,
      "kl": 0.03721154620870948,
      "learning_rate": 3.920164674112893e-05,
      "loss": 0.0015,
      "reward": 3.752653956413269,
      "reward_std": 0.6587705090641975,
      "rewards/ngram_similarity_reward_func": 0.48326437920331955,
      "rewards/reasoning_quality_reward_func": 0.9825437515974045,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 1.5368459466844797,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 352
    },
    {
      "completion_length": 425.84375,
      "epoch": 0.19449035812672177,
      "grad_norm": 0.18134203553199768,
      "kl": 0.03997080260887742,
      "learning_rate": 3.919461931630393e-05,
      "loss": 0.0016,
      "reward": 5.6803165674209595,
      "reward_std": 0.3771057538688183,
      "rewards/ngram_similarity_reward_func": 0.6010917276144028,
      "rewards/reasoning_quality_reward_func": 0.9542249888181686,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.375,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 353
    },
    {
      "completion_length": 397.78125,
      "epoch": 0.19504132231404958,
      "grad_norm": 0.19565853476524353,
      "kl": 0.055060542188584805,
      "learning_rate": 3.918756180543958e-05,
      "loss": 0.0022,
      "reward": 5.114490866661072,
      "reward_std": 0.6970453634858131,
      "rewards/ngram_similarity_reward_func": 0.5895409043878317,
      "rewards/reasoning_quality_reward_func": 0.8663562536239624,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.9085937440395355,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 354
    },
    {
      "completion_length": 465.40625,
      "epoch": 0.19559228650137742,
      "grad_norm": 0.17450600862503052,
      "kl": 0.03273481037467718,
      "learning_rate": 3.918047422088465e-05,
      "loss": 0.0013,
      "reward": 3.84347927570343,
      "reward_std": 0.7384940292686224,
      "rewards/ngram_similarity_reward_func": 0.3972689136862755,
      "rewards/reasoning_quality_reward_func": 0.9557062685489655,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 1.7405042052268982,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 355
    },
    {
      "completion_length": 355.03125,
      "epoch": 0.19614325068870522,
      "grad_norm": 0.1950598806142807,
      "kl": 0.0467569837346673,
      "learning_rate": 3.917335657504054e-05,
      "loss": 0.0019,
      "reward": 6.258630156517029,
      "reward_std": 0.6395473852753639,
      "rewards/ngram_similarity_reward_func": 0.7815356552600861,
      "rewards/reasoning_quality_reward_func": 0.9516656249761581,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.7754289507865906,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 356
    },
    {
      "completion_length": 432.1875,
      "epoch": 0.19669421487603306,
      "grad_norm": 0.1837594360113144,
      "kl": 0.040499878115952015,
      "learning_rate": 3.9166208880361254e-05,
      "loss": 0.0016,
      "reward": 4.799482107162476,
      "reward_std": 0.26129591232165694,
      "rewards/ngram_similarity_reward_func": 0.6680446825921535,
      "rewards/reasoning_quality_reward_func": 0.9939374923706055,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.387499988079071,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 357
    },
    {
      "completion_length": 390.5625,
      "epoch": 0.1972451790633609,
      "grad_norm": 0.184334397315979,
      "kl": 0.04379983339458704,
      "learning_rate": 3.915903114935337e-05,
      "loss": 0.0018,
      "reward": 5.920761704444885,
      "reward_std": 0.545456875115633,
      "rewards/ngram_similarity_reward_func": 0.6328606233000755,
      "rewards/reasoning_quality_reward_func": 0.9884218722581863,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.5494791865348816,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 358
    },
    {
      "completion_length": 340.28125,
      "epoch": 0.1977961432506887,
      "grad_norm": 0.18676677346229553,
      "kl": 0.057850477285683155,
      "learning_rate": 3.915182339457602e-05,
      "loss": 0.0023,
      "reward": 5.79577112197876,
      "reward_std": 0.41157449036836624,
      "rewards/ngram_similarity_reward_func": 0.8389680087566376,
      "rewards/reasoning_quality_reward_func": 0.9568031281232834,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.25,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 359
    },
    {
      "completion_length": 342.90625,
      "epoch": 0.19834710743801653,
      "grad_norm": 0.2628999352455139,
      "kl": 0.06320911832153797,
      "learning_rate": 3.914458562864087e-05,
      "loss": 0.0025,
      "reward": 5.697378993034363,
      "reward_std": 0.4975591003894806,
      "rewards/ngram_similarity_reward_func": 0.6625093370676041,
      "rewards/reasoning_quality_reward_func": 0.9588281363248825,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.3260416388511658,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 360
    },
    {
      "completion_length": 380.1875,
      "epoch": 0.19889807162534434,
      "grad_norm": 0.18406622111797333,
      "kl": 0.04535065311938524,
      "learning_rate": 3.9137317864212085e-05,
      "loss": 0.0018,
      "reward": 5.663205504417419,
      "reward_std": 0.4870543070137501,
      "rewards/ngram_similarity_reward_func": 0.7245699018239975,
      "rewards/reasoning_quality_reward_func": 0.9301468729972839,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.258488714694977,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 361
    },
    {
      "completion_length": 378.03125,
      "epoch": 0.19944903581267218,
      "grad_norm": 0.2156660109758377,
      "kl": 0.049066536128520966,
      "learning_rate": 3.9130020114006346e-05,
      "loss": 0.002,
      "reward": 5.447566390037537,
      "reward_std": 0.8246756792068481,
      "rewards/ngram_similarity_reward_func": 0.6856861710548401,
      "rewards/reasoning_quality_reward_func": 0.8683905899524689,
      "rewards/soft_format_reward_func": 0.2421875,
      "rewards/sql_execution_reward_func": 3.1552083492279053,
      "rewards/xmlcount_reward_func": 0.49609375,
      "step": 362
    },
    {
      "completion_length": 288.59375,
      "epoch": 0.2,
      "grad_norm": 0.22221054136753082,
      "kl": 0.07095225155353546,
      "learning_rate": 3.912269239079279e-05,
      "loss": 0.0028,
      "reward": 5.849725604057312,
      "reward_std": 0.2196935759857297,
      "rewards/ngram_similarity_reward_func": 0.6743568815290928,
      "rewards/reasoning_quality_reward_func": 0.9253687411546707,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.5,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 363
    },
    {
      "completion_length": 432.9375,
      "epoch": 0.20055096418732782,
      "grad_norm": 0.17801330983638763,
      "kl": 0.040391040965914726,
      "learning_rate": 3.911533470739298e-05,
      "loss": 0.0016,
      "reward": 5.935391187667847,
      "reward_std": 0.4171575717628002,
      "rewards/ngram_similarity_reward_func": 0.5597088783979416,
      "rewards/reasoning_quality_reward_func": 0.8340156078338623,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.791666626930237,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 364
    },
    {
      "completion_length": 314.9375,
      "epoch": 0.20110192837465565,
      "grad_norm": 0.19523920118808746,
      "kl": 0.06868137512356043,
      "learning_rate": 3.910794707668093e-05,
      "loss": 0.0027,
      "reward": 5.861039400100708,
      "reward_std": 0.5307717956602573,
      "rewards/ngram_similarity_reward_func": 0.6104863584041595,
      "rewards/reasoning_quality_reward_func": 0.9380531013011932,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.5625,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 365
    },
    {
      "completion_length": 391.03125,
      "epoch": 0.20165289256198346,
      "grad_norm": 0.18844497203826904,
      "kl": 0.04414760787039995,
      "learning_rate": 3.910052951158304e-05,
      "loss": 0.0018,
      "reward": 5.342592835426331,
      "reward_std": 0.9539203234016895,
      "rewards/ngram_similarity_reward_func": 0.41007253900170326,
      "rewards/reasoning_quality_reward_func": 0.9211406111717224,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.261379897594452,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 366
    },
    {
      "completion_length": 354.3125,
      "epoch": 0.2022038567493113,
      "grad_norm": 0.20918360352516174,
      "kl": 0.05260260682553053,
      "learning_rate": 3.909308202507808e-05,
      "loss": 0.0021,
      "reward": 5.627278804779053,
      "reward_std": 0.8334615528583527,
      "rewards/ngram_similarity_reward_func": 0.5731283165514469,
      "rewards/reasoning_quality_reward_func": 0.9694937318563461,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.334656774997711,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 367
    },
    {
      "completion_length": 269.625,
      "epoch": 0.20275482093663913,
      "grad_norm": 0.2272816002368927,
      "kl": 0.06738220155239105,
      "learning_rate": 3.90856046301972e-05,
      "loss": 0.0027,
      "reward": 5.93574583530426,
      "reward_std": 0.28085143584758043,
      "rewards/ngram_similarity_reward_func": 0.8712146580219269,
      "rewards/reasoning_quality_reward_func": 0.8770312517881393,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.4375,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 368
    },
    {
      "completion_length": 356.9375,
      "epoch": 0.20330578512396694,
      "grad_norm": 0.19965340197086334,
      "kl": 0.05732113216072321,
      "learning_rate": 3.9078097340023846e-05,
      "loss": 0.0023,
      "reward": 5.539290904998779,
      "reward_std": 0.8564386926591396,
      "rewards/ngram_similarity_reward_func": 0.7391766905784607,
      "rewards/reasoning_quality_reward_func": 0.9593874961137772,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.0907267034053802,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 369
    },
    {
      "completion_length": 376.28125,
      "epoch": 0.20385674931129477,
      "grad_norm": 0.20409658551216125,
      "kl": 0.055789174512028694,
      "learning_rate": 3.9070560167693796e-05,
      "loss": 0.0022,
      "reward": 6.143639087677002,
      "reward_std": 0.757773332297802,
      "rewards/ngram_similarity_reward_func": 0.798764705657959,
      "rewards/reasoning_quality_reward_func": 0.8943062573671341,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.700568199157715,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 370
    },
    {
      "completion_length": 375.0625,
      "epoch": 0.20440771349862258,
      "grad_norm": 0.18150664865970612,
      "kl": 0.052613090723752975,
      "learning_rate": 3.9062993126395127e-05,
      "loss": 0.0021,
      "reward": 5.779634237289429,
      "reward_std": 0.5700815715827048,
      "rewards/ngram_similarity_reward_func": 0.4929562471807003,
      "rewards/reasoning_quality_reward_func": 0.9038656204938889,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.6328125,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 371
    },
    {
      "completion_length": 337.96875,
      "epoch": 0.2049586776859504,
      "grad_norm": 0.20182238519191742,
      "kl": 0.06167987547814846,
      "learning_rate": 3.905539622936814e-05,
      "loss": 0.0025,
      "reward": 4.734959423542023,
      "reward_std": 0.3173269657418132,
      "rewards/ngram_similarity_reward_func": 0.4699687883257866,
      "rewards/reasoning_quality_reward_func": 0.8446781635284424,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.670312523841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 372
    },
    {
      "completion_length": 349.15625,
      "epoch": 0.20550964187327825,
      "grad_norm": 0.22501292824745178,
      "kl": 0.060175278224051,
      "learning_rate": 3.90477694899054e-05,
      "loss": 0.0024,
      "reward": 4.837795376777649,
      "reward_std": 0.4778463989496231,
      "rewards/ngram_similarity_reward_func": 0.747282363474369,
      "rewards/reasoning_quality_reward_func": 0.9575375020503998,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.382975697517395,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 373
    },
    {
      "completion_length": 363.84375,
      "epoch": 0.20606060606060606,
      "grad_norm": 0.20730702579021454,
      "kl": 0.07064731512218714,
      "learning_rate": 3.904011292135169e-05,
      "loss": 0.0028,
      "reward": 5.595228493213654,
      "reward_std": 0.09934269823133945,
      "rewards/ngram_similarity_reward_func": 0.4779411293566227,
      "rewards/reasoning_quality_reward_func": 0.8672874867916107,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.5,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 374
    },
    {
      "completion_length": 367.15625,
      "epoch": 0.2066115702479339,
      "grad_norm": 0.20052236318588257,
      "kl": 0.064847307279706,
      "learning_rate": 3.903242653710399e-05,
      "loss": 0.0026,
      "reward": 4.825861275196075,
      "reward_std": 0.5914509408175945,
      "rewards/ngram_similarity_reward_func": 0.6120738461613655,
      "rewards/reasoning_quality_reward_func": 0.9169124960899353,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.546875,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 375
    },
    {
      "completion_length": 310.5,
      "epoch": 0.2071625344352617,
      "grad_norm": 0.20714934170246124,
      "kl": 0.0655839079990983,
      "learning_rate": 3.9024710350611426e-05,
      "loss": 0.0026,
      "reward": 5.938927412033081,
      "reward_std": 0.910742374137044,
      "rewards/ngram_similarity_reward_func": 0.8777117431163788,
      "rewards/reasoning_quality_reward_func": 0.9862156063318253,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.325000047683716,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 376
    },
    {
      "completion_length": 360.03125,
      "epoch": 0.20771349862258953,
      "grad_norm": 0.19089394807815552,
      "kl": 0.06358931493014097,
      "learning_rate": 3.901696437537529e-05,
      "loss": 0.0025,
      "reward": 4.880189657211304,
      "reward_std": 0.5088448598980904,
      "rewards/ngram_similarity_reward_func": 0.6904771476984024,
      "rewards/reasoning_quality_reward_func": 0.8803374916315079,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.559374988079071,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 377
    },
    {
      "completion_length": 340.59375,
      "epoch": 0.20826446280991737,
      "grad_norm": 0.22090977430343628,
      "kl": 0.061226640827953815,
      "learning_rate": 3.9009188624949e-05,
      "loss": 0.0024,
      "reward": 4.684970140457153,
      "reward_std": 0.656653918325901,
      "rewards/ngram_similarity_reward_func": 0.6771039515733719,
      "rewards/reasoning_quality_reward_func": 0.9259218573570251,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.331944465637207,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 378
    },
    {
      "completion_length": 291.03125,
      "epoch": 0.20881542699724517,
      "grad_norm": 0.22363604605197906,
      "kl": 0.07800731528550386,
      "learning_rate": 3.900138311293805e-05,
      "loss": 0.0031,
      "reward": 5.534857392311096,
      "reward_std": 0.4977914020419121,
      "rewards/ngram_similarity_reward_func": 0.6027723550796509,
      "rewards/reasoning_quality_reward_func": 0.8819874972105026,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.3000975251197815,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 379
    },
    {
      "completion_length": 336.75,
      "epoch": 0.209366391184573,
      "grad_norm": 0.219398632645607,
      "kl": 0.07098503038287163,
      "learning_rate": 3.899354785300003e-05,
      "loss": 0.0028,
      "reward": 4.709400713443756,
      "reward_std": 1.0458696587011218,
      "rewards/ngram_similarity_reward_func": 0.6132600903511047,
      "rewards/reasoning_quality_reward_func": 0.9242656081914902,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.421874985098839,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 380
    },
    {
      "completion_length": 342.4375,
      "epoch": 0.20991735537190082,
      "grad_norm": 0.23316411674022675,
      "kl": 0.08164038509130478,
      "learning_rate": 3.8985682858844584e-05,
      "loss": 0.0033,
      "reward": 4.354842185974121,
      "reward_std": 0.628161758184433,
      "rewards/ngram_similarity_reward_func": 0.574882909655571,
      "rewards/reasoning_quality_reward_func": 0.8612093478441238,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.1687500029802322,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 381
    },
    {
      "completion_length": 366.0625,
      "epoch": 0.21046831955922865,
      "grad_norm": 0.2040286809206009,
      "kl": 0.06415698956698179,
      "learning_rate": 3.897778814423336e-05,
      "loss": 0.0026,
      "reward": 5.802699327468872,
      "reward_std": 0.36272737942636013,
      "rewards/ngram_similarity_reward_func": 0.7063713818788528,
      "rewards/reasoning_quality_reward_func": 0.9713281244039536,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.375,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 382
    },
    {
      "completion_length": 286.34375,
      "epoch": 0.21101928374655649,
      "grad_norm": 0.21501082181930542,
      "kl": 0.07515604421496391,
      "learning_rate": 3.896986372298001e-05,
      "loss": 0.003,
      "reward": 6.546187996864319,
      "reward_std": 0.2610908932983875,
      "rewards/ngram_similarity_reward_func": 0.8259568363428116,
      "rewards/reasoning_quality_reward_func": 0.9702312499284744,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 4.0,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 383
    },
    {
      "completion_length": 296.375,
      "epoch": 0.2115702479338843,
      "grad_norm": 0.23275244235992432,
      "kl": 0.06781669799238443,
      "learning_rate": 3.896190960895021e-05,
      "loss": 0.0027,
      "reward": 5.380893170833588,
      "reward_std": 0.6650022715330124,
      "rewards/ngram_similarity_reward_func": 0.5775744169950485,
      "rewards/reasoning_quality_reward_func": 0.8564437329769135,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.1968750059604645,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 384
    },
    {
      "completion_length": 334.53125,
      "epoch": 0.21212121212121213,
      "grad_norm": 0.1849033534526825,
      "kl": 0.06413618009537458,
      "learning_rate": 3.895392581606153e-05,
      "loss": 0.0026,
      "reward": 6.0840959548950195,
      "reward_std": 0.5859294245019555,
      "rewards/ngram_similarity_reward_func": 0.712877668440342,
      "rewards/reasoning_quality_reward_func": 0.9988968670368195,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.6223214864730835,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 385
    },
    {
      "completion_length": 315.0,
      "epoch": 0.21267217630853993,
      "grad_norm": 0.210651695728302,
      "kl": 0.06734344735741615,
      "learning_rate": 3.894591235828351e-05,
      "loss": 0.0027,
      "reward": 4.874952614307404,
      "reward_std": 0.47136741131544113,
      "rewards/ngram_similarity_reward_func": 0.6175870895385742,
      "rewards/reasoning_quality_reward_func": 0.9167406260967255,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.590624999254942,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 386
    },
    {
      "completion_length": 270.78125,
      "epoch": 0.21322314049586777,
      "grad_norm": 0.23462246358394623,
      "kl": 0.07329368218779564,
      "learning_rate": 3.893786924963758e-05,
      "loss": 0.0029,
      "reward": 5.433947563171387,
      "reward_std": 1.204606607556343,
      "rewards/ngram_similarity_reward_func": 0.8317920863628387,
      "rewards/reasoning_quality_reward_func": 0.9950374960899353,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.85711807012558,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 387
    },
    {
      "completion_length": 331.9375,
      "epoch": 0.2137741046831956,
      "grad_norm": 0.21732455492019653,
      "kl": 0.06645281240344048,
      "learning_rate": 3.892979650419705e-05,
      "loss": 0.0027,
      "reward": 5.26733934879303,
      "reward_std": 0.9974728748202324,
      "rewards/ngram_similarity_reward_func": 0.6776885911822319,
      "rewards/reasoning_quality_reward_func": 0.9417343586683273,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.8979166746139526,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 388
    },
    {
      "completion_length": 409.8125,
      "epoch": 0.2143250688705234,
      "grad_norm": 0.21375003457069397,
      "kl": 0.052268791012465954,
      "learning_rate": 3.89216941360871e-05,
      "loss": 0.0021,
      "reward": 5.502641677856445,
      "reward_std": 0.7043182365596294,
      "rewards/ngram_similarity_reward_func": 0.7118322998285294,
      "rewards/reasoning_quality_reward_func": 0.9783093631267548,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.0625,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 389
    },
    {
      "completion_length": 375.0,
      "epoch": 0.21487603305785125,
      "grad_norm": 0.21334277093410492,
      "kl": 0.06060289219021797,
      "learning_rate": 3.8913562159484726e-05,
      "loss": 0.0024,
      "reward": 5.219146907329559,
      "reward_std": 0.4498859532177448,
      "rewards/ngram_similarity_reward_func": 0.6608686298131943,
      "rewards/reasoning_quality_reward_func": 0.9207781255245209,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.887499988079071,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 390
    },
    {
      "completion_length": 331.5625,
      "epoch": 0.21542699724517905,
      "grad_norm": 0.19862313568592072,
      "kl": 0.05955408047884703,
      "learning_rate": 3.890540058861874e-05,
      "loss": 0.0024,
      "reward": 5.310173511505127,
      "reward_std": 0.7558363061398268,
      "rewards/ngram_similarity_reward_func": 0.7412958741188049,
      "rewards/reasoning_quality_reward_func": 0.7965093553066254,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.0223684310913086,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 391
    },
    {
      "completion_length": 266.65625,
      "epoch": 0.2159779614325069,
      "grad_norm": 0.20900297164916992,
      "kl": 0.06220496632158756,
      "learning_rate": 3.8897209437769736e-05,
      "loss": 0.0025,
      "reward": 6.1704007387161255,
      "reward_std": 0.608038567006588,
      "rewards/ngram_similarity_reward_func": 0.8892379701137543,
      "rewards/reasoning_quality_reward_func": 0.9764750003814697,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.5625,
      "rewards/xmlcount_reward_func": 0.4921875,
      "step": 392
    },
    {
      "completion_length": 337.09375,
      "epoch": 0.21652892561983472,
      "grad_norm": 0.20373649895191193,
      "kl": 0.05620069894939661,
      "learning_rate": 3.888898872127006e-05,
      "loss": 0.0022,
      "reward": 6.051264643669128,
      "reward_std": 0.6935210712254047,
      "rewards/ngram_similarity_reward_func": 0.7540792673826218,
      "rewards/reasoning_quality_reward_func": 0.9680187404155731,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.5791666507720947,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 393
    },
    {
      "completion_length": 312.625,
      "epoch": 0.21707988980716253,
      "grad_norm": 0.2051205188035965,
      "kl": 0.059074556455016136,
      "learning_rate": 3.888073845350379e-05,
      "loss": 0.0024,
      "reward": 6.2320274114608765,
      "reward_std": 0.4816967062652111,
      "rewards/ngram_similarity_reward_func": 0.831011675298214,
      "rewards/reasoning_quality_reward_func": 0.9869531393051147,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.6640625,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 394
    },
    {
      "completion_length": 352.0625,
      "epoch": 0.21763085399449036,
      "grad_norm": 0.2099621444940567,
      "kl": 0.055240100249648094,
      "learning_rate": 3.8872458648906726e-05,
      "loss": 0.0022,
      "reward": 5.583946108818054,
      "reward_std": 0.9926156178116798,
      "rewards/ngram_similarity_reward_func": 0.6969510465860367,
      "rewards/reasoning_quality_reward_func": 0.8680187165737152,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.268976390361786,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 395
    },
    {
      "completion_length": 366.34375,
      "epoch": 0.21818181818181817,
      "grad_norm": 0.20218908786773682,
      "kl": 0.04617449454963207,
      "learning_rate": 3.8864149321966324e-05,
      "loss": 0.0018,
      "reward": 5.054450631141663,
      "reward_std": 0.9701939821243286,
      "rewards/ngram_similarity_reward_func": 0.522291287779808,
      "rewards/reasoning_quality_reward_func": 0.8852843642234802,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.896875023841858,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 396
    },
    {
      "completion_length": 301.84375,
      "epoch": 0.218732782369146,
      "grad_norm": 0.21052168309688568,
      "kl": 0.06285379361361265,
      "learning_rate": 3.885581048722172e-05,
      "loss": 0.0025,
      "reward": 5.028801798820496,
      "reward_std": 0.3098503816872835,
      "rewards/ngram_similarity_reward_func": 0.7622700333595276,
      "rewards/reasoning_quality_reward_func": 0.9270281195640564,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 2.5895034670829773,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 397
    },
    {
      "completion_length": 315.59375,
      "epoch": 0.21928374655647384,
      "grad_norm": 0.2132742553949356,
      "kl": 0.06167557183653116,
      "learning_rate": 3.8847442159263666e-05,
      "loss": 0.0025,
      "reward": 6.2307208776474,
      "reward_std": 0.503411186626181,
      "rewards/ngram_similarity_reward_func": 0.7272240668535233,
      "rewards/reasoning_quality_reward_func": 0.9909968823194504,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.762499988079071,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 398
    },
    {
      "completion_length": 282.03125,
      "epoch": 0.21983471074380165,
      "grad_norm": 0.19571053981781006,
      "kl": 0.06779203563928604,
      "learning_rate": 3.8839044352734524e-05,
      "loss": 0.0027,
      "reward": 5.848416328430176,
      "reward_std": 0.18905198317952454,
      "rewards/ngram_similarity_reward_func": 0.5929883420467377,
      "rewards/reasoning_quality_reward_func": 0.9351156204938889,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 3.5703125,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 399
    },
    {
      "completion_length": 320.28125,
      "epoch": 0.22038567493112948,
      "grad_norm": 0.1904131919145584,
      "kl": 0.066379071213305,
      "learning_rate": 3.883061708232823e-05,
      "loss": 0.0027,
      "reward": 4.2958656549453735,
      "reward_std": 0.7416478954255581,
      "rewards/ngram_similarity_reward_func": 0.9120374917984009,
      "rewards/reasoning_quality_reward_func": 0.8900781273841858,
      "rewards/soft_format_reward_func": 0.25,
      "rewards/sql_execution_reward_func": 1.743750024586916,
      "rewards/xmlcount_reward_func": 0.5,
      "step": 400
    }
  ],
  "logging_steps": 1,
  "max_steps": 2500,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 2,
  "save_steps": 100,
  "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": 8,
  "trial_name": null,
  "trial_params": null
}