diff --git "a/trainer_state.json" "b/trainer_state.json"
deleted file mode 100644--- "a/trainer_state.json"
+++ /dev/null
@@ -1,12384 +0,0 @@
-{
-    "max_steps": 1125,
-    "num_train_epochs": 1,
-    "log_history": [
-        {
-            "loss": 2.3515895009040833e-06,
-            "reward_mean": 0.7653523087501526,
-            "reward_var": 0.06265226006507874,
-            "avg_think_len": 13.406991958618164,
-            "reward_format_mean": 0.9993489384651184,
-            "reward_time_mean": 0.8125,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.6558104157447815,
-            "step": 0
-        },
-        {
-            "loss": 1.6706217138562351e-06,
-            "reward_mean": 0.7473169565200806,
-            "reward_var": 0.04913057014346123,
-            "avg_think_len": 12.394335746765137,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8374999761581421,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.6205179691314697,
-            "step": 1
-        },
-        {
-            "loss": 1.6912235878407955e-06,
-            "reward_mean": 0.7134044766426086,
-            "reward_var": 0.06620416045188904,
-            "avg_think_len": 11.78555965423584,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.715624988079071,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.5598680377006531,
-            "step": 2
-        },
-        {
-            "loss": 1.823529601097107e-06,
-            "reward_mean": 0.7772846221923828,
-            "reward_var": 0.04757240042090416,
-            "avg_think_len": 13.670331001281738,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.84375,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.6872633695602417,
-            "step": 3
-        },
-        {
-            "loss": 1.8807477317750454e-06,
-            "reward_mean": 0.7286001443862915,
-            "reward_var": 0.04852328076958656,
-            "avg_think_len": 13.854083061218262,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.84375,
-            "reward_correctness_mean": 0.40625,
-            "reward_response_efficiency_mean": 0.6644003987312317,
-            "step": 4
-        },
-        {
-            "loss": 2.5334302335977554e-06,
-            "reward_mean": 0.7333999276161194,
-            "reward_var": 0.07318412512540817,
-            "avg_think_len": 14.393296241760254,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.7281249761581421,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.6117244362831116,
-            "step": 5
-        },
-        {
-            "loss": 2.1229498088359833e-06,
-            "reward_mean": 0.7001112699508667,
-            "reward_var": 0.0722714364528656,
-            "avg_think_len": 13.535761833190918,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.778124988079071,
-            "reward_correctness_mean": 0.421875,
-            "reward_response_efficiency_mean": 0.6004449129104614,
-            "step": 6
-        },
-        {
-            "loss": 2.1795276552438736e-06,
-            "reward_mean": 0.729205846786499,
-            "reward_var": 0.054549455642700195,
-            "avg_think_len": 11.576457977294922,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.784375011920929,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.5074485540390015,
-            "step": 7
-        },
-        {
-            "loss": 1.7225975170731544e-06,
-            "reward_mean": 0.6939795613288879,
-            "reward_var": 0.06416602432727814,
-            "avg_think_len": 12.338702201843262,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.7749999761581421,
-            "reward_correctness_mean": 0.40625,
-            "reward_response_efficiency_mean": 0.5946683883666992,
-            "step": 8
-        },
-        {
-            "loss": 1.65269011631608e-06,
-            "reward_mean": 0.776672899723053,
-            "reward_var": 0.03608284145593643,
-            "avg_think_len": 12.051597595214844,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8843749761581421,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.6285666823387146,
-            "step": 9
-        },
-        {
-            "loss": 2.1009327610954642e-06,
-            "reward_mean": 0.802477240562439,
-            "reward_var": 0.037204064428806305,
-            "avg_think_len": 13.640829086303711,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8999999761581421,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.6380338668823242,
-            "step": 10
-        },
-        {
-            "loss": 1.9243452697992325e-06,
-            "reward_mean": 0.7232353687286377,
-            "reward_var": 0.07444832473993301,
-            "avg_think_len": 12.317909240722656,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.765625,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.5023165941238403,
-            "step": 11
-        },
-        {
-            "loss": 1.7933780327439308e-06,
-            "reward_mean": 0.666974663734436,
-            "reward_var": 0.06747563183307648,
-            "avg_think_len": 13.285401344299316,
-            "reward_format_mean": 0.9986979365348816,
-            "reward_time_mean": 0.6906249523162842,
-            "reward_correctness_mean": 0.4375,
-            "reward_response_efficiency_mean": 0.5410756468772888,
-            "step": 12
-        },
-        {
-            "loss": 1.5980331227183342e-06,
-            "reward_mean": 0.7483532428741455,
-            "reward_var": 0.07305104285478592,
-            "avg_think_len": 11.348151206970215,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.78125,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.5402880907058716,
-            "step": 13
-        },
-        {
-            "loss": 2.2884923964738846e-06,
-            "reward_mean": 0.7830677032470703,
-            "reward_var": 0.014842710457742214,
-            "avg_think_len": 12.698305130004883,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.925000011920929,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.6447707414627075,
-            "step": 14
-        },
-        {
-            "loss": 2.338085323572159e-06,
-            "reward_mean": 0.7507103681564331,
-            "reward_var": 0.029596764594316483,
-            "avg_think_len": 13.557746887207031,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8499999642372131,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.6528413891792297,
-            "step": 15
-        },
-        {
-            "loss": 1.8340651877224445e-06,
-            "reward_mean": 0.7867881655693054,
-            "reward_var": 0.05140697583556175,
-            "avg_think_len": 14.726365089416504,
-            "reward_format_mean": 0.9989583492279053,
-            "reward_time_mean": 0.8656249642372131,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.704444169998169,
-            "step": 16
-        },
-        {
-            "loss": 1.8710270524024963e-06,
-            "reward_mean": 0.802584171295166,
-            "reward_var": 0.04530316963791847,
-            "avg_think_len": 11.859588623046875,
-            "reward_format_mean": 0.9990234375,
-            "reward_time_mean": 0.887499988079071,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.6206881999969482,
-            "step": 17
-        },
-        {
-            "loss": 1.5800469554960728e-06,
-            "reward_mean": 0.6930772662162781,
-            "reward_var": 0.06501504778862,
-            "avg_think_len": 12.933457374572754,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.768750011920929,
-            "reward_correctness_mean": 0.40625,
-            "reward_response_efficiency_mean": 0.5973090529441833,
-            "step": 18
-        },
-        {
-            "loss": 1.5998957678675652e-06,
-            "reward_mean": 0.7092016339302063,
-            "reward_var": 0.06702180951833725,
-            "avg_think_len": 13.914103507995605,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.7437500357627869,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.6243066787719727,
-            "step": 19
-        },
-        {
-            "loss": 1.5909317880868912e-06,
-            "reward_mean": 0.6864709854125977,
-            "reward_var": 0.05510655418038368,
-            "avg_think_len": 11.540730476379395,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8343750238418579,
-            "reward_correctness_mean": 0.359375,
-            "reward_response_efficiency_mean": 0.5521338582038879,
-            "step": 20
-        },
-        {
-            "loss": 1.8083956092596054e-06,
-            "reward_mean": 0.7781481146812439,
-            "reward_var": 0.03408136963844299,
-            "avg_think_len": 11.965822219848633,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.6282173991203308,
-            "step": 21
-        },
-        {
-            "loss": 1.9381986930966377e-06,
-            "reward_mean": 0.6761418581008911,
-            "reward_var": 0.06331852823495865,
-            "avg_think_len": 14.609770774841309,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.75,
-            "reward_correctness_mean": 0.296875,
-            "reward_response_efficiency_mean": 0.6576924920082092,
-            "step": 22
-        },
-        {
-            "loss": 1.7248094081878662e-06,
-            "reward_mean": 0.7607037425041199,
-            "reward_var": 0.0760660246014595,
-            "avg_think_len": 13.108450889587402,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8031250238418579,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.5990650057792664,
-            "step": 23
-        },
-        {
-            "loss": 1.5497207641601562e-06,
-            "reward_mean": 0.6997513175010681,
-            "reward_var": 0.055371351540088654,
-            "avg_think_len": 13.587337493896484,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.75,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.580255389213562,
-            "step": 24
-        },
-        {
-            "loss": 1.944601535797119e-06,
-            "reward_mean": 0.7396306395530701,
-            "reward_var": 0.0738365575671196,
-            "avg_think_len": 13.293149948120117,
-            "reward_format_mean": 0.9998337626457214,
-            "reward_time_mean": 0.784375011920929,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.5961887836456299,
-            "step": 25
-        },
-        {
-            "loss": 1.74669548869133e-06,
-            "reward_mean": 0.7027819156646729,
-            "reward_var": 0.0747334286570549,
-            "avg_think_len": 12.695978164672852,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.7406250238418579,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.5705026388168335,
-            "step": 26
-        },
-        {
-            "loss": 1.7386628314852715e-06,
-            "reward_mean": 0.6944615244865417,
-            "reward_var": 0.04436762258410454,
-            "avg_think_len": 10.807842254638672,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.840624988079071,
-            "reward_correctness_mean": 0.421875,
-            "reward_response_efficiency_mean": 0.5153460502624512,
-            "step": 27
-        },
-        {
-            "loss": 1.6023404896259308e-06,
-            "reward_mean": 0.7718853950500488,
-            "reward_var": 0.06972925364971161,
-            "avg_think_len": 13.151269912719727,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8062499761581421,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.5781665444374084,
-            "step": 28
-        },
-        {
-            "loss": 1.5385448932647705e-06,
-            "reward_mean": 0.781177818775177,
-            "reward_var": 0.05613230913877487,
-            "avg_think_len": 13.92419147491455,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.7937500476837158,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.6434611082077026,
-            "step": 29
-        },
-        {
-            "loss": 1.6526319086551666e-06,
-            "reward_mean": 0.7876669764518738,
-            "reward_var": 0.05077036842703819,
-            "avg_think_len": 13.107754707336426,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8500000238418579,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.644417941570282,
-            "step": 30
-        },
-        {
-            "loss": 1.6819685697555542e-06,
-            "reward_mean": 0.7292056083679199,
-            "reward_var": 0.05025945231318474,
-            "avg_think_len": 12.43581485748291,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.793749988079071,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.5761975049972534,
-            "step": 31
-        },
-        {
-            "loss": 1.8381979316473007e-06,
-            "reward_mean": 0.7169650197029114,
-            "reward_var": 0.0741775631904602,
-            "avg_think_len": 13.911001205444336,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.699999988079071,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.5428599119186401,
-            "step": 32
-        },
-        {
-            "loss": 2.0063453121110797e-06,
-            "reward_mean": 0.7654371857643127,
-            "reward_var": 0.06410399824380875,
-            "avg_think_len": 14.393830299377441,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.7750000357627869,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.6461239457130432,
-            "step": 33
-        },
-        {
-            "loss": 1.8693972378969193e-06,
-            "reward_mean": 0.7214520573616028,
-            "reward_var": 0.06847646832466125,
-            "avg_think_len": 13.804563522338867,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.7749999761581421,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.5795583128929138,
-            "step": 34
-        },
-        {
-            "loss": 1.8926803022623062e-06,
-            "reward_mean": 0.7733328938484192,
-            "reward_var": 0.056004032492637634,
-            "avg_think_len": 12.00368595123291,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.824999988079071,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.5964563488960266,
-            "step": 35
-        },
-        {
-            "loss": 2.0600855350494385e-06,
-            "reward_mean": 0.825847864151001,
-            "reward_var": 0.03843200206756592,
-            "avg_think_len": 13.889095306396484,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9125000238418579,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.7190163731575012,
-            "step": 36
-        },
-        {
-            "loss": 1.5562400221824646e-06,
-            "reward_mean": 0.775917112827301,
-            "reward_var": 0.03234148770570755,
-            "avg_think_len": 12.527803421020508,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.890625,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.6817933320999146,
-            "step": 37
-        },
-        {
-            "loss": 1.3525132089853287e-06,
-            "reward_mean": 0.6870580911636353,
-            "reward_var": 0.05082874745130539,
-            "avg_think_len": 10.49350357055664,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8187500238418579,
-            "reward_correctness_mean": 0.40625,
-            "reward_response_efficiency_mean": 0.5232323408126831,
-            "step": 38
-        },
-        {
-            "loss": 2.100132405757904e-06,
-            "reward_mean": 0.759411096572876,
-            "reward_var": 0.0560121051967144,
-            "avg_think_len": 12.580765724182129,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.84375,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.5532693266868591,
-            "step": 39
-        },
-        {
-            "loss": 2.287793904542923e-06,
-            "reward_mean": 0.7701317667961121,
-            "reward_var": 0.05402165651321411,
-            "avg_think_len": 12.589027404785156,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8187499642372131,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.5742772221565247,
-            "step": 40
-        },
-        {
-            "loss": 2.08243727684021e-06,
-            "reward_mean": 0.7352240085601807,
-            "reward_var": 0.05884271487593651,
-            "avg_think_len": 13.969745635986328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.78125,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.6127713322639465,
-            "step": 41
-        },
-        {
-            "loss": 3.1660893000662327e-06,
-            "reward_mean": 0.7248021364212036,
-            "reward_var": 0.07364002615213394,
-            "avg_think_len": 14.592326164245605,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8031250238418579,
-            "reward_correctness_mean": 0.484375,
-            "reward_response_efficiency_mean": 0.6117082834243774,
-            "step": 42
-        },
-        {
-            "loss": 2.2131571313366294e-06,
-            "reward_mean": 0.791403591632843,
-            "reward_var": 0.06027824431657791,
-            "avg_think_len": 13.919354438781738,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.84375,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.6656143665313721,
-            "step": 43
-        },
-        {
-            "loss": 1.9684084691107273e-06,
-            "reward_mean": 0.7604935169219971,
-            "reward_var": 0.03516477346420288,
-            "avg_think_len": 12.856415748596191,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8531249761581421,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.5794739723205566,
-            "step": 44
-        },
-        {
-            "loss": 1.6869626051629893e-06,
-            "reward_mean": 0.7341602444648743,
-            "reward_var": 0.04462061822414398,
-            "avg_think_len": 13.945775985717773,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.840624988079071,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.6272659301757812,
-            "step": 45
-        },
-        {
-            "loss": 1.8889550119638443e-06,
-            "reward_mean": 0.7766642570495605,
-            "reward_var": 0.013862980529665947,
-            "avg_think_len": 13.704081535339355,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.6691567897796631,
-            "step": 46
-        },
-        {
-            "loss": 1.996289938688278e-06,
-            "reward_mean": 0.8085702061653137,
-            "reward_var": 0.03528052568435669,
-            "avg_think_len": 13.396744728088379,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.893750011920929,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.6999058723449707,
-            "step": 47
-        },
-        {
-            "loss": 1.7459387890994549e-06,
-            "reward_mean": 0.7559542655944824,
-            "reward_var": 0.05038033425807953,
-            "avg_think_len": 13.255315780639648,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8468749523162842,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.6300672888755798,
-            "step": 48
-        },
-        {
-            "loss": 2.3691682144999504e-06,
-            "reward_mean": 0.744377851486206,
-            "reward_var": 0.04699885472655296,
-            "avg_think_len": 13.092660903930664,
-            "reward_format_mean": 0.9998183250427246,
-            "reward_time_mean": 0.8187499642372131,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.5808182954788208,
-            "step": 49
-        },
-        {
-            "loss": 2.1138694137334824e-06,
-            "reward_mean": 0.8044846653938293,
-            "reward_var": 0.04107678681612015,
-            "avg_think_len": 13.051464080810547,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8656249642372131,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.6648138165473938,
-            "step": 50
-        },
-        {
-            "loss": 1.9798753783106804e-06,
-            "reward_mean": 0.7534778118133545,
-            "reward_var": 0.06495855748653412,
-            "avg_think_len": 12.817070007324219,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8312499523162842,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.5889111757278442,
-            "step": 51
-        },
-        {
-            "loss": 1.930864527821541e-06,
-            "reward_mean": 0.7681609988212585,
-            "reward_var": 0.04848902300000191,
-            "avg_think_len": 13.98085880279541,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.778124988079071,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.6382688879966736,
-            "step": 52
-        },
-        {
-            "loss": 1.85379758477211e-06,
-            "reward_mean": 0.7814418077468872,
-            "reward_var": 0.04448670893907547,
-            "avg_think_len": 12.8070707321167,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.903124988079071,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.6445173621177673,
-            "step": 53
-        },
-        {
-            "loss": 2.3418106138706207e-06,
-            "reward_mean": 0.7299690246582031,
-            "reward_var": 0.05045507848262787,
-            "avg_think_len": 13.506301879882812,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.828125,
-            "reward_correctness_mean": 0.484375,
-            "reward_response_efficiency_mean": 0.607375979423523,
-            "step": 54
-        },
-        {
-            "loss": 1.8691644072532654e-06,
-            "reward_mean": 0.7514623999595642,
-            "reward_var": 0.06174487620592117,
-            "avg_think_len": 15.009090423583984,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.7875000238418579,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.6714745759963989,
-            "step": 55
-        },
-        {
-            "loss": 2.2622407414019108e-06,
-            "reward_mean": 0.7651642560958862,
-            "reward_var": 0.06094200909137726,
-            "avg_think_len": 14.182782173156738,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8187500238418579,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.648157000541687,
-            "step": 56
-        },
-        {
-            "loss": 1.9930448615923524e-06,
-            "reward_mean": 0.7437002658843994,
-            "reward_var": 0.04525700956583023,
-            "avg_think_len": 13.282563209533691,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8500000238418579,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.6248011589050293,
-            "step": 57
-        },
-        {
-            "loss": 2.705957740545273e-06,
-            "reward_mean": 0.757960319519043,
-            "reward_var": 0.07012414187192917,
-            "avg_think_len": 15.156267166137695,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.7906249761581421,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.7099664807319641,
-            "step": 58
-        },
-        {
-            "loss": 2.355664037168026e-06,
-            "reward_mean": 0.7737964391708374,
-            "reward_var": 0.04142124578356743,
-            "avg_think_len": 14.692024230957031,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8781249523162842,
-            "reward_correctness_mean": 0.453125,
-            "reward_response_efficiency_mean": 0.7639356851577759,
-            "step": 59
-        },
-        {
-            "loss": 2.6927446015179157e-06,
-            "reward_mean": 0.8173869848251343,
-            "reward_var": 0.03152291104197502,
-            "avg_think_len": 13.037355422973633,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9124999642372131,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.6695478558540344,
-            "step": 60
-        },
-        {
-            "loss": 1.9385479390621185e-06,
-            "reward_mean": 0.6646157503128052,
-            "reward_var": 0.062487732619047165,
-            "avg_think_len": 12.052692413330078,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.778124988079071,
-            "reward_correctness_mean": 0.375,
-            "reward_response_efficiency_mean": 0.5053378343582153,
-            "step": 61
-        },
-        {
-            "loss": 2.366490662097931e-06,
-            "reward_mean": 0.7319700121879578,
-            "reward_var": 0.06375718116760254,
-            "avg_think_len": 13.580440521240234,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8031250238418579,
-            "reward_correctness_mean": 0.484375,
-            "reward_response_efficiency_mean": 0.6403799057006836,
-            "step": 62
-        },
-        {
-            "loss": 2.3224856704473495e-06,
-            "reward_mean": 0.728635311126709,
-            "reward_var": 0.05081430450081825,
-            "avg_think_len": 13.546395301818848,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8343749642372131,
-            "reward_correctness_mean": 0.390625,
-            "reward_response_efficiency_mean": 0.6895413398742676,
-            "step": 63
-        },
-        {
-            "loss": 2.5984481908380985e-06,
-            "reward_mean": 0.8289053440093994,
-            "reward_var": 0.033997539430856705,
-            "avg_think_len": 14.753890037536621,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.918749988079071,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.7249963879585266,
-            "step": 64
-        },
-        {
-            "loss": 2.635992132127285e-06,
-            "reward_mean": 0.773135244846344,
-            "reward_var": 0.0630728155374527,
-            "avg_think_len": 14.171107292175293,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.828125,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.6862911581993103,
-            "step": 65
-        },
-        {
-            "loss": 2.766028046607971e-06,
-            "reward_mean": 0.7541279196739197,
-            "reward_var": 0.05175236612558365,
-            "avg_think_len": 13.269648551940918,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.875,
-            "reward_correctness_mean": 0.484375,
-            "reward_response_efficiency_mean": 0.6571367383003235,
-            "step": 66
-        },
-        {
-            "loss": 2.814223989844322e-06,
-            "reward_mean": 0.8227129578590393,
-            "reward_var": 0.03826793283224106,
-            "avg_think_len": 14.927669525146484,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8999999761581421,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.7658514380455017,
-            "step": 67
-        },
-        {
-            "loss": 2.6426278054714203e-06,
-            "reward_mean": 0.8838137984275818,
-            "reward_var": 0.02438708394765854,
-            "avg_think_len": 13.755559921264648,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.7540052533149719,
-            "step": 68
-        },
-        {
-            "loss": 2.8166105039417744e-06,
-            "reward_mean": 0.7758777141571045,
-            "reward_var": 0.05040447786450386,
-            "avg_think_len": 14.985048294067383,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.871874988079071,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.6847608685493469,
-            "step": 69
-        },
-        {
-            "loss": 2.3720785975456238e-06,
-            "reward_mean": 0.7515663504600525,
-            "reward_var": 0.03732605651021004,
-            "avg_think_len": 14.782647132873535,
-            "reward_format_mean": 0.9997305870056152,
-            "reward_time_mean": 0.8843750357627869,
-            "reward_correctness_mean": 0.4375,
-            "reward_response_efficiency_mean": 0.6846597790718079,
-            "step": 70
-        },
-        {
-            "loss": 2.966262400150299e-06,
-            "reward_mean": 0.8373886346817017,
-            "reward_var": 0.04211440682411194,
-            "avg_think_len": 13.365338325500488,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.90625,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.7089297771453857,
-            "step": 71
-        },
-        {
-            "loss": 2.834014594554901e-06,
-            "reward_mean": 0.7529171109199524,
-            "reward_var": 0.05902469903230667,
-            "avg_think_len": 14.864784240722656,
-            "reward_format_mean": 0.9998337626457214,
-            "reward_time_mean": 0.8031250238418579,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.6305846571922302,
-            "step": 72
-        },
-        {
-            "loss": 3.0902447178959846e-06,
-            "reward_mean": 0.8123693466186523,
-            "reward_var": 0.030252249911427498,
-            "avg_think_len": 13.588669776916504,
-            "reward_format_mean": 0.9991776347160339,
-            "reward_time_mean": 0.893750011920929,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.7002995014190674,
-            "step": 73
-        },
-        {
-            "loss": 2.4614855647087097e-06,
-            "reward_mean": 0.7838570475578308,
-            "reward_var": 0.06278469413518906,
-            "avg_think_len": 13.458584785461426,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.828125,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.6198031306266785,
-            "step": 74
-        },
-        {
-            "loss": 2.658634912222624e-06,
-            "reward_mean": 0.7978320717811584,
-            "reward_var": 0.02276075817644596,
-            "avg_think_len": 14.790016174316406,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9437500238418579,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.7319531440734863,
-            "step": 75
-        },
-        {
-            "loss": 2.691929694265127e-06,
-            "reward_mean": 0.7456363439559937,
-            "reward_var": 0.037588149309158325,
-            "avg_think_len": 14.74563217163086,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8656250238418579,
-            "reward_correctness_mean": 0.40625,
-            "reward_response_efficiency_mean": 0.7106702327728271,
-            "step": 76
-        },
-        {
-            "loss": 2.669592504389584e-06,
-            "reward_mean": 0.8524651527404785,
-            "reward_var": 0.02494010329246521,
-            "avg_think_len": 14.906591415405273,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.934374988079071,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.7098605632781982,
-            "step": 77
-        },
-        {
-            "loss": 3.107357770204544e-06,
-            "reward_mean": 0.8007457852363586,
-            "reward_var": 0.0417976900935173,
-            "avg_think_len": 14.073814392089844,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9093749523162842,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.6686081290245056,
-            "step": 78
-        },
-        {
-            "loss": 2.6148336473852396e-06,
-            "reward_mean": 0.8704825639724731,
-            "reward_var": 0.01936575584113598,
-            "avg_think_len": 15.442326545715332,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.8100553750991821,
-            "step": 79
-        },
-        {
-            "loss": 2.5948684196919203e-06,
-            "reward_mean": 0.7465105056762695,
-            "reward_var": 0.026121195405721664,
-            "avg_think_len": 14.092713356018066,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.856249988079071,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.6610422134399414,
-            "step": 80
-        },
-        {
-            "loss": 3.152294084429741e-06,
-            "reward_mean": 0.8826056718826294,
-            "reward_var": 0.02663850039243698,
-            "avg_think_len": 15.167689323425293,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9593750238418579,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.7585476636886597,
-            "step": 81
-        },
-        {
-            "loss": 3.5348348319530487e-06,
-            "reward_mean": 0.7807178497314453,
-            "reward_var": 0.04511016234755516,
-            "avg_think_len": 16.388896942138672,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.890625,
-            "reward_correctness_mean": 0.453125,
-            "reward_response_efficiency_mean": 0.7791215181350708,
-            "step": 82
-        },
-        {
-            "loss": 3.7490390241146088e-06,
-            "reward_mean": 0.8194796442985535,
-            "reward_var": 0.039470963180065155,
-            "avg_think_len": 14.347061157226562,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8999999761581421,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.7216686010360718,
-            "step": 83
-        },
-        {
-            "loss": 2.9371585696935654e-06,
-            "reward_mean": 0.767212986946106,
-            "reward_var": 0.02296479232609272,
-            "avg_think_len": 13.82998275756836,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.484375,
-            "reward_response_efficiency_mean": 0.6626021265983582,
-            "step": 84
-        },
-        {
-            "loss": 3.7116114981472492e-06,
-            "reward_mean": 0.7434417009353638,
-            "reward_var": 0.04408887028694153,
-            "avg_think_len": 14.485742568969727,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8843749761581421,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.6206420063972473,
-            "step": 85
-        },
-        {
-            "loss": 3.9316946640610695e-06,
-            "reward_mean": 0.7948949933052063,
-            "reward_var": 0.05187084525823593,
-            "avg_think_len": 15.442660331726074,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.875,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.7577050924301147,
-            "step": 86
-        },
-        {
-            "loss": 4.672387149184942e-06,
-            "reward_mean": 0.8577666878700256,
-            "reward_var": 0.03641882911324501,
-            "avg_think_len": 14.840174674987793,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.918749988079071,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.7623170614242554,
-            "step": 87
-        },
-        {
-            "loss": 3.434019163250923e-06,
-            "reward_mean": 0.7639906406402588,
-            "reward_var": 0.03379521518945694,
-            "avg_think_len": 14.595831871032715,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8843749761581421,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.6247129440307617,
-            "step": 88
-        },
-        {
-            "loss": 4.478555638343096e-06,
-            "reward_mean": 0.810924768447876,
-            "reward_var": 0.03692188113927841,
-            "avg_think_len": 16.406963348388672,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.903124988079071,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.8249491453170776,
-            "step": 89
-        },
-        {
-            "loss": 3.3713877201080322e-06,
-            "reward_mean": 0.81424880027771,
-            "reward_var": 0.04461997002363205,
-            "avg_think_len": 16.14203643798828,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8812499642372131,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.7976202368736267,
-            "step": 90
-        },
-        {
-            "loss": 3.5943230614066124e-06,
-            "reward_mean": 0.8701696395874023,
-            "reward_var": 0.02355828881263733,
-            "avg_think_len": 16.571409225463867,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9593750238418579,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.8338037729263306,
-            "step": 91
-        },
-        {
-            "loss": 4.823319613933563e-06,
-            "reward_mean": 0.8263680934906006,
-            "reward_var": 0.03324843570590019,
-            "avg_think_len": 16.27916717529297,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8999999761581421,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.7335976362228394,
-            "step": 92
-        },
-        {
-            "loss": 3.476568963378668e-06,
-            "reward_mean": 0.8010643720626831,
-            "reward_var": 0.02026921696960926,
-            "avg_think_len": 15.573912620544434,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.956250011920929,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.685507595539093,
-            "step": 93
-        },
-        {
-            "loss": 3.72924841940403e-06,
-            "reward_mean": 0.7623965740203857,
-            "reward_var": 0.028365813195705414,
-            "avg_think_len": 15.183035850524902,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.925000011920929,
-            "reward_correctness_mean": 0.421875,
-            "reward_response_efficiency_mean": 0.702711284160614,
-            "step": 94
-        },
-        {
-            "loss": 5.121342837810516e-06,
-            "reward_mean": 0.8899016976356506,
-            "reward_var": 0.02057792618870735,
-            "avg_think_len": 14.9500093460083,
-            "reward_format_mean": 0.9991776347160339,
-            "reward_time_mean": 0.965624988079071,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.8135542273521423,
-            "step": 95
-        },
-        {
-            "loss": 4.317611455917358e-06,
-            "reward_mean": 0.773345410823822,
-            "reward_var": 0.05101017281413078,
-            "avg_think_len": 15.080777168273926,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.871874988079071,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.5652567148208618,
-            "step": 96
-        },
-        {
-            "loss": 5.8675650507211685e-06,
-            "reward_mean": 0.8131956458091736,
-            "reward_var": 0.047767989337444305,
-            "avg_think_len": 15.664886474609375,
-            "reward_format_mean": 0.9998301267623901,
-            "reward_time_mean": 0.871874988079071,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.7404523491859436,
-            "step": 97
-        },
-        {
-            "loss": 4.774192348122597e-06,
-            "reward_mean": 0.8768593072891235,
-            "reward_var": 0.03446066752076149,
-            "avg_think_len": 15.403018951416016,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9249999523162842,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.8011870384216309,
-            "step": 98
-        },
-        {
-            "loss": 5.093286745250225e-06,
-            "reward_mean": 0.8356963396072388,
-            "reward_var": 0.024297166615724564,
-            "avg_think_len": 16.27056884765625,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.940625011920929,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.855285108089447,
-            "step": 99
-        },
-        {
-            "loss": 6.548594683408737e-06,
-            "reward_mean": 0.8459983468055725,
-            "reward_var": 0.03714003413915634,
-            "avg_think_len": 15.524375915527344,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8968750238418579,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.721493124961853,
-            "step": 100
-        },
-        {
-            "loss": 5.544628947973251e-06,
-            "reward_mean": 0.8856593370437622,
-            "reward_var": 0.03317353501915932,
-            "avg_think_len": 16.407577514648438,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.934374988079071,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.8582623600959778,
-            "step": 101
-        },
-        {
-            "loss": 4.792062100023031e-06,
-            "reward_mean": 0.7897721529006958,
-            "reward_var": 0.05056392773985863,
-            "avg_think_len": 16.425939559936523,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.859375,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.7997136116027832,
-            "step": 102
-        },
-        {
-            "loss": 5.621928721666336e-06,
-            "reward_mean": 0.8022594451904297,
-            "reward_var": 0.026303669437766075,
-            "avg_think_len": 17.043153762817383,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9124999642372131,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.7965375185012817,
-            "step": 103
-        },
-        {
-            "loss": 5.420763045549393e-06,
-            "reward_mean": 0.840973973274231,
-            "reward_var": 0.030558161437511444,
-            "avg_think_len": 16.038347244262695,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9031250476837158,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.7888959050178528,
-            "step": 104
-        },
-        {
-            "loss": 4.631117917597294e-06,
-            "reward_mean": 0.8294044733047485,
-            "reward_var": 0.0331323966383934,
-            "avg_think_len": 15.191181182861328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.934374988079071,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.7582428455352783,
-            "step": 105
-        },
-        {
-            "loss": 4.673842340707779e-06,
-            "reward_mean": 0.8717131018638611,
-            "reward_var": 0.025020373985171318,
-            "avg_think_len": 16.112873077392578,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.949999988079071,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.8493527173995972,
-            "step": 106
-        },
-        {
-            "loss": 5.479785613715649e-06,
-            "reward_mean": 0.8336004614830017,
-            "reward_var": 0.02550715208053589,
-            "avg_think_len": 16.622886657714844,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.7406517863273621,
-            "step": 107
-        },
-        {
-            "loss": 5.4389238357543945e-06,
-            "reward_mean": 0.8147554397583008,
-            "reward_var": 0.04394793137907982,
-            "avg_think_len": 15.66575813293457,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.875,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.7277717590332031,
-            "step": 108
-        },
-        {
-            "loss": 6.669666618108749e-06,
-            "reward_mean": 0.8670757412910461,
-            "reward_var": 0.02441396936774254,
-            "avg_think_len": 16.446426391601562,
-            "reward_format_mean": 0.9984375238418579,
-            "reward_time_mean": 0.9281249642372131,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.8229900598526001,
-            "step": 109
-        },
-        {
-            "loss": 5.176756531000137e-06,
-            "reward_mean": 0.8305495977401733,
-            "reward_var": 0.03683268278837204,
-            "avg_think_len": 16.579378128051758,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.909375011920929,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.8190733194351196,
-            "step": 110
-        },
-        {
-            "loss": 6.638932973146439e-06,
-            "reward_mean": 0.8622018098831177,
-            "reward_var": 0.040580615401268005,
-            "avg_think_len": 16.053508758544922,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.893750011920929,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.8050569295883179,
-            "step": 111
-        },
-        {
-            "loss": 5.4924748837947845e-06,
-            "reward_mean": 0.7862641215324402,
-            "reward_var": 0.026207279413938522,
-            "avg_think_len": 17.304733276367188,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.3125,
-            "reward_response_efficiency_mean": 0.8950566649436951,
-            "step": 112
-        },
-        {
-            "loss": 7.499242201447487e-06,
-            "reward_mean": 0.8746403455734253,
-            "reward_var": 0.024354824796319008,
-            "avg_think_len": 17.02338981628418,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.949999988079071,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.8923116326332092,
-            "step": 113
-        },
-        {
-            "loss": 5.9390440583229065e-06,
-            "reward_mean": 0.8250576257705688,
-            "reward_var": 0.020868342369794846,
-            "avg_think_len": 16.67304229736328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.7533554434776306,
-            "step": 114
-        },
-        {
-            "loss": 5.622394382953644e-06,
-            "reward_mean": 0.8510746955871582,
-            "reward_var": 0.02180352620780468,
-            "avg_think_len": 16.700313568115234,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.7792987823486328,
-            "step": 115
-        },
-        {
-            "loss": 6.997492164373398e-06,
-            "reward_mean": 0.8574738502502441,
-            "reward_var": 0.028365764766931534,
-            "avg_think_len": 17.065322875976562,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9593750238418579,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.7830206751823425,
-            "step": 116
-        },
-        {
-            "loss": 6.090151146054268e-06,
-            "reward_mean": 0.8409756422042847,
-            "reward_var": 0.028607802465558052,
-            "avg_think_len": 16.24945831298828,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.8014026880264282,
-            "step": 117
-        },
-        {
-            "loss": 5.640322342514992e-06,
-            "reward_mean": 0.8701163530349731,
-            "reward_var": 0.013573025353252888,
-            "avg_think_len": 17.662660598754883,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9562499523162842,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.883590579032898,
-            "step": 118
-        },
-        {
-            "loss": 7.212162017822266e-06,
-            "reward_mean": 0.8386353850364685,
-            "reward_var": 0.012432902120053768,
-            "avg_think_len": 17.006380081176758,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9906250238418579,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.8014165759086609,
-            "step": 119
-        },
-        {
-            "loss": 5.870126187801361e-06,
-            "reward_mean": 0.8638314008712769,
-            "reward_var": 0.019624903798103333,
-            "avg_think_len": 16.937767028808594,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.965624988079071,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.8959503769874573,
-            "step": 120
-        },
-        {
-            "loss": 7.020076736807823e-06,
-            "reward_mean": 0.8609692454338074,
-            "reward_var": 0.04225597530603409,
-            "avg_think_len": 17.033878326416016,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.887499988079071,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.7751269936561584,
-            "step": 121
-        },
-        {
-            "loss": 7.272115908563137e-06,
-            "reward_mean": 0.893592894077301,
-            "reward_var": 0.027303362265229225,
-            "avg_think_len": 16.509857177734375,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.8556215763092041,
-            "step": 122
-        },
-        {
-            "loss": 6.586196832358837e-06,
-            "reward_mean": 0.8478879928588867,
-            "reward_var": 0.029848186299204826,
-            "avg_think_len": 17.259204864501953,
-            "reward_format_mean": 0.9989583492279053,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.8144684433937073,
-            "step": 123
-        },
-        {
-            "loss": 7.285038009285927e-06,
-            "reward_mean": 0.8481774926185608,
-            "reward_var": 0.022158287465572357,
-            "avg_think_len": 17.2777042388916,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.7833350896835327,
-            "step": 124
-        },
-        {
-            "loss": 6.051268428564072e-06,
-            "reward_mean": 0.8345270156860352,
-            "reward_var": 0.0324590727686882,
-            "avg_think_len": 17.033004760742188,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.90625,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.8068579435348511,
-            "step": 125
-        },
-        {
-            "loss": 9.305542334914207e-06,
-            "reward_mean": 0.8421034812927246,
-            "reward_var": 0.01324191503226757,
-            "avg_think_len": 17.815330505371094,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9749999642372131,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.8934137225151062,
-            "step": 126
-        },
-        {
-            "loss": 7.471477147191763e-06,
-            "reward_mean": 0.8634920120239258,
-            "reward_var": 0.022424854338169098,
-            "avg_think_len": 17.712661743164062,
-            "reward_format_mean": 0.9984558820724487,
-            "reward_time_mean": 0.9156250357627869,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.8680121898651123,
-            "step": 127
-        },
-        {
-            "loss": 7.846974767744541e-06,
-            "reward_mean": 0.8187187314033508,
-            "reward_var": 0.023316845297813416,
-            "avg_think_len": 16.99738121032715,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9593749642372131,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.7998747825622559,
-            "step": 128
-        },
-        {
-            "loss": 8.756294846534729e-06,
-            "reward_mean": 0.8919355273246765,
-            "reward_var": 0.0124326441437006,
-            "avg_think_len": 18.004894256591797,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.8646167516708374,
-            "step": 129
-        },
-        {
-            "loss": 5.481299012899399e-06,
-            "reward_mean": 0.8266474604606628,
-            "reward_var": 0.021117864176630974,
-            "avg_think_len": 17.47282600402832,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.453125,
-            "reward_response_efficiency_mean": 0.9003400206565857,
-            "step": 130
-        },
-        {
-            "loss": 9.919516742229462e-06,
-            "reward_mean": 0.8993935585021973,
-            "reward_var": 0.02800278179347515,
-            "avg_think_len": 18.121477127075195,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9281250238418579,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9194492697715759,
-            "step": 131
-        },
-        {
-            "loss": 7.603317499160767e-06,
-            "reward_mean": 0.8679942488670349,
-            "reward_var": 0.027416495606303215,
-            "avg_think_len": 18.118074417114258,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9406249523162842,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.8282272219657898,
-            "step": 132
-        },
-        {
-            "loss": 7.232418283820152e-06,
-            "reward_mean": 0.8271318674087524,
-            "reward_var": 0.02730587311089039,
-            "avg_think_len": 17.924259185791016,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8968750238418579,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.864777147769928,
-            "step": 133
-        },
-        {
-            "loss": 6.929505616426468e-06,
-            "reward_mean": 0.8585745096206665,
-            "reward_var": 0.029549434781074524,
-            "avg_think_len": 16.403751373291016,
-            "reward_format_mean": 0.9998652935028076,
-            "reward_time_mean": 0.909375011920929,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.7750577926635742,
-            "step": 134
-        },
-        {
-            "loss": 9.3690468929708e-06,
-            "reward_mean": 0.8379561305046082,
-            "reward_var": 0.04035284370183945,
-            "avg_think_len": 17.863502502441406,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9125000238418579,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.8768243789672852,
-            "step": 135
-        },
-        {
-            "loss": 7.812632247805595e-06,
-            "reward_mean": 0.8573186993598938,
-            "reward_var": 0.03737058490514755,
-            "avg_think_len": 17.635393142700195,
-            "reward_format_mean": 0.9994611740112305,
-            "reward_time_mean": 0.903124988079071,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.886063814163208,
-            "step": 136
-        },
-        {
-            "loss": 6.84778206050396e-06,
-            "reward_mean": 0.876197874546051,
-            "reward_var": 0.016208864748477936,
-            "avg_think_len": 18.44736099243164,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9110417366027832,
-            "step": 137
-        },
-        {
-            "loss": 7.70576298236847e-06,
-            "reward_mean": 0.8394706845283508,
-            "reward_var": 0.018052419647574425,
-            "avg_think_len": 17.434192657470703,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.8735077977180481,
-            "step": 138
-        },
-        {
-            "loss": 7.5684838520828635e-06,
-            "reward_mean": 0.8102881908416748,
-            "reward_var": 0.022569358348846436,
-            "avg_think_len": 17.07931900024414,
-            "reward_format_mean": 0.9998224377632141,
-            "reward_time_mean": 0.965624988079071,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.7600801587104797,
-            "step": 139
-        },
-        {
-            "loss": 9.554438292980194e-06,
-            "reward_mean": 0.8889332413673401,
-            "reward_var": 0.014177883975207806,
-            "avg_think_len": 18.45769500732422,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9463580846786499,
-            "step": 140
-        },
-        {
-            "loss": 1.0057119652628899e-05,
-            "reward_mean": 0.8580648899078369,
-            "reward_var": 0.02440396323800087,
-            "avg_think_len": 17.879108428955078,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.965624988079071,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.8728845119476318,
-            "step": 141
-        },
-        {
-            "loss": 8.394941687583923e-06,
-            "reward_mean": 0.8178224563598633,
-            "reward_var": 0.026034247130155563,
-            "avg_think_len": 17.417781829833984,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9437499642372131,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.7650399208068848,
-            "step": 142
-        },
-        {
-            "loss": 8.261296898126602e-06,
-            "reward_mean": 0.8673428297042847,
-            "reward_var": 0.01924438774585724,
-            "avg_think_len": 18.326093673706055,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.971875011920929,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.8412461876869202,
-            "step": 143
-        },
-        {
-            "loss": 8.831266313791275e-06,
-            "reward_mean": 0.8415180444717407,
-            "reward_var": 0.028577860444784164,
-            "avg_think_len": 18.088825225830078,
-            "reward_format_mean": 0.9988632798194885,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.8359590768814087,
-            "step": 144
-        },
-        {
-            "loss": 7.59749673306942e-06,
-            "reward_mean": 0.7883090972900391,
-            "reward_var": 0.04449910297989845,
-            "avg_think_len": 17.817564010620117,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.875,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.7313613295555115,
-            "step": 145
-        },
-        {
-            "loss": 1.0331161320209503e-05,
-            "reward_mean": 0.7999853491783142,
-            "reward_var": 0.05270332470536232,
-            "avg_think_len": 18.129318237304688,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.840624988079071,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.765566349029541,
-            "step": 146
-        },
-        {
-            "loss": 1.3828568626195192e-05,
-            "reward_mean": 0.8272857069969177,
-            "reward_var": 0.019872304052114487,
-            "avg_think_len": 17.6873779296875,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.965624988079071,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.8747679591178894,
-            "step": 147
-        },
-        {
-            "loss": 9.906943887472153e-06,
-            "reward_mean": 0.7864844799041748,
-            "reward_var": 0.02020738646388054,
-            "avg_think_len": 17.612760543823242,
-            "reward_format_mean": 0.9993489384651184,
-            "reward_time_mean": 0.918749988079071,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.7590890526771545,
-            "step": 148
-        },
-        {
-            "loss": 1.128949224948883e-05,
-            "reward_mean": 0.8499423265457153,
-            "reward_var": 0.027879828587174416,
-            "avg_think_len": 18.282291412353516,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.934374988079071,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9341441988945007,
-            "step": 149
-        },
-        {
-            "loss": 1.0180810932070017e-05,
-            "reward_mean": 0.8991268873214722,
-            "reward_var": 0.029697677120566368,
-            "avg_think_len": 18.483394622802734,
-            "reward_format_mean": 0.9991319179534912,
-            "reward_time_mean": 0.9437500238418579,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9036258459091187,
-            "step": 150
-        },
-        {
-            "loss": 8.824281394481659e-06,
-            "reward_mean": 0.9741781949996948,
-            "reward_var": 0.002799608279019594,
-            "avg_think_len": 17.9475154876709,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.984375,
-            "reward_response_efficiency_mean": 0.9123376607894897,
-            "step": 151
-        },
-        {
-            "loss": 9.480369044467807e-06,
-            "reward_mean": 0.9086295366287231,
-            "reward_var": 0.020380476489663124,
-            "avg_think_len": 17.684858322143555,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.890625,
-            "reward_response_efficiency_mean": 0.7751432657241821,
-            "step": 152
-        },
-        {
-            "loss": 8.351169526576996e-06,
-            "reward_mean": 0.9027184844017029,
-            "reward_var": 0.028577154502272606,
-            "avg_think_len": 17.992101669311523,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9468749761581421,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.8671241998672485,
-            "step": 153
-        },
-        {
-            "loss": 8.817994967103004e-06,
-            "reward_mean": 0.8447104692459106,
-            "reward_var": 0.029384581372141838,
-            "avg_think_len": 18.302392959594727,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.90625,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9100918769836426,
-            "step": 154
-        },
-        {
-            "loss": 8.788192644715309e-06,
-            "reward_mean": 0.8389168977737427,
-            "reward_var": 0.034577831625938416,
-            "avg_think_len": 18.265701293945312,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.8712925314903259,
-            "step": 155
-        },
-        {
-            "loss": 1.2122094631195068e-05,
-            "reward_mean": 0.8470774292945862,
-            "reward_var": 0.016636168584227562,
-            "avg_think_len": 18.30124282836914,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.971875011920929,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.8383098244667053,
-            "step": 156
-        },
-        {
-            "loss": 8.345348760485649e-06,
-            "reward_mean": 0.897803783416748,
-            "reward_var": 0.01524113118648529,
-            "avg_think_len": 18.6058406829834,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9468749761581421,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9255903959274292,
-            "step": 157
-        },
-        {
-            "loss": 1.078000059351325e-05,
-            "reward_mean": 0.8698163032531738,
-            "reward_var": 0.022024812176823616,
-            "avg_think_len": 18.1617431640625,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9011402130126953,
-            "step": 158
-        },
-        {
-            "loss": 1.0207077139057219e-05,
-            "reward_mean": 0.8879954814910889,
-            "reward_var": 0.02066461555659771,
-            "avg_think_len": 18.316679000854492,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9582316279411316,
-            "step": 159
-        },
-        {
-            "loss": 1.1023599654436111e-05,
-            "reward_mean": 0.8566635251045227,
-            "reward_var": 0.01989481970667839,
-            "avg_think_len": 18.023147583007812,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.8172793984413147,
-            "step": 160
-        },
-        {
-            "loss": 1.0170857422053814e-05,
-            "reward_mean": 0.7697805762290955,
-            "reward_var": 0.02903856709599495,
-            "avg_think_len": 17.442569732666016,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.90625,
-            "reward_correctness_mean": 0.453125,
-            "reward_response_efficiency_mean": 0.7197472453117371,
-            "step": 161
-        },
-        {
-            "loss": 1.2425938621163368e-05,
-            "reward_mean": 0.9209925532341003,
-            "reward_var": 0.012236557900905609,
-            "avg_think_len": 18.128604888916016,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9027199149131775,
-            "step": 162
-        },
-        {
-            "loss": 1.1704396456480026e-05,
-            "reward_mean": 0.8452965021133423,
-            "reward_var": 0.02473929338157177,
-            "avg_think_len": 17.567523956298828,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.909375011920929,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.831186056137085,
-            "step": 163
-        },
-        {
-            "loss": 9.021256119012833e-06,
-            "reward_mean": 0.90485680103302,
-            "reward_var": 0.024359846487641335,
-            "avg_think_len": 18.709928512573242,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9500000476837158,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9350520968437195,
-            "step": 164
-        },
-        {
-            "loss": 1.1536176316440105e-05,
-            "reward_mean": 0.903253972530365,
-            "reward_var": 0.019380753859877586,
-            "avg_think_len": 18.154651641845703,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9286409616470337,
-            "step": 165
-        },
-        {
-            "loss": 8.245697245001793e-06,
-            "reward_mean": 0.937372088432312,
-            "reward_var": 0.02745956741273403,
-            "avg_think_len": 18.820926666259766,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.940625011920929,
-            "reward_correctness_mean": 0.890625,
-            "reward_response_efficiency_mean": 0.9182382822036743,
-            "step": 166
-        },
-        {
-            "loss": 1.099705696105957e-05,
-            "reward_mean": 0.7954496741294861,
-            "reward_var": 0.0462811216711998,
-            "avg_think_len": 18.28058624267578,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.859375,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.8224236965179443,
-            "step": 167
-        },
-        {
-            "loss": 1.3002660125494003e-05,
-            "reward_mean": 0.837645947933197,
-            "reward_var": 0.0207978468388319,
-            "avg_think_len": 17.919170379638672,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.866208553314209,
-            "step": 168
-        },
-        {
-            "loss": 1.091117155738175e-05,
-            "reward_mean": 0.8102036118507385,
-            "reward_var": 0.02767755649983883,
-            "avg_think_len": 17.87204933166504,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9593749642372131,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.718939483165741,
-            "step": 169
-        },
-        {
-            "loss": 9.411247447133064e-06,
-            "reward_mean": 0.8537107706069946,
-            "reward_var": 0.029290275648236275,
-            "avg_think_len": 18.385337829589844,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9312499761581421,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.8742181658744812,
-            "step": 170
-        },
-        {
-            "loss": 1.0997289791703224e-05,
-            "reward_mean": 0.8411421775817871,
-            "reward_var": 0.028254887089133263,
-            "avg_think_len": 18.381654739379883,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.949999988079071,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.8208188414573669,
-            "step": 171
-        },
-        {
-            "loss": 1.2291493476368487e-05,
-            "reward_mean": 0.8975101709365845,
-            "reward_var": 0.010674620978534222,
-            "avg_think_len": 18.73253631591797,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9025406837463379,
-            "step": 172
-        },
-        {
-            "loss": 1.0891817510128021e-05,
-            "reward_mean": 0.8695266842842102,
-            "reward_var": 0.006810578051954508,
-            "avg_think_len": 18.446502685546875,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9031069278717041,
-            "step": 173
-        },
-        {
-            "loss": 1.2096366845071316e-05,
-            "reward_mean": 0.8804792761802673,
-            "reward_var": 0.010758785530924797,
-            "avg_think_len": 18.0139102935791,
-            "reward_format_mean": 0.9992559552192688,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.8351612687110901,
-            "step": 174
-        },
-        {
-            "loss": 1.1592637747526169e-05,
-            "reward_mean": 0.7853264212608337,
-            "reward_var": 0.04847429692745209,
-            "avg_think_len": 18.310571670532227,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8125,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.7194308042526245,
-            "step": 175
-        },
-        {
-            "loss": 1.3203825801610947e-05,
-            "reward_mean": 0.8323444128036499,
-            "reward_var": 0.04584179446101189,
-            "avg_think_len": 17.45680809020996,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.903124988079071,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.8012526035308838,
-            "step": 176
-        },
-        {
-            "loss": 1.3287630281411111e-05,
-            "reward_mean": 0.8065138459205627,
-            "reward_var": 0.022422799840569496,
-            "avg_think_len": 19.16919708251953,
-            "reward_format_mean": 0.9993749856948853,
-            "reward_time_mean": 0.9500000476837158,
-            "reward_correctness_mean": 0.390625,
-            "reward_response_efficiency_mean": 0.8860554695129395,
-            "step": 177
-        },
-        {
-            "loss": 1.4825258404016495e-05,
-            "reward_mean": 0.9351708292961121,
-            "reward_var": 0.02263047732412815,
-            "avg_think_len": 18.93754768371582,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.9438083171844482,
-            "step": 178
-        },
-        {
-            "loss": 1.2509990483522415e-05,
-            "reward_mean": 0.908720850944519,
-            "reward_var": 0.01724495179951191,
-            "avg_think_len": 18.77414894104004,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9317583441734314,
-            "step": 179
-        },
-        {
-            "loss": 1.1524185538291931e-05,
-            "reward_mean": 0.819033682346344,
-            "reward_var": 0.016916442662477493,
-            "avg_think_len": 18.542882919311523,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.375,
-            "reward_response_efficiency_mean": 0.9167600274085999,
-            "step": 180
-        },
-        {
-            "loss": 1.4217570424079895e-05,
-            "reward_mean": 0.8493883013725281,
-            "reward_var": 0.05139686167240143,
-            "avg_think_len": 18.350921630859375,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.890625,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.8819284439086914,
-            "step": 181
-        },
-        {
-            "loss": 1.2186355888843536e-05,
-            "reward_mean": 0.8728307485580444,
-            "reward_var": 0.01436566561460495,
-            "avg_think_len": 19.000795364379883,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9006977081298828,
-            "step": 182
-        },
-        {
-            "loss": 1.3777520507574081e-05,
-            "reward_mean": 0.8673304319381714,
-            "reward_var": 0.014710918068885803,
-            "avg_think_len": 18.927478790283203,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.9536969065666199,
-            "step": 183
-        },
-        {
-            "loss": 1.3071425200905651e-05,
-            "reward_mean": 0.8293657302856445,
-            "reward_var": 0.017620965838432312,
-            "avg_think_len": 18.941917419433594,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.940625011920929,
-            "reward_correctness_mean": 0.453125,
-            "reward_response_efficiency_mean": 0.9237128496170044,
-            "step": 184
-        },
-        {
-            "loss": 1.6307923942804337e-05,
-            "reward_mean": 0.8935468792915344,
-            "reward_var": 0.03536215424537659,
-            "avg_think_len": 18.987478256225586,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8968750238418579,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9116873741149902,
-            "step": 185
-        },
-        {
-            "loss": 1.6767531633377075e-05,
-            "reward_mean": 0.8589034080505371,
-            "reward_var": 0.04232725501060486,
-            "avg_think_len": 19.26349639892578,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.909375011920929,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.8856135606765747,
-            "step": 186
-        },
-        {
-            "loss": 1.3074080925434828e-05,
-            "reward_mean": 0.852889358997345,
-            "reward_var": 0.025136513635516167,
-            "avg_think_len": 19.010738372802734,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.9115572571754456,
-            "step": 187
-        },
-        {
-            "loss": 1.492304727435112e-05,
-            "reward_mean": 0.874435544013977,
-            "reward_var": 0.027595212683081627,
-            "avg_think_len": 18.924535751342773,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.934374988079071,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9383672475814819,
-            "step": 188
-        },
-        {
-            "loss": 7.3285773396492004e-06,
-            "reward_mean": 0.8519479632377625,
-            "reward_var": 0.0410798154771328,
-            "avg_think_len": 19.106752395629883,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.90625,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.8296669125556946,
-            "step": 189
-        },
-        {
-            "loss": 3.109686076641083e-05,
-            "reward_mean": 0.8465777039527893,
-            "reward_var": 0.02178909443318844,
-            "avg_think_len": 19.016460418701172,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.8706856966018677,
-            "step": 190
-        },
-        {
-            "loss": 1.699896529316902e-05,
-            "reward_mean": 0.9318751692771912,
-            "reward_var": 0.01696857623755932,
-            "avg_think_len": 18.685426712036133,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9462506175041199,
-            "step": 191
-        },
-        {
-            "loss": 1.2684977264143527e-05,
-            "reward_mean": 0.8469463586807251,
-            "reward_var": 0.018581271171569824,
-            "avg_think_len": 18.740915298461914,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.8096603155136108,
-            "step": 192
-        },
-        {
-            "loss": 1.4764722436666489e-05,
-            "reward_mean": 0.8346566557884216,
-            "reward_var": 0.03100533038377762,
-            "avg_think_len": 18.599225997924805,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9156249761581421,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.8761267066001892,
-            "step": 193
-        },
-        {
-            "loss": 1.5023630112409592e-05,
-            "reward_mean": 0.90431147813797,
-            "reward_var": 0.026402000337839127,
-            "avg_think_len": 19.08820152282715,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.949999988079071,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9328707456588745,
-            "step": 194
-        },
-        {
-            "loss": 1.4306511729955673e-05,
-            "reward_mean": 0.8539687991142273,
-            "reward_var": 0.007143731229007244,
-            "avg_think_len": 19.061046600341797,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.453125,
-            "reward_response_efficiency_mean": 0.9627501368522644,
-            "step": 195
-        },
-        {
-            "loss": 1.4017801731824875e-05,
-            "reward_mean": 0.7860679030418396,
-            "reward_var": 0.019247381016612053,
-            "avg_think_len": 18.68524932861328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9156249761581421,
-            "reward_correctness_mean": 0.453125,
-            "reward_response_efficiency_mean": 0.7755217552185059,
-            "step": 196
-        },
-        {
-            "loss": 1.486111432313919e-05,
-            "reward_mean": 0.8946903347969055,
-            "reward_var": 0.018983935937285423,
-            "avg_think_len": 18.773157119750977,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9693863987922668,
-            "step": 197
-        },
-        {
-            "loss": 1.3046199455857277e-05,
-            "reward_mean": 0.8305867910385132,
-            "reward_var": 0.015140753239393234,
-            "avg_think_len": 19.095125198364258,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.453125,
-            "reward_response_efficiency_mean": 0.8848474621772766,
-            "step": 198
-        },
-        {
-            "loss": 1.3420358300209045e-05,
-            "reward_mean": 0.8427436351776123,
-            "reward_var": 0.03271010145545006,
-            "avg_think_len": 18.801456451416016,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.918749988079071,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.8428495526313782,
-            "step": 199
-        },
-        {
-            "loss": 7.524155080318451e-06,
-            "reward_mean": 0.8959214687347412,
-            "reward_var": 0.005540060810744762,
-            "avg_think_len": 18.86780548095703,
-            "reward_format_mean": 0.9998372793197632,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.8525987267494202,
-            "step": 200
-        },
-        {
-            "loss": 1.6278820112347603e-05,
-            "reward_mean": 0.8147538304328918,
-            "reward_var": 0.02444630302488804,
-            "avg_think_len": 18.63440704345703,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.925000011920929,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.865265429019928,
-            "step": 201
-        },
-        {
-            "loss": 1.9085593521595e-05,
-            "reward_mean": 0.8608906269073486,
-            "reward_var": 0.0110493628308177,
-            "avg_think_len": 19.125993728637695,
-            "reward_format_mean": 0.9989583492279053,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.8821043968200684,
-            "step": 202
-        },
-        {
-            "loss": 1.5442725270986557e-05,
-            "reward_mean": 0.882067084312439,
-            "reward_var": 0.019237233325839043,
-            "avg_think_len": 18.727659225463867,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9345183372497559,
-            "step": 203
-        },
-        {
-            "loss": 2.072146162390709e-05,
-            "reward_mean": 0.8093749284744263,
-            "reward_var": 0.009821085259318352,
-            "avg_think_len": 18.731895446777344,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.390625,
-            "reward_response_efficiency_mean": 0.8656244874000549,
-            "step": 204
-        },
-        {
-            "loss": 8.718110620975494e-06,
-            "reward_mean": 0.8955650925636292,
-            "reward_var": 0.019447162747383118,
-            "avg_think_len": 19.00971794128418,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.971875011920929,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9385104179382324,
-            "step": 205
-        },
-        {
-            "loss": 1.5472876839339733e-05,
-            "reward_mean": 0.9075912237167358,
-            "reward_var": 0.0306630190461874,
-            "avg_think_len": 18.679636001586914,
-            "reward_format_mean": 0.9988285303115845,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.8659115433692932,
-            "step": 206
-        },
-        {
-            "loss": 1.4970777556300163e-05,
-            "reward_mean": 0.9285392165184021,
-            "reward_var": 0.009293187409639359,
-            "avg_think_len": 18.428348541259766,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.9016571044921875,
-            "step": 207
-        },
-        {
-            "loss": 1.7762038623914123e-05,
-            "reward_mean": 0.9139779806137085,
-            "reward_var": 0.017193952575325966,
-            "avg_think_len": 19.14598274230957,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9371616840362549,
-            "step": 208
-        },
-        {
-            "loss": 1.8122605979442596e-05,
-            "reward_mean": 0.8819992542266846,
-            "reward_var": 0.01999535784125328,
-            "avg_think_len": 19.122249603271484,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9498719573020935,
-            "step": 209
-        },
-        {
-            "loss": 1.5542376786470413e-05,
-            "reward_mean": 0.8483052253723145,
-            "reward_var": 0.014672146178781986,
-            "avg_think_len": 18.8810977935791,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.918749988079071,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.802595853805542,
-            "step": 210
-        },
-        {
-            "loss": 2.2895634174346924e-05,
-            "reward_mean": 0.8957298994064331,
-            "reward_var": 0.01740330271422863,
-            "avg_think_len": 18.961483001708984,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9593750238418579,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.936044454574585,
-            "step": 211
-        },
-        {
-            "loss": 1.5744008123874664e-05,
-            "reward_mean": 0.8459019064903259,
-            "reward_var": 0.01510237529873848,
-            "avg_think_len": 19.170013427734375,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.987500011920929,
-            "reward_correctness_mean": 0.4375,
-            "reward_response_efficiency_mean": 0.9586074948310852,
-            "step": 212
-        },
-        {
-            "loss": 1.617567613720894e-05,
-            "reward_mean": 0.8903298377990723,
-            "reward_var": 0.015294874086976051,
-            "avg_think_len": 18.804990768432617,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.939444363117218,
-            "step": 213
-        },
-        {
-            "loss": 1.3844575732946396e-05,
-            "reward_mean": 0.8608013987541199,
-            "reward_var": 0.01974719762802124,
-            "avg_think_len": 18.638362884521484,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9275805950164795,
-            "step": 214
-        },
-        {
-            "loss": 1.9755680114030838e-05,
-            "reward_mean": 0.8028340339660645,
-            "reward_var": 0.02062096819281578,
-            "avg_think_len": 19.433494567871094,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.971875011920929,
-            "reward_correctness_mean": 0.3125,
-            "reward_response_efficiency_mean": 0.9269613027572632,
-            "step": 215
-        },
-        {
-            "loss": 1.0578311048448086e-05,
-            "reward_mean": 0.9425342082977295,
-            "reward_var": 0.016657544299960136,
-            "avg_think_len": 19.134990692138672,
-            "reward_format_mean": 0.9986978769302368,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.9745637774467468,
-            "step": 216
-        },
-        {
-            "loss": 1.450791023671627e-05,
-            "reward_mean": 0.8757871985435486,
-            "reward_var": 0.02233031578361988,
-            "avg_think_len": 19.115795135498047,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.8625237345695496,
-            "step": 217
-        },
-        {
-            "loss": 1.478591002523899e-05,
-            "reward_mean": 0.8896237015724182,
-            "reward_var": 0.015676328912377357,
-            "avg_think_len": 19.023887634277344,
-            "reward_format_mean": 0.9990234375,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9500965476036072,
-            "step": 218
-        },
-        {
-            "loss": 1.3374967238632962e-05,
-            "reward_mean": 0.8645636439323425,
-            "reward_var": 0.02706003189086914,
-            "avg_think_len": 19.173599243164062,
-            "reward_format_mean": 0.9994611740112305,
-            "reward_time_mean": 0.9593750238418579,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.8431683778762817,
-            "step": 219
-        },
-        {
-            "loss": 1.7273472622036934e-05,
-            "reward_mean": 0.8838921189308167,
-            "reward_var": 0.011556870304048061,
-            "avg_think_len": 18.558731079101562,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9449436068534851,
-            "step": 220
-        },
-        {
-            "loss": 1.998012885451317e-05,
-            "reward_mean": 0.9441930651664734,
-            "reward_var": 0.007913444191217422,
-            "avg_think_len": 18.95622444152832,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9798972606658936,
-            "step": 221
-        },
-        {
-            "loss": 1.767091453075409e-05,
-            "reward_mean": 0.8446711897850037,
-            "reward_var": 0.037095218896865845,
-            "avg_think_len": 18.999069213867188,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.890625,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.8474346995353699,
-            "step": 222
-        },
-        {
-            "loss": 1.4728866517543793e-05,
-            "reward_mean": 0.8887279033660889,
-            "reward_var": 0.006966181565076113,
-            "avg_think_len": 19.07766342163086,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.8830368518829346,
-            "step": 223
-        },
-        {
-            "loss": 1.9881990738213062e-05,
-            "reward_mean": 0.869768500328064,
-            "reward_var": 0.020983194932341576,
-            "avg_think_len": 19.155574798583984,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9312499761581421,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9228238463401794,
-            "step": 224
-        },
-        {
-            "loss": 1.7877202481031418e-05,
-            "reward_mean": 0.8562892079353333,
-            "reward_var": 0.03165876492857933,
-            "avg_think_len": 19.251358032226562,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.8626568913459778,
-            "step": 225
-        },
-        {
-            "loss": 2.3181317374110222e-05,
-            "reward_mean": 0.8822413682937622,
-            "reward_var": 0.015445350669324398,
-            "avg_think_len": 19.343191146850586,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.987500011920929,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9477154612541199,
-            "step": 226
-        },
-        {
-            "loss": 7.584691047668457e-06,
-            "reward_mean": 0.8821452856063843,
-            "reward_var": 0.018571235239505768,
-            "avg_think_len": 19.05763816833496,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9156249761581421,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9098310470581055,
-            "step": 227
-        },
-        {
-            "loss": 2.081110142171383e-05,
-            "reward_mean": 0.8947812914848328,
-            "reward_var": 0.018662312999367714,
-            "avg_think_len": 19.284530639648438,
-            "reward_format_mean": 0.9988839626312256,
-            "reward_time_mean": 0.965624988079071,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.911491334438324,
-            "step": 228
-        },
-        {
-            "loss": 1.8564751371741295e-05,
-            "reward_mean": 0.8821424841880798,
-            "reward_var": 0.02390131913125515,
-            "avg_think_len": 19.25907325744629,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9504449367523193,
-            "step": 229
-        },
-        {
-            "loss": 1.6524922102689743e-05,
-            "reward_mean": 0.9002629518508911,
-            "reward_var": 0.010323257185518742,
-            "avg_think_len": 19.133724212646484,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9041768312454224,
-            "step": 230
-        },
-        {
-            "loss": 2.3183412849903107e-05,
-            "reward_mean": 0.9155488610267639,
-            "reward_var": 0.019590752199292183,
-            "avg_think_len": 19.195459365844727,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9434455633163452,
-            "step": 231
-        },
-        {
-            "loss": 2.061016857624054e-05,
-            "reward_mean": 0.8606812953948975,
-            "reward_var": 0.02413790300488472,
-            "avg_think_len": 18.878456115722656,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.934374988079071,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.898975133895874,
-            "step": 232
-        },
-        {
-            "loss": 1.832377165555954e-05,
-            "reward_mean": 0.8642477989196777,
-            "reward_var": 0.02810693345963955,
-            "avg_think_len": 18.882787704467773,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9437499642372131,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.810116171836853,
-            "step": 233
-        },
-        {
-            "loss": 2.2161751985549927e-05,
-            "reward_mean": 0.859771728515625,
-            "reward_var": 0.03219769150018692,
-            "avg_think_len": 18.87949562072754,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.8609620928764343,
-            "step": 234
-        },
-        {
-            "loss": 1.856917515397072e-05,
-            "reward_mean": 0.8605369329452515,
-            "reward_var": 0.017516251653432846,
-            "avg_think_len": 19.100526809692383,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9108974933624268,
-            "step": 235
-        },
-        {
-            "loss": 2.1952204406261444e-05,
-            "reward_mean": 0.9085224866867065,
-            "reward_var": 0.011111894622445107,
-            "avg_think_len": 19.07421112060547,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9340899586677551,
-            "step": 236
-        },
-        {
-            "loss": 2.593756653368473e-05,
-            "reward_mean": 0.8944090008735657,
-            "reward_var": 0.025153741240501404,
-            "avg_think_len": 18.905210494995117,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9057612419128418,
-            "step": 237
-        },
-        {
-            "loss": 2.0576873794198036e-05,
-            "reward_mean": 0.9001258611679077,
-            "reward_var": 0.023974742740392685,
-            "avg_think_len": 19.224149703979492,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.8661284446716309,
-            "step": 238
-        },
-        {
-            "loss": 1.9990140572190285e-05,
-            "reward_mean": 0.8316128253936768,
-            "reward_var": 0.0331561304628849,
-            "avg_think_len": 19.330638885498047,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9125000238418579,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.9139512777328491,
-            "step": 239
-        },
-        {
-            "loss": 1.9814353436231613e-05,
-            "reward_mean": 0.8022868037223816,
-            "reward_var": 0.047687072306871414,
-            "avg_think_len": 19.077388763427734,
-            "reward_format_mean": 0.9998617172241211,
-            "reward_time_mean": 0.8218749761581421,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.8249105215072632,
-            "step": 240
-        },
-        {
-            "loss": 1.4424091205000877e-05,
-            "reward_mean": 0.9000381231307983,
-            "reward_var": 0.010962543077766895,
-            "avg_think_len": 19.405214309692383,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9751524925231934,
-            "step": 241
-        },
-        {
-            "loss": 1.490255817770958e-05,
-            "reward_mean": 0.8720400333404541,
-            "reward_var": 0.03472868353128433,
-            "avg_think_len": 19.11297035217285,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8374999761581421,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.8694101572036743,
-            "step": 242
-        },
-        {
-            "loss": 4.6510249376297e-05,
-            "reward_mean": 0.8992578387260437,
-            "reward_var": 0.01917935162782669,
-            "avg_think_len": 19.642122268676758,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9407812356948853,
-            "step": 243
-        },
-        {
-            "loss": 1.9488157704472542e-05,
-            "reward_mean": 0.864406406879425,
-            "reward_var": 0.02188911661505699,
-            "avg_think_len": 19.081634521484375,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9437500238418579,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9045006632804871,
-            "step": 244
-        },
-        {
-            "loss": 2.872571349143982e-05,
-            "reward_mean": 0.9088007807731628,
-            "reward_var": 0.021898850798606873,
-            "avg_think_len": 19.0022029876709,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9164532423019409,
-            "step": 245
-        },
-        {
-            "loss": 1.9337283447384834e-05,
-            "reward_mean": 0.830836296081543,
-            "reward_var": 0.025477180257439613,
-            "avg_think_len": 19.36417579650879,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.949999988079071,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.9045953154563904,
-            "step": 246
-        },
-        {
-            "loss": 3.987690433859825e-05,
-            "reward_mean": 0.9354565143585205,
-            "reward_var": 0.0035749408416450024,
-            "avg_think_len": 19.411588668823242,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9605761170387268,
-            "step": 247
-        },
-        {
-            "loss": 3.6218203604221344e-05,
-            "reward_mean": 0.8052366971969604,
-            "reward_var": 0.011807342991232872,
-            "avg_think_len": 19.01197052001953,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.859375,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.7990718483924866,
-            "step": 248
-        },
-        {
-            "loss": 2.4728477001190186e-05,
-            "reward_mean": 0.9051703214645386,
-            "reward_var": 0.009450028650462627,
-            "avg_think_len": 19.411813735961914,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9800564050674438,
-            "step": 249
-        },
-        {
-            "loss": 2.677924931049347e-05,
-            "reward_mean": 0.8896052241325378,
-            "reward_var": 0.024407781660556793,
-            "avg_think_len": 19.30546760559082,
-            "reward_format_mean": 0.9998022317886353,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9023686647415161,
-            "step": 250
-        },
-        {
-            "loss": 2.0753999706357718e-05,
-            "reward_mean": 0.8219956159591675,
-            "reward_var": 0.023651836439967155,
-            "avg_think_len": 19.154937744140625,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.8192323446273804,
-            "step": 251
-        },
-        {
-            "loss": 2.0924024283885956e-05,
-            "reward_mean": 0.8556656241416931,
-            "reward_var": 0.024420110508799553,
-            "avg_think_len": 19.181791305541992,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.8757872581481934,
-            "step": 252
-        },
-        {
-            "loss": -3.163772635161877e-05,
-            "reward_mean": 0.8955940008163452,
-            "reward_var": 0.007397867739200592,
-            "avg_think_len": 19.4234676361084,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.90625,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.8636263012886047,
-            "step": 253
-        },
-        {
-            "loss": 2.7934089303016663e-05,
-            "reward_mean": 0.8647134304046631,
-            "reward_var": 0.0177652258425951,
-            "avg_think_len": 19.50368309020996,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9619786739349365,
-            "step": 254
-        },
-        {
-            "loss": 2.733105793595314e-05,
-            "reward_mean": 0.90873783826828,
-            "reward_var": 0.009096908383071423,
-            "avg_think_len": 19.268070220947266,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9662013649940491,
-            "step": 255
-        },
-        {
-            "loss": 2.258364111185074e-05,
-            "reward_mean": 0.9295089244842529,
-            "reward_var": 0.00474793603643775,
-            "avg_think_len": 19.506189346313477,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9836604595184326,
-            "step": 256
-        },
-        {
-            "loss": 1.959875226020813e-05,
-            "reward_mean": 0.9186180233955383,
-            "reward_var": 0.014222877100110054,
-            "avg_think_len": 19.504257202148438,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.987500011920929,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9838468432426453,
-            "step": 257
-        },
-        {
-            "loss": 1.6630161553621292e-05,
-            "reward_mean": 0.8683997988700867,
-            "reward_var": 0.023308150470256805,
-            "avg_think_len": 19.257972717285156,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.911099374294281,
-            "step": 258
-        },
-        {
-            "loss": 1.5940051525831223e-05,
-            "reward_mean": 0.8933631181716919,
-            "reward_var": 0.006040377076715231,
-            "avg_think_len": 19.24982261657715,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.8703272938728333,
-            "step": 259
-        },
-        {
-            "loss": -4.761270247399807e-06,
-            "reward_mean": 0.8430534601211548,
-            "reward_var": 0.02126980572938919,
-            "avg_think_len": 19.385799407958984,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9624999761581421,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.784713864326477,
-            "step": 260
-        },
-        {
-            "loss": 2.1173153072595596e-05,
-            "reward_mean": 0.8728353977203369,
-            "reward_var": 0.035426028072834015,
-            "avg_think_len": 19.147750854492188,
-            "reward_format_mean": 0.9997735619544983,
-            "reward_time_mean": 0.9156249761581421,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.872818112373352,
-            "step": 261
-        },
-        {
-            "loss": 1.7437210772186518e-05,
-            "reward_mean": 0.9013829231262207,
-            "reward_var": 0.019724402576684952,
-            "avg_think_len": 19.406265258789062,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.971875011920929,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.8836565017700195,
-            "step": 262
-        },
-        {
-            "loss": 2.519739791750908e-05,
-            "reward_mean": 0.8963683843612671,
-            "reward_var": 0.0067124078050255775,
-            "avg_think_len": 19.30632209777832,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9917234182357788,
-            "step": 263
-        },
-        {
-            "loss": 2.4707638658583164e-05,
-            "reward_mean": 0.8447768688201904,
-            "reward_var": 0.014084737747907639,
-            "avg_think_len": 19.273502349853516,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.949999988079071,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.7728576064109802,
-            "step": 264
-        },
-        {
-            "loss": 2.806028351187706e-05,
-            "reward_mean": 0.8916487097740173,
-            "reward_var": 0.013191222213208675,
-            "avg_think_len": 19.432098388671875,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9790948629379272,
-            "step": 265
-        },
-        {
-            "loss": 3.430526703596115e-06,
-            "reward_mean": 0.8696948289871216,
-            "reward_var": 0.019563976675271988,
-            "avg_think_len": 19.509721755981445,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9475294351577759,
-            "step": 266
-        },
-        {
-            "loss": 2.318633778486401e-05,
-            "reward_mean": 0.9137172698974609,
-            "reward_var": 0.02084244042634964,
-            "avg_think_len": 19.357894897460938,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.971875011920929,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9173693060874939,
-            "step": 267
-        },
-        {
-            "loss": 1.768302172422409e-05,
-            "reward_mean": 0.8099172115325928,
-            "reward_var": 0.03804973512887955,
-            "avg_think_len": 19.21213722229004,
-            "reward_format_mean": 0.9997767806053162,
-            "reward_time_mean": 0.890625,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.8023920059204102,
-            "step": 268
-        },
-        {
-            "loss": 2.619810402393341e-05,
-            "reward_mean": 0.8873074054718018,
-            "reward_var": 0.019119704142212868,
-            "avg_think_len": 19.402957916259766,
-            "reward_format_mean": 0.999627947807312,
-            "reward_time_mean": 0.9312499761581421,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.8996015191078186,
-            "step": 269
-        },
-        {
-            "loss": 3.4580472856760025e-05,
-            "reward_mean": 0.9430974721908569,
-            "reward_var": 0.007026090286672115,
-            "avg_think_len": 19.404117584228516,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.991139829158783,
-            "step": 270
-        },
-        {
-            "loss": 2.664024941623211e-05,
-            "reward_mean": 0.8522539734840393,
-            "reward_var": 0.013232477009296417,
-            "avg_think_len": 19.401268005371094,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.9558910727500916,
-            "step": 271
-        },
-        {
-            "loss": 2.1659303456544876e-05,
-            "reward_mean": 0.881803035736084,
-            "reward_var": 0.019708998501300812,
-            "avg_think_len": 19.297889709472656,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.9022123217582703,
-            "step": 272
-        },
-        {
-            "loss": 3.360491245985031e-05,
-            "reward_mean": 0.9219844341278076,
-            "reward_var": 0.02336719259619713,
-            "avg_think_len": 19.370445251464844,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.9223127961158752,
-            "step": 273
-        },
-        {
-            "loss": 2.5739893317222595e-05,
-            "reward_mean": 0.8404743671417236,
-            "reward_var": 0.017416391521692276,
-            "avg_think_len": 19.429622650146484,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.971875011920929,
-            "reward_correctness_mean": 0.4375,
-            "reward_response_efficiency_mean": 0.9525225758552551,
-            "step": 274
-        },
-        {
-            "loss": 2.3273940314538777e-05,
-            "reward_mean": 0.8791239857673645,
-            "reward_var": 0.011769622564315796,
-            "avg_think_len": 19.244421005249023,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9727458953857422,
-            "step": 275
-        },
-        {
-            "loss": 2.9502087272703648e-05,
-            "reward_mean": 0.8529045581817627,
-            "reward_var": 0.014200961217284203,
-            "avg_think_len": 19.08867835998535,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.8678684234619141,
-            "step": 276
-        },
-        {
-            "loss": 1.7653917893767357e-05,
-            "reward_mean": 0.8643395900726318,
-            "reward_var": 0.01391957513988018,
-            "avg_think_len": 19.469951629638672,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9906250238418579,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9354833364486694,
-            "step": 277
-        },
-        {
-            "loss": 2.6645604521036148e-05,
-            "reward_mean": 0.9006185531616211,
-            "reward_var": 0.010338244959712029,
-            "avg_think_len": 19.677627563476562,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9774742126464844,
-            "step": 278
-        },
-        {
-            "loss": 3.6108773201704025e-05,
-            "reward_mean": 0.9356260299682617,
-            "reward_var": 0.008575264364480972,
-            "avg_think_len": 19.515243530273438,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9768790006637573,
-            "step": 279
-        },
-        {
-            "loss": 2.559390850365162e-05,
-            "reward_mean": 0.7683826684951782,
-            "reward_var": 0.02744620107114315,
-            "avg_think_len": 19.287063598632812,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.828125,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.7766556739807129,
-            "step": 280
-        },
-        {
-            "loss": 2.5461893528699875e-05,
-            "reward_mean": 0.8835293650627136,
-            "reward_var": 0.024646371603012085,
-            "avg_think_len": 19.074729919433594,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.8934925198554993,
-            "step": 281
-        },
-        {
-            "loss": 3.018835559487343e-05,
-            "reward_mean": 0.8521879315376282,
-            "reward_var": 0.031033333390951157,
-            "avg_think_len": 19.27508544921875,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.918749988079071,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.9431266784667969,
-            "step": 282
-        },
-        {
-            "loss": 2.723064972087741e-05,
-            "reward_mean": 0.8975133895874023,
-            "reward_var": 0.011596347205340862,
-            "avg_think_len": 19.155704498291016,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9806786775588989,
-            "step": 283
-        },
-        {
-            "loss": 3.1754374504089355e-05,
-            "reward_mean": 0.8818458318710327,
-            "reward_var": 0.017979245632886887,
-            "avg_think_len": 19.420024871826172,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9492581486701965,
-            "step": 284
-        },
-        {
-            "loss": 2.4252105504274368e-05,
-            "reward_mean": 0.9105850458145142,
-            "reward_var": 0.005793939344584942,
-            "avg_think_len": 19.23759651184082,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.8923399448394775,
-            "step": 285
-        },
-        {
-            "loss": 4.0597282350063324e-05,
-            "reward_mean": 0.8740155100822449,
-            "reward_var": 0.032058365643024445,
-            "avg_think_len": 19.28706169128418,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.903124988079071,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.8898119926452637,
-            "step": 286
-        },
-        {
-            "loss": 0.00015897140838205814,
-            "reward_mean": 0.8761041760444641,
-            "reward_var": 0.020863106474280357,
-            "avg_think_len": 19.547922134399414,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.8637915849685669,
-            "step": 287
-        },
-        {
-            "loss": 3.7441146560013294e-05,
-            "reward_mean": 0.8580065965652466,
-            "reward_var": 0.015978269279003143,
-            "avg_think_len": 19.279163360595703,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.825776219367981,
-            "step": 288
-        },
-        {
-            "loss": 2.4999957531690598e-05,
-            "reward_mean": 0.9548091292381287,
-            "reward_var": 0.010662185959517956,
-            "avg_think_len": 19.49134063720703,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9750000238418579,
-            "reward_correctness_mean": 0.859375,
-            "reward_response_efficiency_mean": 0.9848614931106567,
-            "step": 289
-        },
-        {
-            "loss": 3.719283267855644e-05,
-            "reward_mean": 0.8597429394721985,
-            "reward_var": 0.007227601949125528,
-            "avg_think_len": 19.752531051635742,
-            "reward_format_mean": 0.9998389482498169,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9235076904296875,
-            "step": 290
-        },
-        {
-            "loss": 3.1509436666965485e-05,
-            "reward_mean": 0.8808531165122986,
-            "reward_var": 0.022072207182645798,
-            "avg_think_len": 19.443117141723633,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9140375852584839,
-            "step": 291
-        },
-        {
-            "loss": 5.176849663257599e-05,
-            "reward_mean": 0.9232056736946106,
-            "reward_var": 0.008472403511404991,
-            "avg_think_len": 19.587509155273438,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9615727066993713,
-            "step": 292
-        },
-        {
-            "loss": 2.1472573280334473e-05,
-            "reward_mean": 0.8975754976272583,
-            "reward_var": 0.007743861060589552,
-            "avg_think_len": 19.656879425048828,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9809267520904541,
-            "step": 293
-        },
-        {
-            "loss": 2.21962109208107e-05,
-            "reward_mean": 0.9244059324264526,
-            "reward_var": 0.007623352110385895,
-            "avg_think_len": 19.511981964111328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9788738489151001,
-            "step": 294
-        },
-        {
-            "loss": 2.2555235773324966e-05,
-            "reward_mean": 0.856839120388031,
-            "reward_var": 0.028282586485147476,
-            "avg_think_len": 19.453580856323242,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.918749988079071,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.8367313742637634,
-            "step": 295
-        },
-        {
-            "loss": 2.681888872757554e-05,
-            "reward_mean": 0.9251057505607605,
-            "reward_var": 0.011359913274645805,
-            "avg_think_len": 19.203649520874023,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.956250011920929,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9629229307174683,
-            "step": 296
-        },
-        {
-            "loss": 2.975109964609146e-06,
-            "reward_mean": 0.9298335909843445,
-            "reward_var": 0.009898843243718147,
-            "avg_think_len": 19.548004150390625,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.953709602355957,
-            "step": 297
-        },
-        {
-            "loss": 2.8105685487389565e-05,
-            "reward_mean": 0.8455364108085632,
-            "reward_var": 0.041748110204935074,
-            "avg_think_len": 19.451942443847656,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.887499988079071,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.8227705955505371,
-            "step": 298
-        },
-        {
-            "loss": 4.36282716691494e-05,
-            "reward_mean": 0.9256057739257812,
-            "reward_var": 0.02435019426047802,
-            "avg_think_len": 19.50413703918457,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.925000011920929,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.933673083782196,
-            "step": 299
-        },
-        {
-            "loss": 3.4895725548267365e-05,
-            "reward_mean": 0.8662328720092773,
-            "reward_var": 0.028994452208280563,
-            "avg_think_len": 19.40651512145996,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.890625,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9336812496185303,
-            "step": 300
-        },
-        {
-            "loss": 2.7009285986423492e-05,
-            "reward_mean": 0.9231680631637573,
-            "reward_var": 0.018109140917658806,
-            "avg_think_len": 19.650760650634766,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9426724910736084,
-            "step": 301
-        },
-        {
-            "loss": 3.2546697184443474e-05,
-            "reward_mean": 0.8152293562889099,
-            "reward_var": 0.007724330760538578,
-            "avg_think_len": 19.52743911743164,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.28125,
-            "reward_response_efficiency_mean": 0.9796673655509949,
-            "step": 302
-        },
-        {
-            "loss": 2.810696605592966e-05,
-            "reward_mean": 0.9252316355705261,
-            "reward_var": 0.02750331535935402,
-            "avg_think_len": 19.469135284423828,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.875,
-            "reward_response_efficiency_mean": 0.9040517210960388,
-            "step": 303
-        },
-        {
-            "loss": 3.2193493098020554e-05,
-            "reward_mean": 0.8757368326187134,
-            "reward_var": 0.019366716966032982,
-            "avg_think_len": 19.282686233520508,
-            "reward_format_mean": 0.9998389482498169,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.8624836206436157,
-            "step": 304
-        },
-        {
-            "loss": 3.0366703867912292e-05,
-            "reward_mean": 0.8912941217422485,
-            "reward_var": 0.019833475351333618,
-            "avg_think_len": 19.486024856567383,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9089264869689941,
-            "step": 305
-        },
-        {
-            "loss": 3.7209480069577694e-05,
-            "reward_mean": 0.9578677415847778,
-            "reward_var": 0.01426482480019331,
-            "avg_think_len": 19.624666213989258,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.921875,
-            "reward_response_efficiency_mean": 0.9408460259437561,
-            "step": 306
-        },
-        {
-            "loss": 3.188103437423706e-05,
-            "reward_mean": 0.8774053454399109,
-            "reward_var": 0.01159857027232647,
-            "avg_think_len": 19.2607421875,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9593749642372131,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9408715963363647,
-            "step": 307
-        },
-        {
-            "loss": 3.969110548496246e-05,
-            "reward_mean": 0.9335014820098877,
-            "reward_var": 0.020613914355635643,
-            "avg_think_len": 19.669618606567383,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.965624988079071,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9558809399604797,
-            "step": 308
-        },
-        {
-            "loss": 3.96497780457139e-05,
-            "reward_mean": 0.9055277109146118,
-            "reward_var": 0.007344159297645092,
-            "avg_think_len": 19.361553192138672,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.934610903263092,
-            "step": 309
-        },
-        {
-            "loss": 4.2086467146873474e-05,
-            "reward_mean": 0.8767485618591309,
-            "reward_var": 0.02101827785372734,
-            "avg_think_len": 19.4946346282959,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9437500238418579,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9226194024085999,
-            "step": 310
-        },
-        {
-            "loss": 2.6189954951405525e-05,
-            "reward_mean": 0.9588912129402161,
-            "reward_var": 0.011949165724217892,
-            "avg_think_len": 19.14373016357422,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.90625,
-            "reward_response_efficiency_mean": 0.9449399709701538,
-            "step": 311
-        },
-        {
-            "loss": 4.4220127165317535e-05,
-            "reward_mean": 0.8708769679069519,
-            "reward_var": 0.013759775087237358,
-            "avg_think_len": 19.544509887695312,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.942882776260376,
-            "step": 312
-        },
-        {
-            "loss": 3.9879290852695704e-05,
-            "reward_mean": 0.8061423301696777,
-            "reward_var": 0.008511688560247421,
-            "avg_think_len": 19.66400909423828,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9781249761581421,
-            "reward_correctness_mean": 0.453125,
-            "reward_response_efficiency_mean": 0.7933192849159241,
-            "step": 313
-        },
-        {
-            "loss": 3.602541983127594e-05,
-            "reward_mean": 0.927036702632904,
-            "reward_var": 0.015334436669945717,
-            "avg_think_len": 19.4395694732666,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9781249761581421,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9800218343734741,
-            "step": 314
-        },
-        {
-            "loss": 5.5675627663731575e-05,
-            "reward_mean": 0.8103815913200378,
-            "reward_var": 0.018893856555223465,
-            "avg_think_len": 19.54987907409668,
-            "reward_format_mean": 0.9993206262588501,
-            "reward_time_mean": 0.84375,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.7890806198120117,
-            "step": 315
-        },
-        {
-            "loss": 3.5098870284855366e-05,
-            "reward_mean": 0.9081262946128845,
-            "reward_var": 0.01972869597375393,
-            "avg_think_len": 19.658626556396484,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9450052380561829,
-            "step": 316
-        },
-        {
-            "loss": -1.974683254957199e-05,
-            "reward_mean": 0.8577726483345032,
-            "reward_var": 0.03791548311710358,
-            "avg_think_len": 19.474409103393555,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.862500011920929,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.849840521812439,
-            "step": 317
-        },
-        {
-            "loss": 3.2033771276474e-05,
-            "reward_mean": 0.863330066204071,
-            "reward_var": 0.015433902852237225,
-            "avg_think_len": 19.82097625732422,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.8751952052116394,
-            "step": 318
-        },
-        {
-            "loss": 3.6745332181453705e-05,
-            "reward_mean": 0.8795796036720276,
-            "reward_var": 0.012476250529289246,
-            "avg_think_len": 19.434720993041992,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9401934742927551,
-            "step": 319
-        },
-        {
-            "loss": 1.9259285181760788e-05,
-            "reward_mean": 0.8920231461524963,
-            "reward_var": 0.008791113272309303,
-            "avg_think_len": 19.449113845825195,
-            "reward_format_mean": 0.9987980723381042,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9786697626113892,
-            "step": 320
-        },
-        {
-            "loss": 4.242849536240101e-05,
-            "reward_mean": 0.843324601650238,
-            "reward_var": 0.017542414367198944,
-            "avg_think_len": 19.522554397583008,
-            "reward_format_mean": 0.9990234375,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.484375,
-            "reward_response_efficiency_mean": 0.9055247902870178,
-            "step": 321
-        },
-        {
-            "loss": 2.4189241230487823e-05,
-            "reward_mean": 0.8534282445907593,
-            "reward_var": 0.02174372784793377,
-            "avg_think_len": 19.776790618896484,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.875,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9449633359909058,
-            "step": 322
-        },
-        {
-            "loss": 3.480605664663017e-05,
-            "reward_mean": 0.9091314673423767,
-            "reward_var": 0.025339266285300255,
-            "avg_think_len": 19.645370483398438,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9177758097648621,
-            "step": 323
-        },
-        {
-            "loss": 3.878655843436718e-05,
-            "reward_mean": 0.9110071063041687,
-            "reward_var": 0.021885275840759277,
-            "avg_think_len": 19.566843032836914,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9096534252166748,
-            "step": 324
-        },
-        {
-            "loss": 3.2937852665781975e-05,
-            "reward_mean": 0.9227596521377563,
-            "reward_var": 0.009394248016178608,
-            "avg_think_len": 19.490400314331055,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9566637873649597,
-            "step": 325
-        },
-        {
-            "loss": 3.701820969581604e-05,
-            "reward_mean": 0.8508248925209045,
-            "reward_var": 0.018503179773688316,
-            "avg_think_len": 19.396482467651367,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.8876745104789734,
-            "step": 326
-        },
-        {
-            "loss": 5.844072438776493e-05,
-            "reward_mean": 0.7896108627319336,
-            "reward_var": 0.01091650128364563,
-            "avg_think_len": 19.666534423828125,
-            "reward_format_mean": 0.9998354911804199,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.28125,
-            "reward_response_efficiency_mean": 0.8929828405380249,
-            "step": 327
-        },
-        {
-            "loss": 3.860564902424812e-05,
-            "reward_mean": 0.8869773149490356,
-            "reward_var": 0.012313231825828552,
-            "avg_think_len": 19.47096061706543,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9447842836380005,
-            "step": 328
-        },
-        {
-            "loss": 4.535354673862457e-05,
-            "reward_mean": 0.8653249740600586,
-            "reward_var": 0.011794047430157661,
-            "avg_think_len": 19.636720657348633,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.8831750750541687,
-            "step": 329
-        },
-        {
-            "loss": 4.580459790304303e-05,
-            "reward_mean": 0.8669397234916687,
-            "reward_var": 0.011993715539574623,
-            "avg_think_len": 19.676332473754883,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9677591323852539,
-            "step": 330
-        },
-        {
-            "loss": 3.837421536445618e-05,
-            "reward_mean": 0.8726027011871338,
-            "reward_var": 0.02359873615205288,
-            "avg_think_len": 19.733367919921875,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.965624988079071,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9466609358787537,
-            "step": 331
-        },
-        {
-            "loss": 4.0951184928417206e-05,
-            "reward_mean": 0.9046326875686646,
-            "reward_var": 0.011483115144073963,
-            "avg_think_len": 19.268245697021484,
-            "reward_format_mean": 0.9986978769302368,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9635828137397766,
-            "step": 332
-        },
-        {
-            "loss": 4.188274033367634e-05,
-            "reward_mean": 0.8630773425102234,
-            "reward_var": 0.021931663155555725,
-            "avg_think_len": 19.26033592224121,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.889809250831604,
-            "step": 333
-        },
-        {
-            "loss": 4.393143899505958e-05,
-            "reward_mean": 0.8631026744842529,
-            "reward_var": 0.025221893563866615,
-            "avg_think_len": 19.76863670349121,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.890625,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.936785876750946,
-            "step": 334
-        },
-        {
-            "loss": 3.963429480791092e-05,
-            "reward_mean": 0.901593029499054,
-            "reward_var": 0.010195292532444,
-            "avg_think_len": 19.774572372436523,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9969971179962158,
-            "step": 335
-        },
-        {
-            "loss": 8.921137487050146e-05,
-            "reward_mean": 0.9033202528953552,
-            "reward_var": 0.011599330231547356,
-            "avg_think_len": 19.68967056274414,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9601559042930603,
-            "step": 336
-        },
-        {
-            "loss": 3.6300625652074814e-05,
-            "reward_mean": 0.8820814490318298,
-            "reward_var": 0.01782788708806038,
-            "avg_think_len": 19.76788330078125,
-            "reward_format_mean": 0.9992559552192688,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.8603196144104004,
-            "step": 337
-        },
-        {
-            "loss": 3.1800009310245514e-05,
-            "reward_mean": 0.8255518078804016,
-            "reward_var": 0.024498142302036285,
-            "avg_think_len": 19.62552261352539,
-            "reward_format_mean": 0.996874988079071,
-            "reward_time_mean": 0.934374988079071,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.8709573149681091,
-            "step": 338
-        },
-        {
-            "loss": 2.1631247363984585e-05,
-            "reward_mean": 0.8678321242332458,
-            "reward_var": 0.012292226776480675,
-            "avg_think_len": 19.64860725402832,
-            "reward_format_mean": 0.999218761920929,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9564849138259888,
-            "step": 339
-        },
-        {
-            "loss": 6.902776658535004e-05,
-            "reward_mean": 0.8842005729675293,
-            "reward_var": 0.01026892103254795,
-            "avg_think_len": 19.8344669342041,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9336774349212646,
-            "step": 340
-        },
-        {
-            "loss": 4.3237581849098206e-05,
-            "reward_mean": 0.8779977560043335,
-            "reward_var": 0.0315416157245636,
-            "avg_think_len": 19.271831512451172,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.918241024017334,
-            "step": 341
-        },
-        {
-            "loss": 7.061136420816183e-05,
-            "reward_mean": 0.924504280090332,
-            "reward_var": 0.006790555082261562,
-            "avg_think_len": 19.573745727539062,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9948920607566833,
-            "step": 342
-        },
-        {
-            "loss": 1.733819954097271e-05,
-            "reward_mean": 0.9499961137771606,
-            "reward_var": 0.005929071456193924,
-            "avg_think_len": 19.741483688354492,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9906250238418579,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.981234610080719,
-            "step": 343
-        },
-        {
-            "loss": 3.448408097028732e-05,
-            "reward_mean": 0.8277114629745483,
-            "reward_var": 0.015139682218432426,
-            "avg_think_len": 19.57022476196289,
-            "reward_format_mean": 0.9998337626457214,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.8110120892524719,
-            "step": 344
-        },
-        {
-            "loss": 5.23092458024621e-05,
-            "reward_mean": 0.8636689782142639,
-            "reward_var": 0.02538837492465973,
-            "avg_think_len": 19.548200607299805,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8843749761581421,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.8828006982803345,
-            "step": 345
-        },
-        {
-            "loss": 1.8847640603780746e-05,
-            "reward_mean": 0.8805934190750122,
-            "reward_var": 0.011100249364972115,
-            "avg_think_len": 19.66412353515625,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.8504988551139832,
-            "step": 346
-        },
-        {
-            "loss": 6.630818825215101e-05,
-            "reward_mean": 0.8972488641738892,
-            "reward_var": 0.020990362390875816,
-            "avg_think_len": 19.724294662475586,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9171207547187805,
-            "step": 347
-        },
-        {
-            "loss": 4.1305553168058395e-05,
-            "reward_mean": 0.8623629212379456,
-            "reward_var": 0.017984038218855858,
-            "avg_think_len": 19.858877182006836,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9781249761581421,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.8775767087936401,
-            "step": 348
-        },
-        {
-            "loss": 2.9866117984056473e-05,
-            "reward_mean": 0.8458226919174194,
-            "reward_var": 0.031457286328077316,
-            "avg_think_len": 19.61614227294922,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.90625,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.8832907676696777,
-            "step": 349
-        },
-        {
-            "loss": 3.32733616232872e-05,
-            "reward_mean": 0.9216230511665344,
-            "reward_var": 0.008296589367091656,
-            "avg_think_len": 19.302095413208008,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.859375,
-            "reward_response_efficiency_mean": 0.8271172046661377,
-            "step": 350
-        },
-        {
-            "loss": 2.7176691219210625e-05,
-            "reward_mean": 0.9064250588417053,
-            "reward_var": 0.02341976948082447,
-            "avg_think_len": 19.677350997924805,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9694503545761108,
-            "step": 351
-        },
-        {
-            "loss": 7.263757288455963e-05,
-            "reward_mean": 0.8975852727890015,
-            "reward_var": 0.027193337678909302,
-            "avg_think_len": 19.483606338500977,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9497163891792297,
-            "step": 352
-        },
-        {
-            "loss": 4.0899962186813354e-05,
-            "reward_mean": 0.933242678642273,
-            "reward_var": 0.013471540994942188,
-            "avg_think_len": 19.58091163635254,
-            "reward_format_mean": 0.9989583492279053,
-            "reward_time_mean": 0.9781249761581421,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.9277624487876892,
-            "step": 353
-        },
-        {
-            "loss": 3.3429707400500774e-05,
-            "reward_mean": 0.8820857405662537,
-            "reward_var": 0.015502718277275562,
-            "avg_think_len": 19.614749908447266,
-            "reward_format_mean": 0.9997596144676208,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.8567082285881042,
-            "step": 354
-        },
-        {
-            "loss": 2.9851216822862625e-05,
-            "reward_mean": 0.9141159057617188,
-            "reward_var": 0.006961919367313385,
-            "avg_think_len": 19.58974838256836,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.968963623046875,
-            "step": 355
-        },
-        {
-            "loss": 4.900852218270302e-05,
-            "reward_mean": 0.8957276940345764,
-            "reward_var": 0.013049906119704247,
-            "avg_think_len": 19.75977325439453,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9891608357429504,
-            "step": 356
-        },
-        {
-            "loss": 3.769947215914726e-05,
-            "reward_mean": 0.9235654473304749,
-            "reward_var": 0.010002583265304565,
-            "avg_think_len": 19.428373336791992,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9755116701126099,
-            "step": 357
-        },
-        {
-            "loss": 3.440561704337597e-05,
-            "reward_mean": 0.8542783260345459,
-            "reward_var": 0.014894220978021622,
-            "avg_think_len": 19.44040870666504,
-            "reward_format_mean": 0.9993990659713745,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.484375,
-            "reward_response_efficiency_mean": 0.9489644169807434,
-            "step": 358
-        },
-        {
-            "loss": 6.585102528333664e-05,
-            "reward_mean": 0.8914174437522888,
-            "reward_var": 0.012423601001501083,
-            "avg_think_len": 19.700960159301758,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.987500011920929,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9844196438789368,
-            "step": 359
-        },
-        {
-            "loss": 8.030561730265617e-05,
-            "reward_mean": 0.9399334192276001,
-            "reward_var": 0.00939161330461502,
-            "avg_think_len": 19.81818389892578,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9906250238418579,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9566085934638977,
-            "step": 360
-        },
-        {
-            "loss": 4.054838791489601e-05,
-            "reward_mean": 0.8767420053482056,
-            "reward_var": 0.01330372504889965,
-            "avg_think_len": 19.704843521118164,
-            "reward_format_mean": 0.9984375238418579,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9772806167602539,
-            "step": 361
-        },
-        {
-            "loss": 4.69135120511055e-05,
-            "reward_mean": 0.8885654807090759,
-            "reward_var": 0.012854175642132759,
-            "avg_think_len": 19.607402801513672,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.987500011920929,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.97301185131073,
-            "step": 362
-        },
-        {
-            "loss": 9.921379387378693e-06,
-            "reward_mean": 0.8550281524658203,
-            "reward_var": 0.016728850081562996,
-            "avg_think_len": 19.748960494995117,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.965624988079071,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.8451126217842102,
-            "step": 363
-        },
-        {
-            "loss": 4.1725579649209976e-05,
-            "reward_mean": 0.8961427211761475,
-            "reward_var": 0.0308262649923563,
-            "avg_think_len": 19.706890106201172,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9283207654953003,
-            "step": 364
-        },
-        {
-            "loss": 4.650023765861988e-05,
-            "reward_mean": 0.9050276279449463,
-            "reward_var": 0.01877155900001526,
-            "avg_think_len": 19.49279022216797,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9326105117797852,
-            "step": 365
-        },
-        {
-            "loss": 0.00016647088341414928,
-            "reward_mean": 0.919948935508728,
-            "reward_var": 0.009037558920681477,
-            "avg_think_len": 19.786645889282227,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9610458016395569,
-            "step": 366
-        },
-        {
-            "loss": 4.6741683036088943e-05,
-            "reward_mean": 0.8608707189559937,
-            "reward_var": 0.018810804933309555,
-            "avg_think_len": 19.70323371887207,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.7872329354286194,
-            "step": 367
-        },
-        {
-            "loss": 4.1175633668899536e-05,
-            "reward_mean": 0.888981282711029,
-            "reward_var": 0.009614049457013607,
-            "avg_think_len": 19.68433952331543,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9778000116348267,
-            "step": 368
-        },
-        {
-            "loss": 3.815256059169769e-05,
-            "reward_mean": 0.8818032145500183,
-            "reward_var": 0.01921064779162407,
-            "avg_think_len": 19.771320343017578,
-            "reward_format_mean": 0.9989583492279053,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9501295685768127,
-            "step": 369
-        },
-        {
-            "loss": 3.5779550671577454e-05,
-            "reward_mean": 0.8767154216766357,
-            "reward_var": 0.014581079594790936,
-            "avg_think_len": 19.45549201965332,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.7568618655204773,
-            "step": 370
-        },
-        {
-            "loss": 4.1126273572444916e-05,
-            "reward_mean": 0.8301161527633667,
-            "reward_var": 0.012893098406493664,
-            "avg_think_len": 19.845165252685547,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.965624988079071,
-            "reward_correctness_mean": 0.453125,
-            "reward_response_efficiency_mean": 0.9017148017883301,
-            "step": 371
-        },
-        {
-            "loss": 1.4728051610291004e-05,
-            "reward_mean": 0.9285823106765747,
-            "reward_var": 0.024394836276769638,
-            "avg_think_len": 19.608333587646484,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.933079183101654,
-            "step": 372
-        },
-        {
-            "loss": 2.628704532980919e-05,
-            "reward_mean": 0.9219031929969788,
-            "reward_var": 0.02364412136375904,
-            "avg_think_len": 19.935649871826172,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9376126527786255,
-            "step": 373
-        },
-        {
-            "loss": 4.8473943024873734e-05,
-            "reward_mean": 0.9147375822067261,
-            "reward_var": 0.01124698854982853,
-            "avg_think_len": 19.816282272338867,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9870752096176147,
-            "step": 374
-        },
-        {
-            "loss": 5.354185122996569e-05,
-            "reward_mean": 0.9313425421714783,
-            "reward_var": 0.024511050432920456,
-            "avg_think_len": 19.736862182617188,
-            "reward_format_mean": 0.9993749856948853,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9603700041770935,
-            "step": 375
-        },
-        {
-            "loss": 4.1703227907419205e-05,
-            "reward_mean": 0.934030294418335,
-            "reward_var": 0.007048503495752811,
-            "avg_think_len": 19.40534210205078,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.907995879650116,
-            "step": 376
-        },
-        {
-            "loss": 4.5968685299158096e-05,
-            "reward_mean": 0.8988460302352905,
-            "reward_var": 0.014754736796021461,
-            "avg_think_len": 19.750385284423828,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9078841209411621,
-            "step": 377
-        },
-        {
-            "loss": 6.382912397384644e-05,
-            "reward_mean": 0.9284536242485046,
-            "reward_var": 0.010533700697124004,
-            "avg_think_len": 19.73834800720215,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.987500011920929,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9763143658638,
-            "step": 378
-        },
-        {
-            "loss": 4.820898175239563e-05,
-            "reward_mean": 0.8422837853431702,
-            "reward_var": 0.02796921879053116,
-            "avg_think_len": 19.611610412597656,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9312500357627869,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9066351056098938,
-            "step": 379
-        },
-        {
-            "loss": 3.944197669625282e-05,
-            "reward_mean": 0.7995866537094116,
-            "reward_var": 0.032020796090364456,
-            "avg_think_len": 19.781225204467773,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9125000238418579,
-            "reward_correctness_mean": 0.375,
-            "reward_response_efficiency_mean": 0.9108465313911438,
-            "step": 380
-        },
-        {
-            "loss": 4.952907329425216e-05,
-            "reward_mean": 0.9034271240234375,
-            "reward_var": 0.024897603318095207,
-            "avg_think_len": 19.68795394897461,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.95745849609375,
-            "step": 381
-        },
-        {
-            "loss": 4.314212128520012e-05,
-            "reward_mean": 0.9585378170013428,
-            "reward_var": 0.013638745062053204,
-            "avg_think_len": 19.644702911376953,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.90625,
-            "reward_response_efficiency_mean": 0.959151566028595,
-            "step": 382
-        },
-        {
-            "loss": 3.568106330931187e-05,
-            "reward_mean": 0.9031069278717041,
-            "reward_var": 0.016900958493351936,
-            "avg_think_len": 19.7214298248291,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.8780528903007507,
-            "step": 383
-        },
-        {
-            "loss": 3.428105264902115e-05,
-            "reward_mean": 0.8786014914512634,
-            "reward_var": 0.008853391744196415,
-            "avg_think_len": 19.853763580322266,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9362810850143433,
-            "step": 384
-        },
-        {
-            "loss": 4.369579255580902e-05,
-            "reward_mean": 0.9319255352020264,
-            "reward_var": 0.022173959761857986,
-            "avg_think_len": 19.709884643554688,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.859375,
-            "reward_response_efficiency_mean": 0.8995768427848816,
-            "step": 385
-        },
-        {
-            "loss": 4.74429689347744e-05,
-            "reward_mean": 0.8493735790252686,
-            "reward_var": 0.013774554245173931,
-            "avg_think_len": 19.860576629638672,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9750000238418579,
-            "reward_correctness_mean": 0.453125,
-            "reward_response_efficiency_mean": 0.9693694114685059,
-            "step": 386
-        },
-        {
-            "loss": 3.817956894636154e-05,
-            "reward_mean": 0.9057325720787048,
-            "reward_var": 0.013309132307767868,
-            "avg_think_len": 19.68425941467285,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9698053598403931,
-            "step": 387
-        },
-        {
-            "loss": 4.967954009771347e-05,
-            "reward_mean": 0.8962685465812683,
-            "reward_var": 0.022366918623447418,
-            "avg_think_len": 19.703466415405273,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.940625011920929,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9569491147994995,
-            "step": 388
-        },
-        {
-            "loss": 5.24008646607399e-05,
-            "reward_mean": 0.9227035641670227,
-            "reward_var": 0.00856756791472435,
-            "avg_think_len": 19.95482063293457,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.8970639109611511,
-            "step": 389
-        },
-        {
-            "loss": 4.075956530869007e-05,
-            "reward_mean": 0.8518418073654175,
-            "reward_var": 0.016240321099758148,
-            "avg_think_len": 19.803680419921875,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9750000238418579,
-            "reward_correctness_mean": 0.484375,
-            "reward_response_efficiency_mean": 0.9479920268058777,
-            "step": 390
-        },
-        {
-            "loss": 1.2398231774568558e-05,
-            "reward_mean": 0.9262474179267883,
-            "reward_var": 0.00698811886832118,
-            "avg_think_len": 19.91094970703125,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9862393140792847,
-            "step": 391
-        },
-        {
-            "loss": 3.3768825232982635e-05,
-            "reward_mean": 0.9290317296981812,
-            "reward_var": 0.008396441116929054,
-            "avg_think_len": 19.918237686157227,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9973770976066589,
-            "step": 392
-        },
-        {
-            "loss": 5.1992014050483704e-05,
-            "reward_mean": 0.9145897626876831,
-            "reward_var": 0.00680141057819128,
-            "avg_think_len": 19.826953887939453,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9864841103553772,
-            "step": 393
-        },
-        {
-            "loss": 4.975422052666545e-05,
-            "reward_mean": 0.8936055302619934,
-            "reward_var": 0.01214312668889761,
-            "avg_think_len": 19.919139862060547,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9781249761581421,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9869216680526733,
-            "step": 394
-        },
-        {
-            "loss": 3.853626549243927e-05,
-            "reward_mean": 0.9082131385803223,
-            "reward_var": 0.025934791192412376,
-            "avg_think_len": 19.68440055847168,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.940625011920929,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9422275424003601,
-            "step": 395
-        },
-        {
-            "loss": 2.792244777083397e-05,
-            "reward_mean": 0.9045020341873169,
-            "reward_var": 0.0017730703111737967,
-            "avg_think_len": 19.817651748657227,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9773832559585571,
-            "step": 396
-        },
-        {
-            "loss": 4.801573231816292e-05,
-            "reward_mean": 0.9096477031707764,
-            "reward_var": 0.007098745089024305,
-            "avg_think_len": 20.084177017211914,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9979655742645264,
-            "step": 397
-        },
-        {
-            "loss": 4.732434172183275e-05,
-            "reward_mean": 0.9186035394668579,
-            "reward_var": 0.007693212013691664,
-            "avg_think_len": 19.8566951751709,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.987500011920929,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9212889671325684,
-            "step": 398
-        },
-        {
-            "loss": 3.888411447405815e-05,
-            "reward_mean": 0.9326907396316528,
-            "reward_var": 0.013846738263964653,
-            "avg_think_len": 19.88471794128418,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.987500011920929,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9620131254196167,
-            "step": 399
-        },
-        {
-            "loss": 4.4253189116716385e-05,
-            "reward_mean": 0.8880592584609985,
-            "reward_var": 0.008061670698225498,
-            "avg_think_len": 19.71287727355957,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9897370338439941,
-            "step": 400
-        },
-        {
-            "loss": 4.570581950247288e-05,
-            "reward_mean": 0.9070066213607788,
-            "reward_var": 0.014661259949207306,
-            "avg_think_len": 19.758201599121094,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.8936513066291809,
-            "step": 401
-        },
-        {
-            "loss": 2.7515459805727005e-05,
-            "reward_mean": 0.9245793223381042,
-            "reward_var": 0.011658195406198502,
-            "avg_think_len": 19.632530212402344,
-            "reward_format_mean": 0.9994611740112305,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9832307696342468,
-            "step": 402
-        },
-        {
-            "loss": 4.874297883361578e-05,
-            "reward_mean": 0.9161832928657532,
-            "reward_var": 0.014468548819422722,
-            "avg_think_len": 19.863849639892578,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9772331714630127,
-            "step": 403
-        },
-        {
-            "loss": 2.3427419364452362e-05,
-            "reward_mean": 0.9084482192993164,
-            "reward_var": 0.013342069461941719,
-            "avg_think_len": 19.548707962036133,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9650430083274841,
-            "step": 404
-        },
-        {
-            "loss": 2.8100330382585526e-05,
-            "reward_mean": 0.906502366065979,
-            "reward_var": 0.016326094046235085,
-            "avg_think_len": 19.68236541748047,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9697593450546265,
-            "step": 405
-        },
-        {
-            "loss": 5.96935860812664e-05,
-            "reward_mean": 0.8790698647499084,
-            "reward_var": 0.010683895088732243,
-            "avg_think_len": 19.72271728515625,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.9850293397903442,
-            "step": 406
-        },
-        {
-            "loss": 3.7038931623101234e-05,
-            "reward_mean": 0.8990211486816406,
-            "reward_var": 0.012115188874304295,
-            "avg_think_len": 19.538890838623047,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.971084475517273,
-            "step": 407
-        },
-        {
-            "loss": 1.314561814069748e-06,
-            "reward_mean": 0.8850356936454773,
-            "reward_var": 0.012230885215103626,
-            "avg_think_len": 19.876325607299805,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9463927149772644,
-            "step": 408
-        },
-        {
-            "loss": 1.1368072591722012e-05,
-            "reward_mean": 0.9189023375511169,
-            "reward_var": 0.011916739866137505,
-            "avg_think_len": 19.69725227355957,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9412339925765991,
-            "step": 409
-        },
-        {
-            "loss": 4.002125933766365e-05,
-            "reward_mean": 0.9105644226074219,
-            "reward_var": 0.0069411806762218475,
-            "avg_think_len": 19.733781814575195,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.9860077500343323,
-            "step": 410
-        },
-        {
-            "loss": 3.830238711088896e-05,
-            "reward_mean": 0.8833062052726746,
-            "reward_var": 0.008077388629317284,
-            "avg_think_len": 19.793344497680664,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.8769747614860535,
-            "step": 411
-        },
-        {
-            "loss": 5.2741263061761856e-05,
-            "reward_mean": 0.9030210971832275,
-            "reward_var": 0.0188089981675148,
-            "avg_think_len": 19.88550567626953,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9750000238418579,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9339590072631836,
-            "step": 412
-        },
-        {
-            "loss": 2.6082154363393784e-05,
-            "reward_mean": 0.9013457298278809,
-            "reward_var": 0.014442844316363335,
-            "avg_think_len": 19.882442474365234,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9803828597068787,
-            "step": 413
-        },
-        {
-            "loss": 5.585048347711563e-05,
-            "reward_mean": 0.8984872102737427,
-            "reward_var": 0.019703710451722145,
-            "avg_think_len": 19.77189064025879,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.971875011920929,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9345738887786865,
-            "step": 414
-        },
-        {
-            "loss": 4.391488619148731e-05,
-            "reward_mean": 0.9152179956436157,
-            "reward_var": 0.010101068764925003,
-            "avg_think_len": 19.807777404785156,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.973372220993042,
-            "step": 415
-        },
-        {
-            "loss": 3.798764373641461e-05,
-            "reward_mean": 0.8568681478500366,
-            "reward_var": 0.016725759953260422,
-            "avg_think_len": 19.5751895904541,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9430978298187256,
-            "step": 416
-        },
-        {
-            "loss": 3.673834726214409e-05,
-            "reward_mean": 0.9365928173065186,
-            "reward_var": 0.01414775475859642,
-            "avg_think_len": 19.960474014282227,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9750000238418579,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9901211857795715,
-            "step": 417
-        },
-        {
-            "loss": 8.152936061378568e-05,
-            "reward_mean": 0.9006395936012268,
-            "reward_var": 0.0103303212672472,
-            "avg_think_len": 19.843883514404297,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.987500011920929,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9119334816932678,
-            "step": 418
-        },
-        {
-            "loss": 4.461314529180527e-05,
-            "reward_mean": 0.9254275560379028,
-            "reward_var": 0.022580888122320175,
-            "avg_think_len": 19.81172752380371,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9829607009887695,
-            "step": 419
-        },
-        {
-            "loss": 4.12757508456707e-05,
-            "reward_mean": 0.9458613395690918,
-            "reward_var": 0.010580237954854965,
-            "avg_think_len": 19.524627685546875,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.9709452986717224,
-            "step": 420
-        },
-        {
-            "loss": 5.399511428549886e-05,
-            "reward_mean": 0.9121628999710083,
-            "reward_var": 0.0130409374833107,
-            "avg_think_len": 19.590103149414062,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9299018383026123,
-            "step": 421
-        },
-        {
-            "loss": 4.486153193283826e-05,
-            "reward_mean": 0.8697519302368164,
-            "reward_var": 0.020486488938331604,
-            "avg_think_len": 19.568111419677734,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.916507363319397,
-            "step": 422
-        },
-        {
-            "loss": 4.144664853811264e-05,
-            "reward_mean": 0.8841778039932251,
-            "reward_var": 0.024496857076883316,
-            "avg_think_len": 19.785829544067383,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.875,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9117112755775452,
-            "step": 423
-        },
-        {
-            "loss": 4.3111853301525116e-05,
-            "reward_mean": 0.8910202980041504,
-            "reward_var": 0.014282491989433765,
-            "avg_think_len": 19.79845428466797,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9906249642372131,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9640811085700989,
-            "step": 424
-        },
-        {
-            "loss": 3.7631019949913025e-05,
-            "reward_mean": 0.9101915955543518,
-            "reward_var": 0.010533045046031475,
-            "avg_think_len": 19.729534149169922,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9407666921615601,
-            "step": 425
-        },
-        {
-            "loss": 5.414150655269623e-05,
-            "reward_mean": 0.8376168012619019,
-            "reward_var": 0.02058255299925804,
-            "avg_think_len": 19.916261672973633,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.949999988079071,
-            "reward_correctness_mean": 0.453125,
-            "reward_response_efficiency_mean": 0.9473419785499573,
-            "step": 426
-        },
-        {
-            "loss": 3.012828528881073e-05,
-            "reward_mean": 0.8881330490112305,
-            "reward_var": 0.014871123246848583,
-            "avg_think_len": 19.96894073486328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9900321960449219,
-            "step": 427
-        },
-        {
-            "loss": 2.098295954056084e-05,
-            "reward_mean": 0.8823273777961731,
-            "reward_var": 0.02191162295639515,
-            "avg_think_len": 19.849088668823242,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.8574346899986267,
-            "step": 428
-        },
-        {
-            "loss": 5.284557119011879e-05,
-            "reward_mean": 0.9144213795661926,
-            "reward_var": 0.019451690837740898,
-            "avg_think_len": 19.823516845703125,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9545605778694153,
-            "step": 429
-        },
-        {
-            "loss": 3.807526081800461e-05,
-            "reward_mean": 0.9261701703071594,
-            "reward_var": 0.022109735757112503,
-            "avg_think_len": 19.569913864135742,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.9234306812286377,
-            "step": 430
-        },
-        {
-            "loss": 5.3725321777164936e-05,
-            "reward_mean": 0.8916536569595337,
-            "reward_var": 0.010968961752951145,
-            "avg_think_len": 19.88003158569336,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9572397470474243,
-            "step": 431
-        },
-        {
-            "loss": 2.9470305889844894e-05,
-            "reward_mean": 0.9453324675559998,
-            "reward_var": 0.014657977037131786,
-            "avg_think_len": 19.74852180480957,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.9532045722007751,
-            "step": 432
-        },
-        {
-            "loss": 6.143830250948668e-05,
-            "reward_mean": 0.9461920261383057,
-            "reward_var": 0.008606343530118465,
-            "avg_think_len": 19.74570655822754,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.875,
-            "reward_response_efficiency_mean": 0.9253929257392883,
-            "step": 433
-        },
-        {
-            "loss": 3.549130633473396e-05,
-            "reward_mean": 0.9234561324119568,
-            "reward_var": 0.0070516509003937244,
-            "avg_think_len": 20.076356887817383,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9781996607780457,
-            "step": 434
-        },
-        {
-            "loss": 8.781393989920616e-05,
-            "reward_mean": 0.9137067794799805,
-            "reward_var": 0.017667347565293312,
-            "avg_think_len": 19.890531539916992,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9517021775245667,
-            "step": 435
-        },
-        {
-            "loss": 0.00014249293599277735,
-            "reward_mean": 0.8725283145904541,
-            "reward_var": 0.018346451222896576,
-            "avg_think_len": 19.876890182495117,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.90625,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9588631987571716,
-            "step": 436
-        },
-        {
-            "loss": 3.9211008697748184e-05,
-            "reward_mean": 0.8784116506576538,
-            "reward_var": 0.018845155835151672,
-            "avg_think_len": 19.878158569335938,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.8886467814445496,
-            "step": 437
-        },
-        {
-            "loss": 5.855131894350052e-05,
-            "reward_mean": 0.8968042135238647,
-            "reward_var": 0.014663312584161758,
-            "avg_think_len": 19.74884796142578,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9028423428535461,
-            "step": 438
-        },
-        {
-            "loss": 4.489067941904068e-05,
-            "reward_mean": 0.9694187641143799,
-            "reward_var": 0.005923558492213488,
-            "avg_think_len": 19.971704483032227,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.890625,
-            "reward_response_efficiency_mean": 0.9901753067970276,
-            "step": 439
-        },
-        {
-            "loss": 4.9254391342401505e-05,
-            "reward_mean": 0.9252332448959351,
-            "reward_var": 0.01963832601904869,
-            "avg_think_len": 19.784345626831055,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9665580987930298,
-            "step": 440
-        },
-        {
-            "loss": 0.00010695225500967354,
-            "reward_mean": 0.9369605779647827,
-            "reward_var": 0.006725592073053122,
-            "avg_think_len": 19.975374221801758,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9665921330451965,
-            "step": 441
-        },
-        {
-            "loss": 1.098494976758957e-06,
-            "reward_mean": 0.9120699167251587,
-            "reward_var": 0.01226160116493702,
-            "avg_think_len": 19.922481536865234,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.88265460729599,
-            "step": 442
-        },
-        {
-            "loss": 7.447833195328712e-05,
-            "reward_mean": 0.919570803642273,
-            "reward_var": 0.018495909869670868,
-            "avg_think_len": 19.921710968017578,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9595329165458679,
-            "step": 443
-        },
-        {
-            "loss": 3.4580472856760025e-05,
-            "reward_mean": 0.8892396688461304,
-            "reward_var": 0.01715216599404812,
-            "avg_think_len": 19.846385955810547,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.8850839138031006,
-            "step": 444
-        },
-        {
-            "loss": 4.8815272748470306e-05,
-            "reward_mean": 0.8979631066322327,
-            "reward_var": 0.016248982399702072,
-            "avg_think_len": 19.828998565673828,
-            "reward_format_mean": 0.9998337626457214,
-            "reward_time_mean": 0.965624988079071,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.845143735408783,
-            "step": 445
-        },
-        {
-            "loss": 2.7539790607988834e-05,
-            "reward_mean": 0.9220103621482849,
-            "reward_var": 0.016583813354372978,
-            "avg_think_len": 19.760976791381836,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.971875011920929,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9661665558815002,
-            "step": 446
-        },
-        {
-            "loss": 4.672026261687279e-05,
-            "reward_mean": 0.9065761566162109,
-            "reward_var": 0.009495800361037254,
-            "avg_think_len": 19.60599136352539,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9856799244880676,
-            "step": 447
-        },
-        {
-            "loss": 4.834914579987526e-05,
-            "reward_mean": 0.86695396900177,
-            "reward_var": 0.018955696374177933,
-            "avg_think_len": 19.80076026916504,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9678159952163696,
-            "step": 448
-        },
-        {
-            "loss": 6.255728658288717e-05,
-            "reward_mean": 0.8565671443939209,
-            "reward_var": 0.02197641134262085,
-            "avg_think_len": 19.72352409362793,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9624999761581421,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.854393482208252,
-            "step": 449
-        },
-        {
-            "loss": 4.1049206629395485e-05,
-            "reward_mean": 0.8794679641723633,
-            "reward_var": 0.025386279448866844,
-            "avg_think_len": 19.830368041992188,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9084966778755188,
-            "step": 450
-        },
-        {
-            "loss": 6.973673589527607e-05,
-            "reward_mean": 0.9095425009727478,
-            "reward_var": 0.005737547762691975,
-            "avg_think_len": 19.9132080078125,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9975449442863464,
-            "step": 451
-        },
-        {
-            "loss": 5.995016545057297e-05,
-            "reward_mean": 0.9118812084197998,
-            "reward_var": 0.009780598804354668,
-            "avg_think_len": 19.7432861328125,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9443997144699097,
-            "step": 452
-        },
-        {
-            "loss": 3.255670890212059e-06,
-            "reward_mean": 0.942078173160553,
-            "reward_var": 0.013216744177043438,
-            "avg_think_len": 19.91876792907715,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.987062931060791,
-            "step": 453
-        },
-        {
-            "loss": 3.8164667785167694e-05,
-            "reward_mean": 0.9211180806159973,
-            "reward_var": 0.007774102035909891,
-            "avg_think_len": 19.931785583496094,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9969724416732788,
-            "step": 454
-        },
-        {
-            "loss": 5.6765973567962646e-05,
-            "reward_mean": 0.9057769775390625,
-            "reward_var": 0.019072409719228745,
-            "avg_think_len": 19.798025131225586,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9356078505516052,
-            "step": 455
-        },
-        {
-            "loss": 5.4165953770279884e-05,
-            "reward_mean": 0.9079842567443848,
-            "reward_var": 0.010781561955809593,
-            "avg_think_len": 19.6469669342041,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9756870269775391,
-            "step": 456
-        },
-        {
-            "loss": 3.7499587051570415e-05,
-            "reward_mean": 0.8666459918022156,
-            "reward_var": 0.03257419541478157,
-            "avg_think_len": 19.66286849975586,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.893750011920929,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.8853340148925781,
-            "step": 457
-        },
-        {
-            "loss": 6.506266072392464e-05,
-            "reward_mean": 0.9062055349349976,
-            "reward_var": 0.009693816304206848,
-            "avg_think_len": 19.965038299560547,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.9748219847679138,
-            "step": 458
-        },
-        {
-            "loss": 1.543620601296425e-05,
-            "reward_mean": 0.8880313038825989,
-            "reward_var": 0.022075429558753967,
-            "avg_think_len": 19.978185653686523,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.949999988079071,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9146252870559692,
-            "step": 459
-        },
-        {
-            "loss": 3.762543201446533e-05,
-            "reward_mean": 0.9584645628929138,
-            "reward_var": 0.01333686988800764,
-            "avg_think_len": 19.994380950927734,
-            "reward_format_mean": 0.9994791746139526,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.90625,
-            "reward_response_efficiency_mean": 0.9437541961669922,
-            "step": 460
-        },
-        {
-            "loss": 6.447732448577881e-05,
-            "reward_mean": 0.966962456703186,
-            "reward_var": 0.004751844797283411,
-            "avg_think_len": 19.626142501831055,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.890625,
-            "reward_response_efficiency_mean": 0.9772250652313232,
-            "step": 461
-        },
-        {
-            "loss": 5.144649185240269e-05,
-            "reward_mean": 0.9029447436332703,
-            "reward_var": 0.0166903305798769,
-            "avg_think_len": 19.870779037475586,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9750000238418579,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9336538910865784,
-            "step": 462
-        },
-        {
-            "loss": 7.465877570211887e-05,
-            "reward_mean": 0.9213969111442566,
-            "reward_var": 0.011699219234287739,
-            "avg_think_len": 19.69448471069336,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.926212728023529,
-            "step": 463
-        },
-        {
-            "loss": 3.665406256914139e-05,
-            "reward_mean": 0.9228934049606323,
-            "reward_var": 0.014779506251215935,
-            "avg_think_len": 20.092872619628906,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9728235602378845,
-            "step": 464
-        },
-        {
-            "loss": 3.13029158860445e-05,
-            "reward_mean": 0.9258864521980286,
-            "reward_var": 0.009956574067473412,
-            "avg_think_len": 19.71514320373535,
-            "reward_format_mean": 0.9998117685317993,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.8756095170974731,
-            "step": 465
-        },
-        {
-            "loss": 5.422509275376797e-05,
-            "reward_mean": 0.8828083276748657,
-            "reward_var": 0.03217476233839989,
-            "avg_think_len": 19.76193618774414,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9062333106994629,
-            "step": 466
-        },
-        {
-            "loss": 7.819803431630135e-05,
-            "reward_mean": 0.9618459939956665,
-            "reward_var": 0.00828600861132145,
-            "avg_think_len": 19.964345932006836,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.859375,
-            "reward_response_efficiency_mean": 0.9880086779594421,
-            "step": 467
-        },
-        {
-            "loss": 5.261879414319992e-05,
-            "reward_mean": 0.9394612312316895,
-            "reward_var": 0.0082209762185812,
-            "avg_think_len": 19.944950103759766,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9797197580337524,
-            "step": 468
-        },
-        {
-            "loss": 4.461035132408142e-05,
-            "reward_mean": 0.9659682512283325,
-            "reward_var": 0.006292285863310099,
-            "avg_think_len": 19.94292640686035,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.890625,
-            "reward_response_efficiency_mean": 0.9732481837272644,
-            "step": 469
-        },
-        {
-            "loss": 6.695184856653214e-05,
-            "reward_mean": 0.8940539956092834,
-            "reward_var": 0.013552683405578136,
-            "avg_think_len": 19.995256423950195,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9980911612510681,
-            "step": 470
-        },
-        {
-            "loss": 2.445071004331112e-05,
-            "reward_mean": 0.8648049831390381,
-            "reward_var": 0.015430103056132793,
-            "avg_think_len": 19.83599281311035,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.9748448729515076,
-            "step": 471
-        },
-        {
-            "loss": 0.00011142785660922527,
-            "reward_mean": 0.8862766027450562,
-            "reward_var": 0.012299190275371075,
-            "avg_think_len": 19.824533462524414,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9513564109802246,
-            "step": 472
-        },
-        {
-            "loss": 8.148816414177418e-05,
-            "reward_mean": 0.9092778563499451,
-            "reward_var": 0.021923692896962166,
-            "avg_think_len": 19.879013061523438,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9339863061904907,
-            "step": 473
-        },
-        {
-            "loss": 7.897987961769104e-05,
-            "reward_mean": 0.9136167764663696,
-            "reward_var": 0.011748920194804668,
-            "avg_think_len": 19.738338470458984,
-            "reward_format_mean": 0.9998553395271301,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.8796120882034302,
-            "step": 474
-        },
-        {
-            "loss": 3.5152770578861237e-05,
-            "reward_mean": 0.8494949340820312,
-            "reward_var": 0.0074028451927006245,
-            "avg_think_len": 19.868762969970703,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.421875,
-            "reward_response_efficiency_mean": 0.9761049747467041,
-            "step": 475
-        },
-        {
-            "loss": 6.02498184889555e-05,
-            "reward_mean": 0.9329198002815247,
-            "reward_var": 0.012634338811039925,
-            "avg_think_len": 19.933258056640625,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9816792607307434,
-            "step": 476
-        },
-        {
-            "loss": 4.2464002035558224e-05,
-            "reward_mean": 0.8271296620368958,
-            "reward_var": 0.018269231542944908,
-            "avg_think_len": 19.828533172607422,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9906250238418579,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.7241437435150146,
-            "step": 477
-        },
-        {
-            "loss": 4.597799852490425e-05,
-            "reward_mean": 0.8901934027671814,
-            "reward_var": 0.010426554828882217,
-            "avg_think_len": 20.097366333007812,
-            "reward_format_mean": 0.9991776347160339,
-            "reward_time_mean": 0.9906250238418579,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9615963697433472,
-            "step": 478
-        },
-        {
-            "loss": 7.313070818781853e-05,
-            "reward_mean": 0.8616881370544434,
-            "reward_var": 0.02244938537478447,
-            "avg_think_len": 19.997344970703125,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.965624988079071,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9655025005340576,
-            "step": 479
-        },
-        {
-            "loss": 7.869116961956024e-05,
-            "reward_mean": 0.9186766147613525,
-            "reward_var": 0.01206020638346672,
-            "avg_think_len": 19.693580627441406,
-            "reward_format_mean": 0.9998686909675598,
-            "reward_time_mean": 0.9750000238418579,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9654624462127686,
-            "step": 480
-        },
-        {
-            "loss": 8.413864998146892e-05,
-            "reward_mean": 0.8703961372375488,
-            "reward_var": 0.006283564958721399,
-            "avg_think_len": 19.885147094726562,
-            "reward_format_mean": 0.9989583492279053,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.484375,
-            "reward_response_efficiency_mean": 0.9982512593269348,
-            "step": 481
-        },
-        {
-            "loss": 5.6252931244671345e-05,
-            "reward_mean": 0.8589581251144409,
-            "reward_var": 0.0055833421647548676,
-            "avg_think_len": 20.103818893432617,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.4375,
-            "reward_response_efficiency_mean": 0.9983327984809875,
-            "step": 482
-        },
-        {
-            "loss": 0.00010250462219119072,
-            "reward_mean": 0.9274094104766846,
-            "reward_var": 0.008478578180074692,
-            "avg_think_len": 19.934844970703125,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9596375823020935,
-            "step": 483
-        },
-        {
-            "loss": 1.6804784536361694e-05,
-            "reward_mean": 0.8899657726287842,
-            "reward_var": 0.005843807477504015,
-            "avg_think_len": 19.856945037841797,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9973630309104919,
-            "step": 484
-        },
-        {
-            "loss": 0.00012795208021998405,
-            "reward_mean": 0.9547804594039917,
-            "reward_var": 0.003962322603911161,
-            "avg_think_len": 19.834054946899414,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.9784966707229614,
-            "step": 485
-        },
-        {
-            "loss": 2.7840957045555115e-05,
-            "reward_mean": 0.8896687626838684,
-            "reward_var": 0.00989848468452692,
-            "avg_think_len": 20.064308166503906,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9649250507354736,
-            "step": 486
-        },
-        {
-            "loss": 5.642324686050415e-05,
-            "reward_mean": 0.8589166402816772,
-            "reward_var": 0.01625777781009674,
-            "avg_think_len": 19.901195526123047,
-            "reward_format_mean": 0.9991319179534912,
-            "reward_time_mean": 0.987500011920929,
-            "reward_correctness_mean": 0.484375,
-            "reward_response_efficiency_mean": 0.964659571647644,
-            "step": 487
-        },
-        {
-            "loss": 9.105727076530457e-05,
-            "reward_mean": 0.888776421546936,
-            "reward_var": 0.008477160707116127,
-            "avg_think_len": 19.97047233581543,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9613555669784546,
-            "step": 488
-        },
-        {
-            "loss": 6.86800922267139e-05,
-            "reward_mean": 0.9428228735923767,
-            "reward_var": 0.005064234137535095,
-            "avg_think_len": 20.010082244873047,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.974416196346283,
-            "step": 489
-        },
-        {
-            "loss": 5.558924749493599e-05,
-            "reward_mean": 0.9530274868011475,
-            "reward_var": 0.0082959970459342,
-            "avg_think_len": 19.70298957824707,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.875,
-            "reward_response_efficiency_mean": 0.9527347087860107,
-            "step": 490
-        },
-        {
-            "loss": 6.992649286985397e-05,
-            "reward_mean": 0.8818283081054688,
-            "reward_var": 0.006172745488584042,
-            "avg_think_len": 19.756303787231445,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9554383754730225,
-            "step": 491
-        },
-        {
-            "loss": 9.123235940933228e-05,
-            "reward_mean": 0.8807119727134705,
-            "reward_var": 0.009747888892889023,
-            "avg_think_len": 19.777141571044922,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.975972592830658,
-            "step": 492
-        },
-        {
-            "loss": 7.833901327103376e-05,
-            "reward_mean": 0.8879071474075317,
-            "reward_var": 0.01013962272554636,
-            "avg_think_len": 19.770824432373047,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9735034704208374,
-            "step": 493
-        },
-        {
-            "loss": 3.7701334804296494e-05,
-            "reward_mean": 0.9162071347236633,
-            "reward_var": 0.017635919153690338,
-            "avg_think_len": 19.775936126708984,
-            "reward_format_mean": 0.999804675579071,
-            "reward_time_mean": 0.9624999761581421,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.936898946762085,
-            "step": 494
-        },
-        {
-            "loss": 2.0605511963367462e-05,
-            "reward_mean": 0.8992611765861511,
-            "reward_var": 0.008968651294708252,
-            "avg_think_len": 19.872697830200195,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9564197063446045,
-            "step": 495
-        },
-        {
-            "loss": 3.3180927857756615e-05,
-            "reward_mean": 0.8897787928581238,
-            "reward_var": 0.013218306936323643,
-            "avg_think_len": 20.02341651916504,
-            "reward_format_mean": 0.9989583492279053,
-            "reward_time_mean": 0.9781249761581421,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9726569056510925,
-            "step": 496
-        },
-        {
-            "loss": 4.175398498773575e-05,
-            "reward_mean": 0.8670286536216736,
-            "reward_var": 0.011208820156753063,
-            "avg_think_len": 20.054372787475586,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.9681143760681152,
-            "step": 497
-        },
-        {
-            "loss": 4.483573138713837e-05,
-            "reward_mean": 0.8693352937698364,
-            "reward_var": 0.010553169995546341,
-            "avg_think_len": 19.8763370513916,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9148408770561218,
-            "step": 498
-        },
-        {
-            "loss": 7.058167830109596e-05,
-            "reward_mean": 0.8880413770675659,
-            "reward_var": 0.010419381782412529,
-            "avg_think_len": 19.80984878540039,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9896656274795532,
-            "step": 499
-        },
-        {
-            "loss": 4.7116074711084366e-05,
-            "reward_mean": 0.946499764919281,
-            "reward_var": 0.005965035408735275,
-            "avg_think_len": 19.830923080444336,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9891241192817688,
-            "step": 500
-        },
-        {
-            "loss": 6.372621282935143e-05,
-            "reward_mean": 0.9126339554786682,
-            "reward_var": 0.006560968700796366,
-            "avg_think_len": 19.96689796447754,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9817857146263123,
-            "step": 501
-        },
-        {
-            "loss": 6.734323687851429e-05,
-            "reward_mean": 0.8918426632881165,
-            "reward_var": 0.01784271001815796,
-            "avg_think_len": 19.89215850830078,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.9579956531524658,
-            "step": 502
-        },
-        {
-            "loss": 6.280872185016051e-05,
-            "reward_mean": 0.9644026160240173,
-            "reward_var": 0.005187463015317917,
-            "avg_think_len": 19.87496566772461,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.859375,
-            "reward_response_efficiency_mean": 0.9982356429100037,
-            "step": 503
-        },
-        {
-            "loss": 6.630620919167995e-05,
-            "reward_mean": 0.9049258232116699,
-            "reward_var": 0.007648036815226078,
-            "avg_think_len": 19.884336471557617,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9822032451629639,
-            "step": 504
-        },
-        {
-            "loss": 7.532211020588875e-05,
-            "reward_mean": 0.9194352030754089,
-            "reward_var": 0.014752279967069626,
-            "avg_think_len": 20.016422271728516,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9902409315109253,
-            "step": 505
-        },
-        {
-            "loss": 5.870312452316284e-05,
-            "reward_mean": 0.8437022566795349,
-            "reward_var": 0.022714732214808464,
-            "avg_think_len": 19.728389739990234,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.4375,
-            "reward_response_efficiency_mean": 0.9841840267181396,
-            "step": 506
-        },
-        {
-            "loss": 5.4555945098400116e-05,
-            "reward_mean": 0.8977859616279602,
-            "reward_var": 0.006398526951670647,
-            "avg_think_len": 19.895540237426758,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9973939061164856,
-            "step": 507
-        },
-        {
-            "loss": 5.512195639312267e-05,
-            "reward_mean": 0.8772433996200562,
-            "reward_var": 0.019385041669011116,
-            "avg_think_len": 20.029335021972656,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9620985984802246,
-            "step": 508
-        },
-        {
-            "loss": 8.064508438110352e-05,
-            "reward_mean": 0.8164517879486084,
-            "reward_var": 0.03877562656998634,
-            "avg_think_len": 19.849985122680664,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8812500238418579,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.8220571279525757,
-            "step": 509
-        },
-        {
-            "loss": 6.051763193681836e-05,
-            "reward_mean": 0.8337306976318359,
-            "reward_var": 0.021288547664880753,
-            "avg_think_len": 19.863618850708008,
-            "reward_format_mean": 0.9998566508293152,
-            "reward_time_mean": 0.9624999761581421,
-            "reward_correctness_mean": 0.484375,
-            "reward_response_efficiency_mean": 0.8881914615631104,
-            "step": 510
-        },
-        {
-            "loss": 5.5400654673576355e-05,
-            "reward_mean": 0.9085975289344788,
-            "reward_var": 0.014942717738449574,
-            "avg_think_len": 19.86190414428711,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9625151753425598,
-            "step": 511
-        },
-        {
-            "loss": 7.586297579109669e-05,
-            "reward_mean": 0.9280349016189575,
-            "reward_var": 0.011165600270032883,
-            "avg_think_len": 19.912193298339844,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9777650833129883,
-            "step": 512
-        },
-        {
-            "loss": 6.523169577121735e-05,
-            "reward_mean": 0.8725718259811401,
-            "reward_var": 0.0069367531687021255,
-            "avg_think_len": 20.211090087890625,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.9902874231338501,
-            "step": 513
-        },
-        {
-            "loss": 6.417324766516685e-05,
-            "reward_mean": 0.910973072052002,
-            "reward_var": 0.015792222693562508,
-            "avg_think_len": 19.975387573242188,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.949999988079071,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9907671809196472,
-            "step": 514
-        },
-        {
-            "loss": 4.94103878736496e-05,
-            "reward_mean": 0.8624205589294434,
-            "reward_var": 0.006873494479805231,
-            "avg_think_len": 19.891010284423828,
-            "reward_format_mean": 0.9998497366905212,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.453125,
-            "reward_response_efficiency_mean": 0.9967074394226074,
-            "step": 515
-        },
-        {
-            "loss": 7.505295798182487e-05,
-            "reward_mean": 0.9052271842956543,
-            "reward_var": 0.015519972890615463,
-            "avg_think_len": 19.917089462280273,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9552836418151855,
-            "step": 516
-        },
-        {
-            "loss": 5.730893462896347e-05,
-            "reward_mean": 0.855698823928833,
-            "reward_var": 0.020550444722175598,
-            "avg_think_len": 19.731952667236328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.484375,
-            "reward_response_efficiency_mean": 0.969670295715332,
-            "step": 517
-        },
-        {
-            "loss": 5.6519173085689545e-05,
-            "reward_mean": 0.8710190057754517,
-            "reward_var": 0.011123813688755035,
-            "avg_think_len": 19.833045959472656,
-            "reward_format_mean": 0.9998566508293152,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9685945510864258,
-            "step": 518
-        },
-        {
-            "loss": 5.85382804274559e-05,
-            "reward_mean": 0.9398465752601624,
-            "reward_var": 0.010034173727035522,
-            "avg_think_len": 19.95684051513672,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9687610268592834,
-            "step": 519
-        },
-        {
-            "loss": 5.884189158678055e-05,
-            "reward_mean": 0.8548965454101562,
-            "reward_var": 0.021894436329603195,
-            "avg_think_len": 19.777067184448242,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9508363008499146,
-            "step": 520
-        },
-        {
-            "loss": 3.6574434489011765e-05,
-            "reward_mean": 0.8912020921707153,
-            "reward_var": 0.012954158708453178,
-            "avg_think_len": 19.87497329711914,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9710586667060852,
-            "step": 521
-        },
-        {
-            "loss": 9.599560871720314e-05,
-            "reward_mean": 0.8764561414718628,
-            "reward_var": 0.012262404896318913,
-            "avg_think_len": 19.88456153869629,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9433244466781616,
-            "step": 522
-        },
-        {
-            "loss": 9.036227129399776e-05,
-            "reward_mean": 0.8562390804290771,
-            "reward_var": 0.011325652711093426,
-            "avg_think_len": 19.810178756713867,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.8937062621116638,
-            "step": 523
-        },
-        {
-            "loss": 8.697900921106339e-05,
-            "reward_mean": 0.889457643032074,
-            "reward_var": 0.011495808139443398,
-            "avg_think_len": 19.98743438720703,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9015806913375854,
-            "step": 524
-        },
-        {
-            "loss": 5.667912773787975e-05,
-            "reward_mean": 0.8478749990463257,
-            "reward_var": 0.012329384684562683,
-            "avg_think_len": 20.04956817626953,
-            "reward_format_mean": 0.9998686909675598,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.847881555557251,
-            "step": 525
-        },
-        {
-            "loss": 8.397223427891731e-05,
-            "reward_mean": 0.9022558927536011,
-            "reward_var": 0.009241756983101368,
-            "avg_think_len": 19.742765426635742,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9840233325958252,
-            "step": 526
-        },
-        {
-            "loss": 5.077826790511608e-05,
-            "reward_mean": 0.8804749846458435,
-            "reward_var": 0.012435052543878555,
-            "avg_think_len": 19.683197021484375,
-            "reward_format_mean": 0.9998675584793091,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9470325708389282,
-            "step": 527
-        },
-        {
-            "loss": 6.636464968323708e-05,
-            "reward_mean": 0.8518575429916382,
-            "reward_var": 0.028465241193771362,
-            "avg_think_len": 19.911487579345703,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9543050527572632,
-            "step": 528
-        },
-        {
-            "loss": 9.460048750042915e-05,
-            "reward_mean": 0.8861473798751831,
-            "reward_var": 0.006005186587572098,
-            "avg_think_len": 19.937929153442383,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9820897579193115,
-            "step": 529
-        },
-        {
-            "loss": 6.958981975913048e-05,
-            "reward_mean": 0.8870944380760193,
-            "reward_var": 0.011335273273289204,
-            "avg_think_len": 19.868793487548828,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9890028834342957,
-            "step": 530
-        },
-        {
-            "loss": -6.859190762042999e-06,
-            "reward_mean": 0.8778648376464844,
-            "reward_var": 0.008306116797029972,
-            "avg_think_len": 19.807527542114258,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.964584231376648,
-            "step": 531
-        },
-        {
-            "loss": 6.51446171104908e-05,
-            "reward_mean": 0.9428983330726624,
-            "reward_var": 0.004427643027156591,
-            "avg_think_len": 19.913330078125,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9903430938720703,
-            "step": 532
-        },
-        {
-            "loss": 6.142142228782177e-05,
-            "reward_mean": 0.9009384512901306,
-            "reward_var": 0.01704949513077736,
-            "avg_think_len": 19.90130043029785,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.931878924369812,
-            "step": 533
-        },
-        {
-            "loss": 7.069949060678482e-05,
-            "reward_mean": 0.9129154086112976,
-            "reward_var": 0.010418176651000977,
-            "avg_think_len": 20.073116302490234,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9797869324684143,
-            "step": 534
-        },
-        {
-            "loss": 8.54248646646738e-05,
-            "reward_mean": 0.8701379895210266,
-            "reward_var": 0.011903546750545502,
-            "avg_think_len": 19.951541900634766,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9649268984794617,
-            "step": 535
-        },
-        {
-            "loss": 6.277486681938171e-05,
-            "reward_mean": 0.8885308504104614,
-            "reward_var": 0.018854832276701927,
-            "avg_think_len": 19.805694580078125,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9603733420372009,
-            "step": 536
-        },
-        {
-            "loss": 7.715122774243355e-05,
-            "reward_mean": 0.9364112615585327,
-            "reward_var": 0.015241141431033611,
-            "avg_think_len": 19.947879791259766,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9643952250480652,
-            "step": 537
-        },
-        {
-            "loss": 5.381088703870773e-05,
-            "reward_mean": 0.9036719799041748,
-            "reward_var": 0.007524642627686262,
-            "avg_think_len": 19.951650619506836,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.989687979221344,
-            "step": 538
-        },
-        {
-            "loss": 4.314957186579704e-05,
-            "reward_mean": 0.9193597435951233,
-            "reward_var": 0.010874152183532715,
-            "avg_think_len": 19.49085807800293,
-            "reward_format_mean": 0.9991319179534912,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.8658072352409363,
-            "step": 539
-        },
-        {
-            "loss": 5.117850378155708e-05,
-            "reward_mean": 0.9157167673110962,
-            "reward_var": 0.019365794956684113,
-            "avg_think_len": 19.949554443359375,
-            "reward_format_mean": 0.9992897510528564,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9604524374008179,
-            "step": 540
-        },
-        {
-            "loss": 4.870491102337837e-05,
-            "reward_mean": 0.9253065586090088,
-            "reward_var": 0.009613582864403725,
-            "avg_think_len": 19.842918395996094,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9981014132499695,
-            "step": 541
-        },
-        {
-            "loss": 7.404704228974879e-05,
-            "reward_mean": 0.9547768235206604,
-            "reward_var": 0.005248498171567917,
-            "avg_think_len": 20.038440704345703,
-            "reward_format_mean": 0.9994419813156128,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.9759151339530945,
-            "step": 542
-        },
-        {
-            "loss": 5.006883293390274e-05,
-            "reward_mean": 0.9431237578392029,
-            "reward_var": 0.00959047768265009,
-            "avg_think_len": 19.972475051879883,
-            "reward_format_mean": 0.9993489384651184,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9606459140777588,
-            "step": 543
-        },
-        {
-            "loss": 4.563387483358383e-05,
-            "reward_mean": 0.8221720457077026,
-            "reward_var": 0.02218155935406685,
-            "avg_think_len": 19.889814376831055,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.949999988079071,
-            "reward_correctness_mean": 0.390625,
-            "reward_response_efficiency_mean": 0.94806307554245,
-            "step": 544
-        },
-        {
-            "loss": 3.884173929691315e-05,
-            "reward_mean": 0.9274579286575317,
-            "reward_var": 0.024548998102545738,
-            "avg_think_len": 19.97148895263672,
-            "reward_format_mean": 0.9998244047164917,
-            "reward_time_mean": 0.949999988079071,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9475073218345642,
-            "step": 545
-        },
-        {
-            "loss": 0.00011284090578556061,
-            "reward_mean": 0.9025294780731201,
-            "reward_var": 0.004660357255488634,
-            "avg_think_len": 20.112079620361328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.95386803150177,
-            "step": 546
-        },
-        {
-            "loss": 7.447763346135616e-05,
-            "reward_mean": 0.9117408990859985,
-            "reward_var": 0.005824057850986719,
-            "avg_think_len": 19.928129196166992,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.990713357925415,
-            "step": 547
-        },
-        {
-            "loss": 3.8412632420659065e-05,
-            "reward_mean": 0.968360960483551,
-            "reward_var": 0.001955632120370865,
-            "avg_think_len": 20.039936065673828,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.875,
-            "reward_response_efficiency_mean": 0.9984438419342041,
-            "step": 548
-        },
-        {
-            "loss": 5.49146207049489e-05,
-            "reward_mean": 0.9255159497261047,
-            "reward_var": 0.005708605051040649,
-            "avg_think_len": 20.157289505004883,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9989389181137085,
-            "step": 549
-        },
-        {
-            "loss": 6.597687024623156e-05,
-            "reward_mean": 0.9135103225708008,
-            "reward_var": 0.003820835379883647,
-            "avg_think_len": 19.88027000427246,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9352911710739136,
-            "step": 550
-        },
-        {
-            "loss": -4.2541068978607655e-07,
-            "reward_mean": 0.8936873078346252,
-            "reward_var": 0.02069958485662937,
-            "avg_think_len": 20.033124923706055,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9593750238418579,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.9591240882873535,
-            "step": 551
-        },
-        {
-            "loss": 6.072595715522766e-05,
-            "reward_mean": 0.835639476776123,
-            "reward_var": 0.033893074840307236,
-            "avg_think_len": 19.970287322998047,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.8894327878952026,
-            "step": 552
-        },
-        {
-            "loss": 5.930929910391569e-05,
-            "reward_mean": 0.8661022186279297,
-            "reward_var": 0.013674912042915821,
-            "avg_think_len": 19.965961456298828,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9331590533256531,
-            "step": 553
-        },
-        {
-            "loss": 5.027838051319122e-05,
-            "reward_mean": 0.933918297290802,
-            "reward_var": 0.021491728723049164,
-            "avg_think_len": 20.058015823364258,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.970048189163208,
-            "step": 554
-        },
-        {
-            "loss": 6.635233876295388e-05,
-            "reward_mean": 0.9062802791595459,
-            "reward_var": 0.013919839635491371,
-            "avg_think_len": 20.05113410949707,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9376212358474731,
-            "step": 555
-        },
-        {
-            "loss": 6.038229912519455e-05,
-            "reward_mean": 0.8782325387001038,
-            "reward_var": 0.013203169219195843,
-            "avg_think_len": 19.785133361816406,
-            "reward_format_mean": 0.99951171875,
-            "reward_time_mean": 0.9906250238418579,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9446682333946228,
-            "step": 556
-        },
-        {
-            "loss": 5.081715062260628e-05,
-            "reward_mean": 0.855124831199646,
-            "reward_var": 0.012861900962889194,
-            "avg_think_len": 19.886289596557617,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.9517494440078735,
-            "step": 557
-        },
-        {
-            "loss": 5.8461620938032866e-05,
-            "reward_mean": 0.9183643460273743,
-            "reward_var": 0.008123248815536499,
-            "avg_think_len": 20.06377410888672,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9703324437141418,
-            "step": 558
-        },
-        {
-            "loss": 5.577743286266923e-05,
-            "reward_mean": 0.9225172996520996,
-            "reward_var": 0.004023809917271137,
-            "avg_think_len": 19.910017013549805,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9900690913200378,
-            "step": 559
-        },
-        {
-            "loss": 9.502374450676143e-05,
-            "reward_mean": 0.9118530750274658,
-            "reward_var": 0.012483688071370125,
-            "avg_think_len": 19.950050354003906,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9442871809005737,
-            "step": 560
-        },
-        {
-            "loss": 6.107613444328308e-05,
-            "reward_mean": 0.8967015743255615,
-            "reward_var": 0.010750520043075085,
-            "avg_think_len": 19.813724517822266,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.987500011920929,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9899313449859619,
-            "step": 561
-        },
-        {
-            "loss": 8.976680692285299e-05,
-            "reward_mean": 0.8820075392723083,
-            "reward_var": 0.022611869499087334,
-            "avg_think_len": 19.936708450317383,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9342802166938782,
-            "step": 562
-        },
-        {
-            "loss": 0.00012128730304539204,
-            "reward_mean": 0.9503582119941711,
-            "reward_var": 0.009096693247556686,
-            "avg_think_len": 19.97325897216797,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9750000238418579,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.9826825857162476,
-            "step": 563
-        },
-        {
-            "loss": 5.76418824493885e-06,
-            "reward_mean": 0.931140661239624,
-            "reward_var": 0.0028940928168594837,
-            "avg_think_len": 20.079824447631836,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9901878833770752,
-            "step": 564
-        },
-        {
-            "loss": 6.315927021205425e-05,
-            "reward_mean": 0.9266148805618286,
-            "reward_var": 0.006857044063508511,
-            "avg_think_len": 19.933624267578125,
-            "reward_format_mean": 0.9997859597206116,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.972298800945282,
-            "step": 565
-        },
-        {
-            "loss": 6.288988515734673e-05,
-            "reward_mean": 0.9000642895698547,
-            "reward_var": 0.004280885681509972,
-            "avg_think_len": 20.127172470092773,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9908820986747742,
-            "step": 566
-        },
-        {
-            "loss": 5.708634853363037e-05,
-            "reward_mean": 0.9463211297988892,
-            "reward_var": 0.00864927377551794,
-            "avg_think_len": 20.04168701171875,
-            "reward_format_mean": 0.9993990659713745,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9733853340148926,
-            "step": 567
-        },
-        {
-            "loss": 5.809427239000797e-05,
-            "reward_mean": 0.8594256043434143,
-            "reward_var": 0.018419742584228516,
-            "avg_think_len": 19.85871696472168,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9377024173736572,
-            "step": 568
-        },
-        {
-            "loss": 7.696961984038353e-05,
-            "reward_mean": 0.8823477625846863,
-            "reward_var": 0.015536662191152573,
-            "avg_think_len": 19.93097686767578,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9825161099433899,
-            "step": 569
-        },
-        {
-            "loss": 7.667113095521927e-05,
-            "reward_mean": 0.9604964852333069,
-            "reward_var": 0.00474938889965415,
-            "avg_think_len": 20.106740951538086,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.859375,
-            "reward_response_efficiency_mean": 0.9826110601425171,
-            "step": 570
-        },
-        {
-            "loss": 3.59867699444294e-05,
-            "reward_mean": 0.8862808346748352,
-            "reward_var": 0.007130990270525217,
-            "avg_think_len": 20.052427291870117,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.9982481002807617,
-            "step": 571
-        },
-        {
-            "loss": 8.956249803304672e-05,
-            "reward_mean": 0.9544686079025269,
-            "reward_var": 0.0042886328883469105,
-            "avg_think_len": 20.070520401000977,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.9741244912147522,
-            "step": 572
-        },
-        {
-            "loss": 5.578482523560524e-05,
-            "reward_mean": 0.8784635663032532,
-            "reward_var": 0.009175444953143597,
-            "avg_think_len": 19.919795989990234,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.957604169845581,
-            "step": 573
-        },
-        {
-            "loss": 7.425155490636826e-05,
-            "reward_mean": 0.9459469318389893,
-            "reward_var": 0.004880972672253847,
-            "avg_think_len": 19.890457153320312,
-            "reward_format_mean": 0.9998437166213989,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9870690107345581,
-            "step": 574
-        },
-        {
-            "loss": 2.0013656467199326e-05,
-            "reward_mean": 0.9065117835998535,
-            "reward_var": 0.008837230503559113,
-            "avg_think_len": 20.099048614501953,
-            "reward_format_mean": 0.9984375238418579,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.9869848489761353,
-            "step": 575
-        },
-        {
-            "loss": 6.033666431903839e-05,
-            "reward_mean": 0.8569911122322083,
-            "reward_var": 0.018645670264959335,
-            "avg_think_len": 20.060991287231445,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.9748392701148987,
-            "step": 576
-        },
-        {
-            "loss": 6.079464219510555e-05,
-            "reward_mean": 0.9030594229698181,
-            "reward_var": 0.011602659709751606,
-            "avg_think_len": 19.847564697265625,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9716123342514038,
-            "step": 577
-        },
-        {
-            "loss": 8.683465421199799e-05,
-            "reward_mean": 0.8912401795387268,
-            "reward_var": 0.006821486633270979,
-            "avg_think_len": 20.129289627075195,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9868357181549072,
-            "step": 578
-        },
-        {
-            "loss": 4.2722560465335846e-05,
-            "reward_mean": 0.8479716181755066,
-            "reward_var": 0.05022681504487991,
-            "avg_think_len": 19.861339569091797,
-            "reward_format_mean": 0.9997373819351196,
-            "reward_time_mean": 0.875,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.8765240907669067,
-            "step": 579
-        },
-        {
-            "loss": 6.58971257507801e-05,
-            "reward_mean": 0.8992409706115723,
-            "reward_var": 0.031718119978904724,
-            "avg_think_len": 19.79915428161621,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8968750238418579,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9344637989997864,
-            "step": 580
-        },
-        {
-            "loss": 3.562972415238619e-05,
-            "reward_mean": 0.9402509331703186,
-            "reward_var": 0.009955242276191711,
-            "avg_think_len": 19.907426834106445,
-            "reward_format_mean": 0.9991776347160339,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9680760502815247,
-            "step": 581
-        },
-        {
-            "loss": 5.465769208967686e-05,
-            "reward_mean": 0.8763164281845093,
-            "reward_var": 0.02340395376086235,
-            "avg_think_len": 19.966663360595703,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.8958906531333923,
-            "step": 582
-        },
-        {
-            "loss": -6.300397217273712e-07,
-            "reward_mean": 0.9575693607330322,
-            "reward_var": 0.0067146820947527885,
-            "avg_think_len": 19.950626373291016,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.9865277409553528,
-            "step": 583
-        },
-        {
-            "loss": 6.110314279794693e-05,
-            "reward_mean": 0.9198673367500305,
-            "reward_var": 0.013669224455952644,
-            "avg_think_len": 20.057714462280273,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9763442873954773,
-            "step": 584
-        },
-        {
-            "loss": 4.105060361325741e-05,
-            "reward_mean": 0.856329083442688,
-            "reward_var": 0.02207511104643345,
-            "avg_think_len": 19.897663116455078,
-            "reward_format_mean": 0.999849796295166,
-            "reward_time_mean": 0.949999988079071,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.8817164897918701,
-            "step": 585
-        },
-        {
-            "loss": 6.282143294811249e-05,
-            "reward_mean": 0.8934276700019836,
-            "reward_var": 0.013291753828525543,
-            "avg_think_len": 19.86494255065918,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9393357038497925,
-            "step": 586
-        },
-        {
-            "loss": 7.565412670373917e-05,
-            "reward_mean": 0.9269188046455383,
-            "reward_var": 0.020627474412322044,
-            "avg_think_len": 20.030790328979492,
-            "reward_format_mean": 0.9984375238418579,
-            "reward_time_mean": 0.965624988079071,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9779878258705139,
-            "step": 587
-        },
-        {
-            "loss": 6.232038140296936e-05,
-            "reward_mean": 0.8419971466064453,
-            "reward_var": 0.021833181381225586,
-            "avg_think_len": 20.015344619750977,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.9304887056350708,
-            "step": 588
-        },
-        {
-            "loss": 5.834503099322319e-05,
-            "reward_mean": 0.9243828654289246,
-            "reward_var": 0.007724681869149208,
-            "avg_think_len": 19.872318267822266,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.997531533241272,
-            "step": 589
-        },
-        {
-            "loss": 5.6910605053417385e-05,
-            "reward_mean": 0.8998351693153381,
-            "reward_var": 0.014487640000879765,
-            "avg_think_len": 19.904224395751953,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.9587156772613525,
-            "step": 590
-        },
-        {
-            "loss": 7.442920468747616e-05,
-            "reward_mean": 0.954349160194397,
-            "reward_var": 0.0038796900771558285,
-            "avg_think_len": 19.93866729736328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.9892712831497192,
-            "step": 591
-        },
-        {
-            "loss": 9.233597666025162e-05,
-            "reward_mean": 0.8570672273635864,
-            "reward_var": 0.022733498364686966,
-            "avg_think_len": 20.043560028076172,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.97514408826828,
-            "step": 592
-        },
-        {
-            "loss": 6.020069122314453e-05,
-            "reward_mean": 0.9197532534599304,
-            "reward_var": 0.012380777858197689,
-            "avg_think_len": 20.115745544433594,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9781249761581421,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9977629780769348,
-            "step": 593
-        },
-        {
-            "loss": 5.5711854656692594e-05,
-            "reward_mean": 0.8382391929626465,
-            "reward_var": 0.014955183491110802,
-            "avg_think_len": 19.88722038269043,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.8373318314552307,
-            "step": 594
-        },
-        {
-            "loss": 6.496114656329155e-05,
-            "reward_mean": 0.9388960003852844,
-            "reward_var": 0.00928298570215702,
-            "avg_think_len": 19.985567092895508,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9743338823318481,
-            "step": 595
-        },
-        {
-            "loss": 0.00011083763092756271,
-            "reward_mean": 0.8955939412117004,
-            "reward_var": 0.01036019530147314,
-            "avg_think_len": 20.032926559448242,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9573756456375122,
-            "step": 596
-        },
-        {
-            "loss": 5.995435640215874e-05,
-            "reward_mean": 0.9447363615036011,
-            "reward_var": 0.013200887478888035,
-            "avg_think_len": 19.937976837158203,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9820705056190491,
-            "step": 597
-        },
-        {
-            "loss": 7.297424599528313e-05,
-            "reward_mean": 0.9213653802871704,
-            "reward_var": 0.007837743498384953,
-            "avg_think_len": 19.84002113342285,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9979612231254578,
-            "step": 598
-        },
-        {
-            "loss": 3.889109939336777e-05,
-            "reward_mean": 0.948670506477356,
-            "reward_var": 0.002085309475660324,
-            "avg_think_len": 19.915630340576172,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9978068470954895,
-            "step": 599
-        },
-        {
-            "loss": 5.16730360686779e-05,
-            "reward_mean": 0.8867471218109131,
-            "reward_var": 0.016151050105690956,
-            "avg_think_len": 19.843576431274414,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9532390236854553,
-            "step": 600
-        },
-        {
-            "loss": 2.965587191283703e-05,
-            "reward_mean": 0.9036998748779297,
-            "reward_var": 0.006497017107903957,
-            "avg_think_len": 19.8781795501709,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9741746187210083,
-            "step": 601
-        },
-        {
-            "loss": 5.85170928388834e-05,
-            "reward_mean": 0.8999002575874329,
-            "reward_var": 0.02133123017847538,
-            "avg_think_len": 19.978408813476562,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.8964760899543762,
-            "step": 602
-        },
-        {
-            "loss": -0.00010644143912941217,
-            "reward_mean": 0.9430214166641235,
-            "reward_var": 0.0074902744963765144,
-            "avg_think_len": 19.885818481445312,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9908356070518494,
-            "step": 603
-        },
-        {
-            "loss": 5.8498524595052004e-05,
-            "reward_mean": 0.9396746754646301,
-            "reward_var": 0.016321878880262375,
-            "avg_think_len": 20.011960983276367,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.9461987018585205,
-            "step": 604
-        },
-        {
-            "loss": 8.654780685901642e-05,
-            "reward_mean": 0.857108473777771,
-            "reward_var": 0.012747514992952347,
-            "avg_think_len": 20.049266815185547,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9315587878227234,
-            "step": 605
-        },
-        {
-            "loss": 4.5546796172857285e-05,
-            "reward_mean": 0.9067516326904297,
-            "reward_var": 0.004874875769019127,
-            "avg_think_len": 19.878053665161133,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9863816499710083,
-            "step": 606
-        },
-        {
-            "loss": 0.00016203915583901107,
-            "reward_mean": 0.9208805561065674,
-            "reward_var": 0.0169694721698761,
-            "avg_think_len": 19.89718246459961,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9335221648216248,
-            "step": 607
-        },
-        {
-            "loss": 8.53012315928936e-05,
-            "reward_mean": 0.8932071328163147,
-            "reward_var": 0.006758034694939852,
-            "avg_think_len": 19.876914978027344,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9790785908699036,
-            "step": 608
-        },
-        {
-            "loss": 5.8690086007118225e-05,
-            "reward_mean": 0.858004093170166,
-            "reward_var": 0.026131073012948036,
-            "avg_think_len": 19.79706573486328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9624999761581421,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9382662773132324,
-            "step": 609
-        },
-        {
-            "loss": 5.10229729115963e-05,
-            "reward_mean": 0.8795334100723267,
-            "reward_var": 0.007321957498788834,
-            "avg_think_len": 19.952970504760742,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9868836402893066,
-            "step": 610
-        },
-        {
-            "loss": 4.940340295433998e-05,
-            "reward_mean": 0.943004846572876,
-            "reward_var": 0.005195297300815582,
-            "avg_think_len": 19.98050308227539,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9751442074775696,
-            "step": 611
-        },
-        {
-            "loss": 9.277090430259705e-05,
-            "reward_mean": 0.8738960027694702,
-            "reward_var": 0.014628310687839985,
-            "avg_think_len": 20.100296020507812,
-            "reward_format_mean": 0.9998405575752258,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9332432746887207,
-            "step": 612
-        },
-        {
-            "loss": 5.750218406319618e-05,
-            "reward_mean": 0.8939490914344788,
-            "reward_var": 0.011868403293192387,
-            "avg_think_len": 19.91632652282715,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.997671365737915,
-            "step": 613
-        },
-        {
-            "loss": 7.782306056469679e-05,
-            "reward_mean": 0.9017570614814758,
-            "reward_var": 0.010737732984125614,
-            "avg_think_len": 20.015384674072266,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9976532459259033,
-            "step": 614
-        },
-        {
-            "loss": 7.060705684125423e-05,
-            "reward_mean": 0.8770586848258972,
-            "reward_var": 0.016415979713201523,
-            "avg_think_len": 19.959701538085938,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9613593220710754,
-            "step": 615
-        },
-        {
-            "loss": 5.659274756908417e-05,
-            "reward_mean": 0.8589619994163513,
-            "reward_var": 0.015909994021058083,
-            "avg_think_len": 20.09747886657715,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9514727592468262,
-            "step": 616
-        },
-        {
-            "loss": 4.9977563321590424e-05,
-            "reward_mean": 0.9096951484680176,
-            "reward_var": 0.025591503828763962,
-            "avg_think_len": 19.99032211303711,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9512805938720703,
-            "step": 617
-        },
-        {
-            "loss": 8.413102477788925e-05,
-            "reward_mean": 0.9439646601676941,
-            "reward_var": 0.004832098260521889,
-            "avg_think_len": 20.033170700073242,
-            "reward_format_mean": 0.9994212985038757,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.979562520980835,
-            "step": 618
-        },
-        {
-            "loss": -1.9519298803061247e-05,
-            "reward_mean": 0.9214394092559814,
-            "reward_var": 0.009215054102241993,
-            "avg_think_len": 19.950777053833008,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9982577562332153,
-            "step": 619
-        },
-        {
-            "loss": 4.43565659224987e-06,
-            "reward_mean": 0.9483681917190552,
-            "reward_var": 0.01632063090801239,
-            "avg_think_len": 19.8349666595459,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.890625,
-            "reward_response_efficiency_mean": 0.9340976476669312,
-            "step": 620
-        },
-        {
-            "loss": 6.327778100967407e-05,
-            "reward_mean": 0.9308417439460754,
-            "reward_var": 0.005276086740195751,
-            "avg_think_len": 20.1192626953125,
-            "reward_format_mean": 0.9984375238418579,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9749297499656677,
-            "step": 621
-        },
-        {
-            "loss": 4.8446934670209885e-05,
-            "reward_mean": 0.9585886597633362,
-            "reward_var": 0.005962246097624302,
-            "avg_think_len": 20.036396026611328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.9906046390533447,
-            "step": 622
-        },
-        {
-            "loss": 6.516184657812119e-05,
-            "reward_mean": 0.9310053586959839,
-            "reward_var": 0.00951379630714655,
-            "avg_think_len": 19.911344528198242,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9802714586257935,
-            "step": 623
-        },
-        {
-            "loss": 9.161746129393578e-05,
-            "reward_mean": 0.9214295148849487,
-            "reward_var": 0.008078924380242825,
-            "avg_think_len": 20.063962936401367,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9982179403305054,
-            "step": 624
-        },
-        {
-            "loss": 5.770474672317505e-06,
-            "reward_mean": 0.9070110321044922,
-            "reward_var": 0.008697832934558392,
-            "avg_think_len": 20.04216766357422,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9905440211296082,
-            "step": 625
-        },
-        {
-            "loss": 6.469033542089164e-05,
-            "reward_mean": 0.9462868571281433,
-            "reward_var": 0.011818256229162216,
-            "avg_think_len": 20.023895263671875,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9882718324661255,
-            "step": 626
-        },
-        {
-            "loss": 9.859632700681686e-05,
-            "reward_mean": 0.9373430609703064,
-            "reward_var": 0.008231950923800468,
-            "avg_think_len": 19.96550750732422,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9524971842765808,
-            "step": 627
-        },
-        {
-            "loss": 0.0001092695165425539,
-            "reward_mean": 0.9459253549575806,
-            "reward_var": 0.01701231859624386,
-            "avg_think_len": 19.743759155273438,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.859375,
-            "reward_response_efficiency_mean": 0.9555766582489014,
-            "step": 628
-        },
-        {
-            "loss": 6.772694177925587e-05,
-            "reward_mean": 0.8503286242485046,
-            "reward_var": 0.019991787150502205,
-            "avg_think_len": 19.806968688964844,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.9638144969940186,
-            "step": 629
-        },
-        {
-            "loss": 5.989836063235998e-05,
-            "reward_mean": 0.9172838926315308,
-            "reward_var": 0.006285243202000856,
-            "avg_think_len": 19.865535736083984,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9972609281539917,
-            "step": 630
-        },
-        {
-            "loss": 1.7112120985984802e-05,
-            "reward_mean": 0.9524967670440674,
-            "reward_var": 0.0022227426525205374,
-            "avg_think_len": 20.110092163085938,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9974871873855591,
-            "step": 631
-        },
-        {
-            "loss": 6.12335279583931e-05,
-            "reward_mean": 0.9452186226844788,
-            "reward_var": 0.009818491525948048,
-            "avg_think_len": 20.089954376220703,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.9683743119239807,
-            "step": 632
-        },
-        {
-            "loss": 8.926354348659515e-05,
-            "reward_mean": 0.930594801902771,
-            "reward_var": 0.005045011639595032,
-            "avg_think_len": 20.125104904174805,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9880041480064392,
-            "step": 633
-        },
-        {
-            "loss": 4.7127483412623405e-05,
-            "reward_mean": 0.9382466673851013,
-            "reward_var": 0.006706704385578632,
-            "avg_think_len": 19.92145347595215,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.97173672914505,
-            "step": 634
-        },
-        {
-            "loss": 0.00012241071090102196,
-            "reward_mean": 0.9254439473152161,
-            "reward_var": 0.007392674684524536,
-            "avg_think_len": 19.990772247314453,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9986507892608643,
-            "step": 635
-        },
-        {
-            "loss": 5.420076195150614e-05,
-            "reward_mean": 0.8924010992050171,
-            "reward_var": 0.01860567182302475,
-            "avg_think_len": 20.201904296875,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9749999642372131,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9227295517921448,
-            "step": 636
-        },
-        {
-            "loss": 4.014023579657078e-05,
-            "reward_mean": 0.9293210506439209,
-            "reward_var": 0.02564121223986149,
-            "avg_think_len": 19.909948348999023,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.875,
-            "reward_response_efficiency_mean": 0.9047841429710388,
-            "step": 637
-        },
-        {
-            "loss": 9.542517364025116e-05,
-            "reward_mean": 0.8980475664138794,
-            "reward_var": 0.006697743199765682,
-            "avg_think_len": 20.182519912719727,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9984400868415833,
-            "step": 638
-        },
-        {
-            "loss": 2.044485881924629e-06,
-            "reward_mean": 0.8751107454299927,
-            "reward_var": 0.015629298985004425,
-            "avg_think_len": 19.917280197143555,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9566928148269653,
-            "step": 639
-        },
-        {
-            "loss": 2.5954563170671463e-05,
-            "reward_mean": 0.8964157700538635,
-            "reward_var": 0.013273454271256924,
-            "avg_think_len": 19.95168685913086,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9781249761581421,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9825381636619568,
-            "step": 640
-        },
-        {
-            "loss": 5.3394585847854614e-05,
-            "reward_mean": 0.9171786904335022,
-            "reward_var": 0.02133997529745102,
-            "avg_think_len": 19.8536376953125,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9343397617340088,
-            "step": 641
-        },
-        {
-            "loss": 5.81950880587101e-05,
-            "reward_mean": 0.905698299407959,
-            "reward_var": 0.0052914684638381,
-            "avg_think_len": 19.895872116088867,
-            "reward_format_mean": 0.9994611740112305,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.967082142829895,
-            "step": 642
-        },
-        {
-            "loss": 3.9188715163618326e-05,
-            "reward_mean": 0.8941737413406372,
-            "reward_var": 0.009807541966438293,
-            "avg_think_len": 20.039962768554688,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9673203229904175,
-            "step": 643
-        },
-        {
-            "loss": 5.489867180585861e-05,
-            "reward_mean": 0.878598153591156,
-            "reward_var": 0.004929061979055405,
-            "avg_think_len": 19.90131950378418,
-            "reward_format_mean": 0.9998117685317993,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9989557862281799,
-            "step": 644
-        },
-        {
-            "loss": 6.983615458011627e-05,
-            "reward_mean": 0.8639957904815674,
-            "reward_var": 0.01617198809981346,
-            "avg_think_len": 19.905017852783203,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9716084599494934,
-            "step": 645
-        },
-        {
-            "loss": 8.823396638035774e-05,
-            "reward_mean": 0.9130431413650513,
-            "reward_var": 0.009738655760884285,
-            "avg_think_len": 19.934589385986328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9802976846694946,
-            "step": 646
-        },
-        {
-            "loss": 6.39110803604126e-05,
-            "reward_mean": 0.8168105483055115,
-            "reward_var": 0.01817232370376587,
-            "avg_think_len": 19.994430541992188,
-            "reward_format_mean": 0.9998354911804199,
-            "reward_time_mean": 0.949999988079071,
-            "reward_correctness_mean": 0.453125,
-            "reward_response_efficiency_mean": 0.8642816543579102,
-            "step": 647
-        },
-        {
-            "loss": 9.585125371813774e-05,
-            "reward_mean": 0.9486526846885681,
-            "reward_var": 0.00516506889835,
-            "avg_think_len": 19.858558654785156,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9977357983589172,
-            "step": 648
-        },
-        {
-            "loss": 7.607508450746536e-05,
-            "reward_mean": 0.8960850834846497,
-            "reward_var": 0.0044561452232301235,
-            "avg_think_len": 20.069692611694336,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9905902743339539,
-            "step": 649
-        },
-        {
-            "loss": 7.467251271009445e-05,
-            "reward_mean": 0.8881610035896301,
-            "reward_var": 0.01611368916928768,
-            "avg_think_len": 20.047441482543945,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9901441335678101,
-            "step": 650
-        },
-        {
-            "loss": 6.555160507559776e-05,
-            "reward_mean": 0.8733296990394592,
-            "reward_var": 0.009324650280177593,
-            "avg_think_len": 20.06318473815918,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9776938557624817,
-            "step": 651
-        },
-        {
-            "loss": 7.764296606183052e-05,
-            "reward_mean": 0.9109283685684204,
-            "reward_var": 0.008592374622821808,
-            "avg_think_len": 20.004257202148438,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.8780883550643921,
-            "step": 652
-        },
-        {
-            "loss": -1.5348196029663086e-06,
-            "reward_mean": 0.9506195783615112,
-            "reward_var": 0.005834557581692934,
-            "avg_think_len": 19.916086196899414,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9899784922599792,
-            "step": 653
-        },
-        {
-            "loss": 7.777102291584015e-05,
-            "reward_mean": 0.9332281351089478,
-            "reward_var": 0.007689318619668484,
-            "avg_think_len": 20.107545852661133,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9985374808311462,
-            "step": 654
-        },
-        {
-            "loss": 5.496549420058727e-05,
-            "reward_mean": 0.9170026779174805,
-            "reward_var": 0.011197138577699661,
-            "avg_think_len": 20.089679718017578,
-            "reward_format_mean": 0.998319149017334,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9665663838386536,
-            "step": 655
-        },
-        {
-            "loss": 6.706966087222099e-05,
-            "reward_mean": 0.917367696762085,
-            "reward_var": 0.002655215561389923,
-            "avg_think_len": 19.93579864501953,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9975960850715637,
-            "step": 656
-        },
-        {
-            "loss": 4.702107980847359e-05,
-            "reward_mean": 0.912156343460083,
-            "reward_var": 0.005243335384875536,
-            "avg_think_len": 20.02996253967285,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.961125373840332,
-            "step": 657
-        },
-        {
-            "loss": 0.00010914012091234326,
-            "reward_mean": 0.8898864984512329,
-            "reward_var": 0.017482275143265724,
-            "avg_think_len": 19.908483505249023,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.949999988079071,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.9532960653305054,
-            "step": 658
-        },
-        {
-            "loss": 8.76900739967823e-05,
-            "reward_mean": 0.9073746204376221,
-            "reward_var": 0.020233498886227608,
-            "avg_think_len": 20.109649658203125,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9419984221458435,
-            "step": 659
-        },
-        {
-            "loss": 8.697505109012127e-05,
-            "reward_mean": 0.9031223654747009,
-            "reward_var": 0.016375020146369934,
-            "avg_think_len": 20.014976501464844,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9718645811080933,
-            "step": 660
-        },
-        {
-            "loss": 8.279411122202873e-05,
-            "reward_mean": 0.9447530508041382,
-            "reward_var": 0.0050356267020106316,
-            "avg_think_len": 20.081472396850586,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9977623224258423,
-            "step": 661
-        },
-        {
-            "loss": 6.603915244340897e-05,
-            "reward_mean": 0.9058306217193604,
-            "reward_var": 0.008362394757568836,
-            "avg_think_len": 19.913564682006836,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9826973676681519,
-            "step": 662
-        },
-        {
-            "loss": 7.17034563422203e-05,
-            "reward_mean": 0.892959475517273,
-            "reward_var": 0.006163205951452255,
-            "avg_think_len": 20.00595474243164,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9468379616737366,
-            "step": 663
-        },
-        {
-            "loss": 7.46389850974083e-05,
-            "reward_mean": 0.9040868878364563,
-            "reward_var": 0.008793439716100693,
-            "avg_think_len": 19.99941062927246,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.9600974917411804,
-            "step": 664
-        },
-        {
-            "loss": 6.367024616338313e-05,
-            "reward_mean": 0.9280040860176086,
-            "reward_var": 0.011354417540133,
-            "avg_think_len": 19.966550827026367,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9620164632797241,
-            "step": 665
-        },
-        {
-            "loss": 7.645034929737449e-05,
-            "reward_mean": 0.8874024152755737,
-            "reward_var": 0.016492823138833046,
-            "avg_think_len": 20.034774780273438,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.965624988079071,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.9277346730232239,
-            "step": 666
-        },
-        {
-            "loss": 6.030220538377762e-05,
-            "reward_mean": 0.9143765568733215,
-            "reward_var": 0.021211151033639908,
-            "avg_think_len": 19.88866424560547,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9543813467025757,
-            "step": 667
-        },
-        {
-            "loss": 6.429385393857956e-05,
-            "reward_mean": 0.931145966053009,
-            "reward_var": 0.0028969780541956425,
-            "avg_think_len": 20.064777374267578,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9902089238166809,
-            "step": 668
-        },
-        {
-            "loss": 6.761215627193451e-05,
-            "reward_mean": 0.9184740781784058,
-            "reward_var": 0.011765541508793831,
-            "avg_think_len": 20.219213485717773,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9863964319229126,
-            "step": 669
-        },
-        {
-            "loss": 4.707032348960638e-05,
-            "reward_mean": 0.922006368637085,
-            "reward_var": 0.011403077282011509,
-            "avg_think_len": 19.91292381286621,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9380254149436951,
-            "step": 670
-        },
-        {
-            "loss": 0.00012556981528177857,
-            "reward_mean": 0.8645516037940979,
-            "reward_var": 0.006898940075188875,
-            "avg_think_len": 19.98041534423828,
-            "reward_format_mean": 0.9998452663421631,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.8646107912063599,
-            "step": 671
-        },
-        {
-            "loss": 0.00015989295206964016,
-            "reward_mean": 0.896246075630188,
-            "reward_var": 0.023873623460531235,
-            "avg_think_len": 20.066097259521484,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9599841833114624,
-            "step": 672
-        },
-        {
-            "loss": 4.548662400338799e-05,
-            "reward_mean": 0.8595499396324158,
-            "reward_var": 0.016215335577726364,
-            "avg_think_len": 20.070480346679688,
-            "reward_format_mean": 0.9998675584793091,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.8289573788642883,
-            "step": 673
-        },
-        {
-            "loss": 5.007954314351082e-05,
-            "reward_mean": 0.919705867767334,
-            "reward_var": 0.02264164574444294,
-            "avg_think_len": 20.09725570678711,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9288235306739807,
-            "step": 674
-        },
-        {
-            "loss": 4.3143751099705696e-05,
-            "reward_mean": 0.9145858883857727,
-            "reward_var": 0.02293235808610916,
-            "avg_think_len": 19.91767120361328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9239687323570251,
-            "step": 675
-        },
-        {
-            "loss": 2.5523360818624496e-05,
-            "reward_mean": 0.9055313467979431,
-            "reward_var": 0.007290907669812441,
-            "avg_think_len": 20.036846160888672,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9971256256103516,
-            "step": 676
-        },
-        {
-            "loss": 6.36526383459568e-05,
-            "reward_mean": 0.8857026696205139,
-            "reward_var": 0.016927003860473633,
-            "avg_think_len": 19.984622955322266,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9490604996681213,
-            "step": 677
-        },
-        {
-            "loss": 0.00025302532594650984,
-            "reward_mean": 0.9797163009643555,
-            "reward_var": 0.006964486092329025,
-            "avg_think_len": 20.078805923461914,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9906250238418579,
-            "reward_correctness_mean": 0.9375,
-            "reward_response_efficiency_mean": 0.9907397627830505,
-            "step": 678
-        },
-        {
-            "loss": 0.00010242825374007225,
-            "reward_mean": 0.9228785037994385,
-            "reward_var": 0.012973057106137276,
-            "avg_think_len": 20.113506317138672,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9781249761581421,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9633886814117432,
-            "step": 679
-        },
-        {
-            "loss": 7.18957744538784e-05,
-            "reward_mean": 0.8914133906364441,
-            "reward_var": 0.014044913463294506,
-            "avg_think_len": 20.09954071044922,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9875285625457764,
-            "step": 680
-        },
-        {
-            "loss": 8.035625796765089e-05,
-            "reward_mean": 0.889701247215271,
-            "reward_var": 0.013559624552726746,
-            "avg_think_len": 19.938159942626953,
-            "reward_format_mean": 0.9984375238418579,
-            "reward_time_mean": 0.987500011920929,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9634926319122314,
-            "step": 681
-        },
-        {
-            "loss": 4.355784767540172e-05,
-            "reward_mean": 0.9403440952301025,
-            "reward_var": 0.005994941107928753,
-            "avg_think_len": 20.189510345458984,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9957509636878967,
-            "step": 682
-        },
-        {
-            "loss": 0.0001103677786886692,
-            "reward_mean": 0.9447402954101562,
-            "reward_var": 0.007796697784215212,
-            "avg_think_len": 20.030292510986328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9820862412452698,
-            "step": 683
-        },
-        {
-            "loss": 5.9278449043631554e-05,
-            "reward_mean": 0.8653202056884766,
-            "reward_var": 0.00661889323964715,
-            "avg_think_len": 20.204166412353516,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.9675310850143433,
-            "step": 684
-        },
-        {
-            "loss": 0.00013440113980323076,
-            "reward_mean": 0.9004026651382446,
-            "reward_var": 0.010876568965613842,
-            "avg_think_len": 20.071866989135742,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9766106605529785,
-            "step": 685
-        },
-        {
-            "loss": 8.35559330880642e-05,
-            "reward_mean": 0.9029154181480408,
-            "reward_var": 0.006435340270400047,
-            "avg_think_len": 20.141828536987305,
-            "reward_format_mean": 0.9984375238418579,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9882240891456604,
-            "step": 686
-        },
-        {
-            "loss": 7.875356823205948e-05,
-            "reward_mean": 0.8645539879798889,
-            "reward_var": 0.015018622390925884,
-            "avg_think_len": 20.038610458374023,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.9738412499427795,
-            "step": 687
-        },
-        {
-            "loss": 7.300404831767082e-05,
-            "reward_mean": 0.889298677444458,
-            "reward_var": 0.0068125189282000065,
-            "avg_think_len": 20.193729400634766,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9790698289871216,
-            "step": 688
-        },
-        {
-            "loss": -1.2302771210670471e-05,
-            "reward_mean": 0.8824493885040283,
-            "reward_var": 0.007237358018755913,
-            "avg_think_len": 20.056808471679688,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9985476732254028,
-            "step": 689
-        },
-        {
-            "loss": 7.84359872341156e-05,
-            "reward_mean": 0.9205293655395508,
-            "reward_var": 0.01486114040017128,
-            "avg_think_len": 19.81539535522461,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9071175456047058,
-            "step": 690
-        },
-        {
-            "loss": 7.039029151201248e-05,
-            "reward_mean": 0.9294606447219849,
-            "reward_var": 0.010306047275662422,
-            "avg_think_len": 20.056779861450195,
-            "reward_format_mean": 0.9988839626312256,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9689589142799377,
-            "step": 691
-        },
-        {
-            "loss": 0.0001361240865662694,
-            "reward_mean": 0.8929969072341919,
-            "reward_var": 0.01600329577922821,
-            "avg_think_len": 19.9874267578125,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9501124620437622,
-            "step": 692
-        },
-        {
-            "loss": 5.0603412091732025e-06,
-            "reward_mean": 0.905467689037323,
-            "reward_var": 0.012597601860761642,
-            "avg_think_len": 20.061721801757812,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.965624988079071,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9531208276748657,
-            "step": 693
-        },
-        {
-            "loss": 1.469138078391552e-05,
-            "reward_mean": 0.925541877746582,
-            "reward_var": 0.009067917242646217,
-            "avg_think_len": 20.04952621459961,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9834174513816833,
-            "step": 694
-        },
-        {
-            "loss": 7.294910028576851e-05,
-            "reward_mean": 0.8472431898117065,
-            "reward_var": 0.007378058973699808,
-            "avg_think_len": 20.03715705871582,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.390625,
-            "reward_response_efficiency_mean": 0.9983475804328918,
-            "step": 695
-        },
-        {
-            "loss": 6.376532837748528e-05,
-            "reward_mean": 0.9175801277160645,
-            "reward_var": 0.012229645624756813,
-            "avg_think_len": 19.972042083740234,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9515705108642578,
-            "step": 696
-        },
-        {
-            "loss": 4.472234286367893e-05,
-            "reward_mean": 0.917726457118988,
-            "reward_var": 0.006229874212294817,
-            "avg_think_len": 20.077919006347656,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.8927807211875916,
-            "step": 697
-        },
-        {
-            "loss": 0.0002110847271978855,
-            "reward_mean": 0.9501710534095764,
-            "reward_var": 0.0074009522795677185,
-            "avg_think_len": 20.004125595092773,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.9725594520568848,
-            "step": 698
-        },
-        {
-            "loss": 4.330161027610302e-05,
-            "reward_mean": 0.9065316915512085,
-            "reward_var": 0.006402480415999889,
-            "avg_think_len": 20.209653854370117,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9855020046234131,
-            "step": 699
-        },
-        {
-            "loss": 7.065199315547943e-05,
-            "reward_mean": 0.8933171629905701,
-            "reward_var": 0.01525941677391529,
-            "avg_think_len": 20.004209518432617,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9482688307762146,
-            "step": 700
-        },
-        {
-            "loss": 4.6848319470882416e-05,
-            "reward_mean": 0.9241767525672913,
-            "reward_var": 0.01112560648471117,
-            "avg_think_len": 20.0164737701416,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.915457010269165,
-            "step": 701
-        },
-        {
-            "loss": 0.00010322360321879387,
-            "reward_mean": 0.8779693841934204,
-            "reward_var": 0.01658713072538376,
-            "avg_think_len": 20.11404800415039,
-            "reward_format_mean": 0.9998355507850647,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9495416879653931,
-            "step": 702
-        },
-        {
-            "loss": 6.070174276828766e-05,
-            "reward_mean": 0.9489613771438599,
-            "reward_var": 0.0039707375690341,
-            "avg_think_len": 19.838178634643555,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9864705204963684,
-            "step": 703
-        },
-        {
-            "loss": 6.762240082025528e-05,
-            "reward_mean": 0.898453414440155,
-            "reward_var": 0.012815441004931927,
-            "avg_think_len": 20.080955505371094,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.953188419342041,
-            "step": 704
-        },
-        {
-            "loss": 9.146216325461864e-05,
-            "reward_mean": 0.9683434367179871,
-            "reward_var": 0.0058526042848825455,
-            "avg_think_len": 19.89478874206543,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.875,
-            "reward_response_efficiency_mean": 0.9983738660812378,
-            "step": 705
-        },
-        {
-            "loss": 6.480154115706682e-05,
-            "reward_mean": 0.8954971432685852,
-            "reward_var": 0.009537745267152786,
-            "avg_think_len": 20.055295944213867,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9882387518882751,
-            "step": 706
-        },
-        {
-            "loss": 7.231661584228277e-05,
-            "reward_mean": 0.8912531137466431,
-            "reward_var": 0.01824202388525009,
-            "avg_think_len": 19.9530086517334,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9712622761726379,
-            "step": 707
-        },
-        {
-            "loss": 7.569370791316032e-05,
-            "reward_mean": 0.9835161566734314,
-            "reward_var": 0.003067722776904702,
-            "avg_think_len": 19.938262939453125,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.9375,
-            "reward_response_efficiency_mean": 0.9965648055076599,
-            "step": 708
-        },
-        {
-            "loss": 6.211816798895597e-05,
-            "reward_mean": 0.8731185793876648,
-            "reward_var": 0.0273609422147274,
-            "avg_think_len": 19.924585342407227,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9143494367599487,
-            "step": 709
-        },
-        {
-            "loss": 4.3031584937125444e-05,
-            "reward_mean": 0.8583948016166687,
-            "reward_var": 0.023505453020334244,
-            "avg_think_len": 19.994556427001953,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9648292064666748,
-            "step": 710
-        },
-        {
-            "loss": 5.005464481655508e-05,
-            "reward_mean": 0.943025529384613,
-            "reward_var": 0.006241052411496639,
-            "avg_think_len": 20.062410354614258,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9908522367477417,
-            "step": 711
-        },
-        {
-            "loss": 9.311805479228497e-05,
-            "reward_mean": 0.8786168098449707,
-            "reward_var": 0.008226907812058926,
-            "avg_think_len": 20.071590423583984,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9988422393798828,
-            "step": 712
-        },
-        {
-            "loss": 3.547128289937973e-05,
-            "reward_mean": 0.8899966478347778,
-            "reward_var": 0.01200688723474741,
-            "avg_think_len": 20.108413696289062,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9974867105484009,
-            "step": 713
-        },
-        {
-            "loss": 4.86886128783226e-05,
-            "reward_mean": 0.8911250829696655,
-            "reward_var": 0.020987868309020996,
-            "avg_think_len": 19.988468170166016,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9551254510879517,
-            "step": 714
-        },
-        {
-            "loss": -1.2803880963474512e-05,
-            "reward_mean": 0.8833051323890686,
-            "reward_var": 0.01878795027732849,
-            "avg_think_len": 19.9448184967041,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9238455295562744,
-            "step": 715
-        },
-        {
-            "loss": 5.175964906811714e-05,
-            "reward_mean": 0.8908472657203674,
-            "reward_var": 0.008038693107664585,
-            "avg_think_len": 20.213109970092773,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.8758888840675354,
-            "step": 716
-        },
-        {
-            "loss": 7.941527292132378e-05,
-            "reward_mean": 0.9292401671409607,
-            "reward_var": 0.012764506042003632,
-            "avg_think_len": 20.102041244506836,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9825857877731323,
-            "step": 717
-        },
-        {
-            "loss": 6.107671651989222e-05,
-            "reward_mean": 0.8616675734519958,
-            "reward_var": 0.014707041904330254,
-            "avg_think_len": 20.0481014251709,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.8997950553894043,
-            "step": 718
-        },
-        {
-            "loss": 6.771576590836048e-05,
-            "reward_mean": 0.8275463581085205,
-            "reward_var": 0.02220618724822998,
-            "avg_think_len": 20.13033676147461,
-            "reward_format_mean": 0.9984375238418579,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.421875,
-            "reward_response_efficiency_mean": 0.9367479681968689,
-            "step": 719
-        },
-        {
-            "loss": 8.534640073776245e-05,
-            "reward_mean": 0.9469426274299622,
-            "reward_var": 0.005820032209157944,
-            "avg_think_len": 20.16145896911621,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9908952713012695,
-            "step": 720
-        },
-        {
-            "loss": 5.707051604986191e-05,
-            "reward_mean": 0.9266346096992493,
-            "reward_var": 0.005736537743359804,
-            "avg_think_len": 20.12229347229004,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9784131646156311,
-            "step": 721
-        },
-        {
-            "loss": 8.106837049126625e-05,
-            "reward_mean": 0.8701041340827942,
-            "reward_var": 0.007483497262001038,
-            "avg_think_len": 19.780475616455078,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.484375,
-            "reward_response_efficiency_mean": 0.9960417747497559,
-            "step": 722
-        },
-        {
-            "loss": -7.687648758292198e-05,
-            "reward_mean": 0.8721997737884521,
-            "reward_var": 0.014032384380698204,
-            "avg_think_len": 20.077468872070312,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.9575493931770325,
-            "step": 723
-        },
-        {
-            "loss": 7.207365706562996e-05,
-            "reward_mean": 0.9526603817939758,
-            "reward_var": 0.0041970317251980305,
-            "avg_think_len": 20.090579986572266,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9981414079666138,
-            "step": 724
-        },
-        {
-            "loss": 9.656185284256935e-05,
-            "reward_mean": 0.9090673923492432,
-            "reward_var": 0.020747652277350426,
-            "avg_think_len": 20.044363021850586,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9643943309783936,
-            "step": 725
-        },
-        {
-            "loss": 5.777599290013313e-05,
-            "reward_mean": 0.949822187423706,
-            "reward_var": 0.009258689358830452,
-            "avg_think_len": 20.151498794555664,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9867889881134033,
-            "step": 726
-        },
-        {
-            "loss": 7.317191921174526e-05,
-            "reward_mean": 0.8516185879707336,
-            "reward_var": 0.012906073592603207,
-            "avg_think_len": 19.912376403808594,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.9533491134643555,
-            "step": 727
-        },
-        {
-            "loss": 5.532801151275635e-05,
-            "reward_mean": 0.8251526355743408,
-            "reward_var": 0.017185453325510025,
-            "avg_think_len": 19.86951446533203,
-            "reward_format_mean": 0.9998605251312256,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.421875,
-            "reward_response_efficiency_mean": 0.9257498979568481,
-            "step": 728
-        },
-        {
-            "loss": -5.337828770279884e-05,
-            "reward_mean": 0.9111661911010742,
-            "reward_var": 0.014846551232039928,
-            "avg_think_len": 20.101669311523438,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9727898836135864,
-            "step": 729
-        },
-        {
-            "loss": 7.899233605712652e-05,
-            "reward_mean": 0.9253044724464417,
-            "reward_var": 0.0020957274828106165,
-            "avg_think_len": 19.94830894470215,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9980926513671875,
-            "step": 730
-        },
-        {
-            "loss": 6.197625771164894e-05,
-            "reward_mean": 0.8981238603591919,
-            "reward_var": 0.010895948857069016,
-            "avg_think_len": 19.954910278320312,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9987455606460571,
-            "step": 731
-        },
-        {
-            "loss": 0.00016243010759353638,
-            "reward_mean": 0.8740832805633545,
-            "reward_var": 0.010547526180744171,
-            "avg_think_len": 20.04700469970703,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.9650827646255493,
-            "step": 732
-        },
-        {
-            "loss": 0.00015146099030971527,
-            "reward_mean": 0.9079146981239319,
-            "reward_var": 0.008219058625400066,
-            "avg_think_len": 20.148588180541992,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9910338521003723,
-            "step": 733
-        },
-        {
-            "loss": 5.345826502889395e-05,
-            "reward_mean": 0.8500069379806519,
-            "reward_var": 0.01615169644355774,
-            "avg_think_len": 19.978681564331055,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9156526923179626,
-            "step": 734
-        },
-        {
-            "loss": 5.5711716413497925e-05,
-            "reward_mean": 0.9040402173995972,
-            "reward_var": 0.018470266833901405,
-            "avg_think_len": 20.043235778808594,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.965624988079071,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9474108219146729,
-            "step": 735
-        },
-        {
-            "loss": 6.670970469713211e-05,
-            "reward_mean": 0.8761236071586609,
-            "reward_var": 0.006389940623193979,
-            "avg_think_len": 20.02431869506836,
-            "reward_format_mean": 0.999849796295166,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.9577695727348328,
-            "step": 736
-        },
-        {
-            "loss": 8.832523599267006e-05,
-            "reward_mean": 0.9331321716308594,
-            "reward_var": 0.007372037507593632,
-            "avg_think_len": 20.211889266967773,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9981537461280823,
-            "step": 737
-        },
-        {
-            "loss": 2.099527046084404e-05,
-            "reward_mean": 0.9058738946914673,
-            "reward_var": 0.007250954397022724,
-            "avg_think_len": 19.99491310119629,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9984956383705139,
-            "step": 738
-        },
-        {
-            "loss": 3.1948089599609375e-05,
-            "reward_mean": 0.9216229319572449,
-            "reward_var": 0.015468493103981018,
-            "avg_think_len": 20.00888442993164,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9677414894104004,
-            "step": 739
-        },
-        {
-            "loss": 4.277890548110008e-05,
-            "reward_mean": 0.8936486840248108,
-            "reward_var": 0.009811402298510075,
-            "avg_think_len": 19.995887756347656,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9906250238418579,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9902194738388062,
-            "step": 740
-        },
-        {
-            "loss": 5.7246070355176926e-05,
-            "reward_mean": 0.9410608410835266,
-            "reward_var": 0.006837853230535984,
-            "avg_think_len": 20.14539337158203,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9986186027526855,
-            "step": 741
-        },
-        {
-            "loss": 6.004888564348221e-05,
-            "reward_mean": 0.8651273846626282,
-            "reward_var": 0.029757827520370483,
-            "avg_think_len": 20.072898864746094,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9605096578598022,
-            "step": 742
-        },
-        {
-            "loss": 0.0001183563144877553,
-            "reward_mean": 0.9376345276832581,
-            "reward_var": 0.004719728138297796,
-            "avg_think_len": 20.089496612548828,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9849130511283875,
-            "step": 743
-        },
-        {
-            "loss": 8.064275607466698e-05,
-            "reward_mean": 0.9074466228485107,
-            "reward_var": 0.013107425533235073,
-            "avg_think_len": 20.073863983154297,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9422866702079773,
-            "step": 744
-        },
-        {
-            "loss": 0.00010824017226696014,
-            "reward_mean": 0.8694136142730713,
-            "reward_var": 0.014366509392857552,
-            "avg_think_len": 20.100013732910156,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9906250238418579,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9557797312736511,
-            "step": 745
-        },
-        {
-            "loss": 0.00015393132343888283,
-            "reward_mean": 0.9053469896316528,
-            "reward_var": 0.009233941324055195,
-            "avg_think_len": 20.074600219726562,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.996387779712677,
-            "step": 746
-        },
-        {
-            "loss": 7.805600762367249e-05,
-            "reward_mean": 0.9086822271347046,
-            "reward_var": 0.010064498521387577,
-            "avg_think_len": 20.088815689086914,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9784788489341736,
-            "step": 747
-        },
-        {
-            "loss": -4.2246654629707336e-05,
-            "reward_mean": 0.9441023468971252,
-            "reward_var": 0.010090116411447525,
-            "avg_think_len": 20.142074584960938,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.859375,
-            "reward_response_efficiency_mean": 0.9170343279838562,
-            "step": 748
-        },
-        {
-            "loss": 7.884437218308449e-05,
-            "reward_mean": 0.9331312775611877,
-            "reward_var": 0.009378261864185333,
-            "avg_think_len": 20.01303482055664,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9981501698493958,
-            "step": 749
-        },
-        {
-            "loss": 7.784133777022362e-05,
-            "reward_mean": 0.9142245650291443,
-            "reward_var": 0.022762611508369446,
-            "avg_think_len": 20.010093688964844,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9693982005119324,
-            "step": 750
-        },
-        {
-            "loss": 3.0314549803733826e-05,
-            "reward_mean": 0.9189445376396179,
-            "reward_var": 0.0075948406010866165,
-            "avg_think_len": 19.85269546508789,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9882781505584717,
-            "step": 751
-        },
-        {
-            "loss": 5.778716877102852e-05,
-            "reward_mean": 0.7824016213417053,
-            "reward_var": 0.013214527629315853,
-            "avg_think_len": 19.936935424804688,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.1875,
-            "reward_response_efficiency_mean": 0.9577316045761108,
-            "step": 752
-        },
-        {
-            "loss": 7.264688611030579e-05,
-            "reward_mean": 0.9038501977920532,
-            "reward_var": 0.008674582466483116,
-            "avg_think_len": 20.059425354003906,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9591507911682129,
-            "step": 753
-        },
-        {
-            "loss": 0.00010602176189422607,
-            "reward_mean": 0.9021819829940796,
-            "reward_var": 0.004225847776979208,
-            "avg_think_len": 20.1218204498291,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9712280631065369,
-            "step": 754
-        },
-        {
-            "loss": 0.00012551969848573208,
-            "reward_mean": 0.9410971999168396,
-            "reward_var": 0.0072594378143548965,
-            "avg_think_len": 20.05573081970215,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9518887996673584,
-            "step": 755
-        },
-        {
-            "loss": 6.927689537405968e-05,
-            "reward_mean": 0.83600914478302,
-            "reward_var": 0.029694661498069763,
-            "avg_think_len": 19.948036193847656,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.934374988079071,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.9409118294715881,
-            "step": 756
-        },
-        {
-            "loss": 4.7688838094472885e-05,
-            "reward_mean": 0.9506810903549194,
-            "reward_var": 0.007956248708069324,
-            "avg_think_len": 19.880146026611328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.990224301815033,
-            "step": 757
-        },
-        {
-            "loss": 3.320048563182354e-05,
-            "reward_mean": 0.921631395816803,
-            "reward_var": 0.005715799517929554,
-            "avg_think_len": 20.095434188842773,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9365255832672119,
-            "step": 758
-        },
-        {
-            "loss": -2.917577512562275e-05,
-            "reward_mean": 0.8682227730751038,
-            "reward_var": 0.01614736206829548,
-            "avg_think_len": 19.947158813476562,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.9885163307189941,
-            "step": 759
-        },
-        {
-            "loss": 7.279356941580772e-05,
-            "reward_mean": 0.8893479108810425,
-            "reward_var": 0.004780171439051628,
-            "avg_think_len": 20.149999618530273,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9980168342590332,
-            "step": 760
-        },
-        {
-            "loss": 4.613958299160004e-05,
-            "reward_mean": 0.8833764791488647,
-            "reward_var": 0.02068871259689331,
-            "avg_think_len": 19.811744689941406,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9624999761581421,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9460054636001587,
-            "step": 761
-        },
-        {
-            "loss": 2.384372055530548e-05,
-            "reward_mean": 0.9069235920906067,
-            "reward_var": 0.007477891631424427,
-            "avg_think_len": 19.871597290039062,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9870691895484924,
-            "step": 762
-        },
-        {
-            "loss": 9.83739155344665e-05,
-            "reward_mean": 0.9250965714454651,
-            "reward_var": 0.007210154086351395,
-            "avg_think_len": 20.084365844726562,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9503864049911499,
-            "step": 763
-        },
-        {
-            "loss": 0.00010727811604738235,
-            "reward_mean": 0.8980116844177246,
-            "reward_var": 0.008047407492995262,
-            "avg_think_len": 20.03084945678711,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9982966184616089,
-            "step": 764
-        },
-        {
-            "loss": 5.983887240290642e-05,
-            "reward_mean": 0.8769919276237488,
-            "reward_var": 0.018108204007148743,
-            "avg_think_len": 19.998558044433594,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9610930681228638,
-            "step": 765
-        },
-        {
-            "loss": 9.280722588300705e-05,
-            "reward_mean": 0.8402060866355896,
-            "reward_var": 0.00823965109884739,
-            "avg_think_len": 20.205001831054688,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.421875,
-            "reward_response_efficiency_mean": 0.9389495253562927,
-            "step": 766
-        },
-        {
-            "loss": 5.157291889190674e-05,
-            "reward_mean": 0.9293023943901062,
-            "reward_var": 0.003629673970863223,
-            "avg_think_len": 20.099905014038086,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9984595775604248,
-            "step": 767
-        },
-        {
-            "loss": 0.00011562695726752281,
-            "reward_mean": 0.910706639289856,
-            "reward_var": 0.009414373897016048,
-            "avg_think_len": 20.07086181640625,
-            "reward_format_mean": 0.9994419813156128,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.987134575843811,
-            "step": 768
-        },
-        {
-            "loss": 4.4825486838817596e-05,
-            "reward_mean": 0.9257357120513916,
-            "reward_var": 0.016320858150720596,
-            "avg_think_len": 20.162609100341797,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9685678482055664,
-            "step": 769
-        },
-        {
-            "loss": 3.317114897072315e-05,
-            "reward_mean": 0.8826317191123962,
-            "reward_var": 0.008326807990670204,
-            "avg_think_len": 20.126365661621094,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.9836519956588745,
-            "step": 770
-        },
-        {
-            "loss": 0.00010414887219667435,
-            "reward_mean": 0.918595552444458,
-            "reward_var": 0.015359174460172653,
-            "avg_think_len": 20.085390090942383,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.8931320905685425,
-            "step": 771
-        },
-        {
-            "loss": 6.689177826046944e-05,
-            "reward_mean": 0.9244421720504761,
-            "reward_var": 0.011802558787167072,
-            "avg_think_len": 20.22390365600586,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9790187478065491,
-            "step": 772
-        },
-        {
-            "loss": 2.0791776478290558e-05,
-            "reward_mean": 0.9428604245185852,
-            "reward_var": 0.008628729730844498,
-            "avg_think_len": 20.00548553466797,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.9433165788650513,
-            "step": 773
-        },
-        {
-            "loss": 5.984632298350334e-05,
-            "reward_mean": 0.8954288363456726,
-            "reward_var": 0.012288283556699753,
-            "avg_think_len": 20.032512664794922,
-            "reward_format_mean": 0.9998301863670349,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9100101590156555,
-            "step": 774
-        },
-        {
-            "loss": 5.187513306736946e-05,
-            "reward_mean": 0.8913507461547852,
-            "reward_var": 0.013028815388679504,
-            "avg_think_len": 19.96420669555664,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9872779250144958,
-            "step": 775
-        },
-        {
-            "loss": 4.416145384311676e-05,
-            "reward_mean": 0.9330853223800659,
-            "reward_var": 0.006832440849393606,
-            "avg_think_len": 20.092493057250977,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9979662895202637,
-            "step": 776
-        },
-        {
-            "loss": 7.378542795777321e-05,
-            "reward_mean": 0.9528869390487671,
-            "reward_var": 0.006414593197405338,
-            "avg_think_len": 20.179855346679688,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9990474581718445,
-            "step": 777
-        },
-        {
-            "loss": 9.677186608314514e-05,
-            "reward_mean": 0.9176170825958252,
-            "reward_var": 0.0057205455377697945,
-            "avg_think_len": 20.012514114379883,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9985930919647217,
-            "step": 778
-        },
-        {
-            "loss": 9.459164721192792e-05,
-            "reward_mean": 0.8578194379806519,
-            "reward_var": 0.010501453652977943,
-            "avg_think_len": 20.07196044921875,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.453125,
-            "reward_response_efficiency_mean": 0.9781527519226074,
-            "step": 779
-        },
-        {
-            "loss": 4.736008122563362e-05,
-            "reward_mean": 0.9149186611175537,
-            "reward_var": 0.009514826349914074,
-            "avg_think_len": 20.057239532470703,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9877994656562805,
-            "step": 780
-        },
-        {
-            "loss": -3.964683855883777e-06,
-            "reward_mean": 0.9130496978759766,
-            "reward_var": 0.0024439096450805664,
-            "avg_think_len": 20.214401245117188,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9646992087364197,
-            "step": 781
-        },
-        {
-            "loss": 8.4059895016253e-05,
-            "reward_mean": 0.9210495352745056,
-            "reward_var": 0.008106407709419727,
-            "avg_think_len": 20.14358139038086,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9906250238418579,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9904483556747437,
-            "step": 782
-        },
-        {
-            "loss": 7.573774200864136e-05,
-            "reward_mean": 0.8628436326980591,
-            "reward_var": 0.015621723607182503,
-            "avg_think_len": 20.064895629882812,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.8451244235038757,
-            "step": 783
-        },
-        {
-            "loss": 2.3720785975456238e-05,
-            "reward_mean": 0.9761745929718018,
-            "reward_var": 0.005033088847994804,
-            "avg_think_len": 20.10761260986328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.90625,
-            "reward_response_efficiency_mean": 0.9984486699104309,
-            "step": 784
-        },
-        {
-            "loss": 0.00014406302943825722,
-            "reward_mean": 0.9188916087150574,
-            "reward_var": 0.012221170589327812,
-            "avg_think_len": 19.993579864501953,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9880664348602295,
-            "step": 785
-        },
-        {
-            "loss": 0.00010442337952554226,
-            "reward_mean": 0.9097336530685425,
-            "reward_var": 0.007749313488602638,
-            "avg_think_len": 19.98638916015625,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9670599102973938,
-            "step": 786
-        },
-        {
-            "loss": 0.00012833825894631445,
-            "reward_mean": 0.9655964374542236,
-            "reward_var": 0.009073038585484028,
-            "avg_think_len": 20.1568603515625,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.90625,
-            "reward_response_efficiency_mean": 0.971760630607605,
-            "step": 787
-        },
-        {
-            "loss": 3.2978132367134094e-05,
-            "reward_mean": 0.9274208545684814,
-            "reward_var": 0.007232675328850746,
-            "avg_think_len": 19.957279205322266,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9596835374832153,
-            "step": 788
-        },
-        {
-            "loss": 6.408337503671646e-05,
-            "reward_mean": 0.8498486280441284,
-            "reward_var": 0.01915060542523861,
-            "avg_think_len": 19.963048934936523,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.965624988079071,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9181445837020874,
-            "step": 789
-        },
-        {
-            "loss": 3.565358929336071e-05,
-            "reward_mean": 0.9191591143608093,
-            "reward_var": 0.015120146796107292,
-            "avg_think_len": 20.00661277770996,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9750000238418579,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9828869104385376,
-            "step": 790
-        },
-        {
-            "loss": 7.340777665376663e-05,
-            "reward_mean": 0.9601882100105286,
-            "reward_var": 0.006261499132961035,
-            "avg_think_len": 20.065195083618164,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.875,
-            "reward_response_efficiency_mean": 0.9688778519630432,
-            "step": 791
-        },
-        {
-            "loss": 7.456028833985329e-05,
-            "reward_mean": 0.8892087936401367,
-            "reward_var": 0.015233270823955536,
-            "avg_think_len": 19.927906036376953,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9474602341651917,
-            "step": 792
-        },
-        {
-            "loss": 3.728969022631645e-05,
-            "reward_mean": 0.9116075038909912,
-            "reward_var": 0.013213914819061756,
-            "avg_think_len": 20.023605346679688,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9901798963546753,
-            "step": 793
-        },
-        {
-            "loss": 4.0951184928417206e-05,
-            "reward_mean": 0.899749219417572,
-            "reward_var": 0.013136610388755798,
-            "avg_think_len": 19.889070510864258,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9624999761581421,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9177471399307251,
-            "step": 794
-        },
-        {
-            "loss": 4.781363531947136e-05,
-            "reward_mean": 0.9388226866722107,
-            "reward_var": 0.008060231804847717,
-            "avg_think_len": 20.096839904785156,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9584158062934875,
-            "step": 795
-        },
-        {
-            "loss": 8.542882278561592e-05,
-            "reward_mean": 0.8903046250343323,
-            "reward_var": 0.016935249790549278,
-            "avg_think_len": 19.944841384887695,
-            "reward_format_mean": 0.9998539686203003,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9832395315170288,
-            "step": 796
-        },
-        {
-            "loss": 5.156034603714943e-05,
-            "reward_mean": 0.9094349145889282,
-            "reward_var": 0.01765594631433487,
-            "avg_think_len": 20.135557174682617,
-            "reward_format_mean": 0.9997159242630005,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9505240321159363,
-            "step": 797
-        },
-        {
-            "loss": 0.00010877428576350212,
-            "reward_mean": 0.9236046671867371,
-            "reward_var": 0.014078555628657341,
-            "avg_think_len": 20.112703323364258,
-            "reward_format_mean": 0.9998421669006348,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.8977015614509583,
-            "step": 798
-        },
-        {
-            "loss": 6.941007450222969e-05,
-            "reward_mean": 0.8672409057617188,
-            "reward_var": 0.015002356842160225,
-            "avg_think_len": 20.150814056396484,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9220884442329407,
-            "step": 799
-        },
-        {
-            "loss": 6.52708113193512e-05,
-            "reward_mean": 0.9013052582740784,
-            "reward_var": 0.010304149240255356,
-            "avg_think_len": 20.14089584350586,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9802212715148926,
-            "step": 800
-        },
-        {
-            "loss": 9.960890747606754e-05,
-            "reward_mean": 0.917590320110321,
-            "reward_var": 0.006804110016673803,
-            "avg_think_len": 20.01895523071289,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9984862208366394,
-            "step": 801
-        },
-        {
-            "loss": 8.429284207522869e-06,
-            "reward_mean": 0.9428123235702515,
-            "reward_var": 0.00443669967353344,
-            "avg_think_len": 20.0191593170166,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9899994134902954,
-            "step": 802
-        },
-        {
-            "loss": 5.971640348434448e-06,
-            "reward_mean": 0.9037929177284241,
-            "reward_var": 0.0048344978131353855,
-            "avg_think_len": 20.028175354003906,
-            "reward_format_mean": 0.99913489818573,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9754118919372559,
-            "step": 803
-        },
-        {
-            "loss": 4.735402762889862e-05,
-            "reward_mean": 0.9247298240661621,
-            "reward_var": 0.009639669209718704,
-            "avg_think_len": 19.94532585144043,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.948919415473938,
-            "step": 804
-        },
-        {
-            "loss": 8.892454206943512e-05,
-            "reward_mean": 0.8834150433540344,
-            "reward_var": 0.012218650430440903,
-            "avg_think_len": 20.167184829711914,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9555349349975586,
-            "step": 805
-        },
-        {
-            "loss": 4.703877493739128e-05,
-            "reward_mean": 0.9664703607559204,
-            "reward_var": 0.005120229907333851,
-            "avg_think_len": 20.10000991821289,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.875,
-            "reward_response_efficiency_mean": 0.9908814430236816,
-            "step": 806
-        },
-        {
-            "loss": 0.00016234011854976416,
-            "reward_mean": 0.954765796661377,
-            "reward_var": 0.002896468387916684,
-            "avg_think_len": 20.1285343170166,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.9909380674362183,
-            "step": 807
-        },
-        {
-            "loss": 8.272816194221377e-05,
-            "reward_mean": 0.8981493711471558,
-            "reward_var": 0.010122640058398247,
-            "avg_think_len": 19.99848747253418,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9832228422164917,
-            "step": 808
-        },
-        {
-            "loss": 6.692332681268454e-05,
-            "reward_mean": 0.8593502044677734,
-            "reward_var": 0.026057882234454155,
-            "avg_think_len": 19.85141944885254,
-            "reward_format_mean": 0.9998337626457214,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9531919956207275,
-            "step": 809
-        },
-        {
-            "loss": 9.223166853189468e-05,
-            "reward_mean": 0.9065835475921631,
-            "reward_var": 0.008838065899908543,
-            "avg_think_len": 20.063854217529297,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.9857094287872314,
-            "step": 810
-        },
-        {
-            "loss": 6.165634840726852e-05,
-            "reward_mean": 0.9410060048103333,
-            "reward_var": 0.008504809811711311,
-            "avg_think_len": 19.997148513793945,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9983991980552673,
-            "step": 811
-        },
-        {
-            "loss": 2.8694048523902893e-05,
-            "reward_mean": 0.8942643404006958,
-            "reward_var": 0.015137744136154652,
-            "avg_think_len": 20.08423614501953,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9833073616027832,
-            "step": 812
-        },
-        {
-            "loss": 5.0337985157966614e-05,
-            "reward_mean": 0.9035418033599854,
-            "reward_var": 0.013232184574007988,
-            "avg_think_len": 19.849510192871094,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9891674518585205,
-            "step": 813
-        },
-        {
-            "loss": 0.00010170880705118179,
-            "reward_mean": 0.9441356062889099,
-            "reward_var": 0.005900296848267317,
-            "avg_think_len": 20.077159881591797,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.9640423655509949,
-            "step": 814
-        },
-        {
-            "loss": 5.33335842192173e-05,
-            "reward_mean": 0.8453222513198853,
-            "reward_var": 0.02045384794473648,
-            "avg_think_len": 20.01993751525879,
-            "reward_format_mean": 0.9996565580368042,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.834757387638092,
-            "step": 815
-        },
-        {
-            "loss": 6.266171112656593e-05,
-            "reward_mean": 0.8884530067443848,
-            "reward_var": 0.005813511088490486,
-            "avg_think_len": 19.985828399658203,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9444369077682495,
-            "step": 816
-        },
-        {
-            "loss": 1.2137927114963531e-05,
-            "reward_mean": 0.9445193409919739,
-            "reward_var": 0.008772075176239014,
-            "avg_think_len": 19.997983932495117,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9968276619911194,
-            "step": 817
-        },
-        {
-            "loss": -2.151448279619217e-05,
-            "reward_mean": 0.9339240193367004,
-            "reward_var": 0.00879648420959711,
-            "avg_think_len": 20.09349250793457,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9544464945793152,
-            "step": 818
-        },
-        {
-            "loss": 9.954115375876427e-05,
-            "reward_mean": 0.9195728302001953,
-            "reward_var": 0.009866884909570217,
-            "avg_think_len": 20.157272338867188,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9907910227775574,
-            "step": 819
-        },
-        {
-            "loss": 4.266700489097275e-05,
-            "reward_mean": 0.9224466681480408,
-            "reward_var": 0.007468800991773605,
-            "avg_think_len": 19.924482345581055,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9397866129875183,
-            "step": 820
-        },
-        {
-            "loss": 8.788518607616425e-05,
-            "reward_mean": 0.9526236057281494,
-            "reward_var": 0.0030648040119558573,
-            "avg_think_len": 20.02740478515625,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9979944229125977,
-            "step": 821
-        },
-        {
-            "loss": 7.419055327773094e-05,
-            "reward_mean": 0.9376868605613708,
-            "reward_var": 0.010393060743808746,
-            "avg_think_len": 20.096603393554688,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.988247275352478,
-            "step": 822
-        },
-        {
-            "loss": 9.528493683319539e-05,
-            "reward_mean": 0.8984339833259583,
-            "reward_var": 0.021793490275740623,
-            "avg_think_len": 20.11931610107422,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9531108140945435,
-            "step": 823
-        },
-        {
-            "loss": 0.0001512260059826076,
-            "reward_mean": 0.9224989414215088,
-            "reward_var": 0.004452274646610022,
-            "avg_think_len": 20.179668426513672,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9868707656860352,
-            "step": 824
-        },
-        {
-            "loss": 7.074803579598665e-05,
-            "reward_mean": 0.8812389373779297,
-            "reward_var": 0.013623167760670185,
-            "avg_think_len": 20.017127990722656,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9780809879302979,
-            "step": 825
-        },
-        {
-            "loss": 8.188632637029514e-05,
-            "reward_mean": 0.9153236150741577,
-            "reward_var": 0.013629918918013573,
-            "avg_think_len": 19.988521575927734,
-            "reward_format_mean": 0.9997106790542603,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9740837216377258,
-            "step": 826
-        },
-        {
-            "loss": 8.697866724105552e-05,
-            "reward_mean": 0.9098590612411499,
-            "reward_var": 0.019701115787029266,
-            "avg_think_len": 19.992050170898438,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.918749988079071,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9394361972808838,
-            "step": 827
-        },
-        {
-            "loss": 6.75637274980545e-05,
-            "reward_mean": 0.8708019256591797,
-            "reward_var": 0.007499877829104662,
-            "avg_think_len": 20.1225528717041,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.9519578218460083,
-            "step": 828
-        },
-        {
-            "loss": 6.415368989109993e-05,
-            "reward_mean": 0.8647738695144653,
-            "reward_var": 0.024131938815116882,
-            "avg_think_len": 20.244977951049805,
-            "reward_format_mean": 0.9998592138290405,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9279862642288208,
-            "step": 829
-        },
-        {
-            "loss": 6.84238038957119e-05,
-            "reward_mean": 0.9391838908195496,
-            "reward_var": 0.0053947120904922485,
-            "avg_think_len": 20.057621002197266,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9911108016967773,
-            "step": 830
-        },
-        {
-            "loss": 6.298068910837173e-05,
-            "reward_mean": 0.9042699933052063,
-            "reward_var": 0.01027875766158104,
-            "avg_think_len": 20.10988998413086,
-            "reward_format_mean": 0.9988839626312256,
-            "reward_time_mean": 0.987500011920929,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9900710582733154,
-            "step": 831
-        },
-        {
-            "loss": 0.00010941363871097565,
-            "reward_mean": 0.922200620174408,
-            "reward_var": 0.007790267001837492,
-            "avg_think_len": 20.156543731689453,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9856773018836975,
-            "step": 832
-        },
-        {
-            "loss": 5.726376548409462e-05,
-            "reward_mean": 0.9087549448013306,
-            "reward_var": 0.006275765597820282,
-            "avg_think_len": 19.81119155883789,
-            "reward_format_mean": 0.9994611740112305,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.998058557510376,
-            "step": 833
-        },
-        {
-            "loss": 3.8869213312864304e-05,
-            "reward_mean": 0.9082181453704834,
-            "reward_var": 0.018612893298268318,
-            "avg_think_len": 20.022104263305664,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9453725814819336,
-            "step": 834
-        },
-        {
-            "loss": 3.0667055398225784e-05,
-            "reward_mean": 0.8853037357330322,
-            "reward_var": 0.014397557824850082,
-            "avg_think_len": 20.08745002746582,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9787151217460632,
-            "step": 835
-        },
-        {
-            "loss": 7.067667320370674e-05,
-            "reward_mean": 0.9176622629165649,
-            "reward_var": 0.023427262902259827,
-            "avg_think_len": 20.105924606323242,
-            "reward_format_mean": 0.9983552694320679,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9379185438156128,
-            "step": 836
-        },
-        {
-            "loss": 4.239077679812908e-05,
-            "reward_mean": 0.922610878944397,
-            "reward_var": 0.020678462460637093,
-            "avg_think_len": 20.02434730529785,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.965624988079071,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.927943229675293,
-            "step": 837
-        },
-        {
-            "loss": 5.549611523747444e-05,
-            "reward_mean": 0.8545433878898621,
-            "reward_var": 0.016368914395570755,
-            "avg_think_len": 20.011062622070312,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9781249761581421,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9244234561920166,
-            "step": 838
-        },
-        {
-            "loss": 0.00010816473513841629,
-            "reward_mean": 0.9206356406211853,
-            "reward_var": 0.010562250390648842,
-            "avg_think_len": 20.23549461364746,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.979417622089386,
-            "step": 839
-        },
-        {
-            "loss": 0.0001609226455911994,
-            "reward_mean": 0.9392078518867493,
-            "reward_var": 0.01071553211659193,
-            "avg_think_len": 20.149402618408203,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9912062883377075,
-            "step": 840
-        },
-        {
-            "loss": 1.68713741004467e-05,
-            "reward_mean": 0.8746179342269897,
-            "reward_var": 0.008663719519972801,
-            "avg_think_len": 20.0194149017334,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9828469157218933,
-            "step": 841
-        },
-        {
-            "loss": 0.00010773399844765663,
-            "reward_mean": 0.9489008188247681,
-            "reward_var": 0.004326456226408482,
-            "avg_think_len": 20.13713264465332,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9987284541130066,
-            "step": 842
-        },
-        {
-            "loss": 4.776963032782078e-05,
-            "reward_mean": 0.8379930257797241,
-            "reward_var": 0.015567025169730186,
-            "avg_think_len": 20.146635055541992,
-            "reward_format_mean": 0.9996874928474426,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.40625,
-            "reward_response_efficiency_mean": 0.9772847890853882,
-            "step": 843
-        },
-        {
-            "loss": -4.805624485015869e-06,
-            "reward_mean": 0.8572993874549866,
-            "reward_var": 0.01330655999481678,
-            "avg_think_len": 20.175838470458984,
-            "reward_format_mean": 0.9998592138290405,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.9762134552001953,
-            "step": 844
-        },
-        {
-            "loss": 6.443483289331198e-05,
-            "reward_mean": 0.8694795370101929,
-            "reward_var": 0.017942463979125023,
-            "avg_think_len": 19.916702270507812,
-            "reward_format_mean": 0.9998579621315002,
-            "reward_time_mean": 0.971875011920929,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.9593102931976318,
-            "step": 845
-        },
-        {
-            "loss": 9.16937569854781e-05,
-            "reward_mean": 0.9527356028556824,
-            "reward_var": 0.006399884819984436,
-            "avg_think_len": 19.98532485961914,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9984423518180847,
-            "step": 846
-        },
-        {
-            "loss": 8.49762000143528e-05,
-            "reward_mean": 0.8752162456512451,
-            "reward_var": 0.013553112745285034,
-            "avg_think_len": 19.936538696289062,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9883646965026855,
-            "step": 847
-        },
-        {
-            "loss": 6.817933171987534e-05,
-            "reward_mean": 0.8932223320007324,
-            "reward_var": 0.017950283363461494,
-            "avg_think_len": 19.98700523376465,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.963514506816864,
-            "step": 848
-        },
-        {
-            "loss": 7.532304152846336e-05,
-            "reward_mean": 0.9058864116668701,
-            "reward_var": 0.008372459560632706,
-            "avg_think_len": 20.12486457824707,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9985458850860596,
-            "step": 849
-        },
-        {
-            "loss": 5.390268051996827e-05,
-            "reward_mean": 0.8716285824775696,
-            "reward_var": 0.007534292060881853,
-            "avg_think_len": 20.010122299194336,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.8771395683288574,
-            "step": 850
-        },
-        {
-            "loss": 7.202662527561188e-05,
-            "reward_mean": 0.8793367147445679,
-            "reward_var": 0.013721526600420475,
-            "avg_think_len": 20.056066513061523,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.986096978187561,
-            "step": 851
-        },
-        {
-            "loss": 6.81350938975811e-05,
-            "reward_mean": 0.8618079423904419,
-            "reward_var": 0.021014725789427757,
-            "avg_think_len": 20.085031509399414,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9784815907478333,
-            "step": 852
-        },
-        {
-            "loss": 9.592436254024506e-05,
-            "reward_mean": 0.8877320289611816,
-            "reward_var": 0.013168266974389553,
-            "avg_think_len": 20.080312728881836,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9728026986122131,
-            "step": 853
-        },
-        {
-            "loss": 0.00010482314974069595,
-            "reward_mean": 0.9277626276016235,
-            "reward_var": 0.007795404642820358,
-            "avg_think_len": 20.211055755615234,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9766755700111389,
-            "step": 854
-        },
-        {
-            "loss": 7.388810627162457e-05,
-            "reward_mean": 0.8522975444793701,
-            "reward_var": 0.02298242598772049,
-            "avg_think_len": 20.092252731323242,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9560654163360596,
-            "step": 855
-        },
-        {
-            "loss": 5.5571552366018295e-05,
-            "reward_mean": 0.9428439140319824,
-            "reward_var": 0.011658048257231712,
-            "avg_think_len": 19.909645080566406,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9901255965232849,
-            "step": 856
-        },
-        {
-            "loss": 6.962008774280548e-05,
-            "reward_mean": 0.9241783022880554,
-            "reward_var": 0.007165761198848486,
-            "avg_think_len": 20.149553298950195,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.931088387966156,
-            "step": 857
-        },
-        {
-            "loss": 5.218200385570526e-05,
-            "reward_mean": 0.8481568098068237,
-            "reward_var": 0.01965506188571453,
-            "avg_think_len": 20.17230224609375,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.9551270604133606,
-            "step": 858
-        },
-        {
-            "loss": 7.926137186586857e-05,
-            "reward_mean": 0.9268490076065063,
-            "reward_var": 0.005178053863346577,
-            "avg_think_len": 20.10061264038086,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9730208516120911,
-            "step": 859
-        },
-        {
-            "loss": 3.0416063964366913e-05,
-            "reward_mean": 0.9459465146064758,
-            "reward_var": 0.0075218030251562595,
-            "avg_think_len": 19.975984573364258,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9869109988212585,
-            "step": 860
-        },
-        {
-            "loss": 8.849035657476634e-05,
-            "reward_mean": 0.9535320401191711,
-            "reward_var": 0.009277158416807652,
-            "avg_think_len": 20.119644165039062,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.875,
-            "reward_response_efficiency_mean": 0.9547534584999084,
-            "step": 861
-        },
-        {
-            "loss": 0.00010650733020156622,
-            "reward_mean": 0.9879213571548462,
-            "reward_var": 0.004878593608736992,
-            "avg_think_len": 20.16741943359375,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.96875,
-            "reward_response_efficiency_mean": 0.9985604286193848,
-            "step": 862
-        },
-        {
-            "loss": 6.508547812700272e-05,
-            "reward_mean": 0.8750743865966797,
-            "reward_var": 0.008712628856301308,
-            "avg_think_len": 20.10645866394043,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9221727848052979,
-            "step": 863
-        },
-        {
-            "loss": 7.431686390191317e-05,
-            "reward_mean": 0.8607576489448547,
-            "reward_var": 0.017573071643710136,
-            "avg_think_len": 20.191242218017578,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.974280595779419,
-            "step": 864
-        },
-        {
-            "loss": 0.00011442182585597038,
-            "reward_mean": 0.9310110807418823,
-            "reward_var": 0.009557586163282394,
-            "avg_think_len": 20.140153884887695,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.958419144153595,
-            "step": 865
-        },
-        {
-            "loss": 4.574097692966461e-05,
-            "reward_mean": 0.9496046900749207,
-            "reward_var": 0.012350311502814293,
-            "avg_think_len": 20.17732810974121,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.9859188795089722,
-            "step": 866
-        },
-        {
-            "loss": 5.909707397222519e-05,
-            "reward_mean": 0.8447063565254211,
-            "reward_var": 0.015265759080648422,
-            "avg_think_len": 20.043386459350586,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.4375,
-            "reward_response_efficiency_mean": 0.9882006645202637,
-            "step": 867
-        },
-        {
-            "loss": 5.124695599079132e-05,
-            "reward_mean": 0.9058447480201721,
-            "reward_var": 0.006423196289688349,
-            "avg_think_len": 20.07097053527832,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9983789920806885,
-            "step": 868
-        },
-        {
-            "loss": 0.00010128412395715714,
-            "reward_mean": 0.8677113652229309,
-            "reward_var": 0.01154590118676424,
-            "avg_think_len": 19.937156677246094,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.484375,
-            "reward_response_efficiency_mean": 0.9864703416824341,
-            "step": 869
-        },
-        {
-            "loss": 6.788643077015877e-05,
-            "reward_mean": 0.8514608144760132,
-            "reward_var": 0.013580575585365295,
-            "avg_think_len": 20.14016342163086,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.9370933175086975,
-            "step": 870
-        },
-        {
-            "loss": 7.662968710064888e-05,
-            "reward_mean": 0.883774995803833,
-            "reward_var": 0.01746929995715618,
-            "avg_think_len": 19.975812911987305,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9569748640060425,
-            "step": 871
-        },
-        {
-            "loss": 8.303765207529068e-05,
-            "reward_mean": 0.8195396065711975,
-            "reward_var": 0.02522299252450466,
-            "avg_think_len": 20.061193466186523,
-            "reward_format_mean": 0.9990828037261963,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.453125,
-            "reward_response_efficiency_mean": 0.9040757417678833,
-            "step": 872
-        },
-        {
-            "loss": 3.622658550739288e-05,
-            "reward_mean": 0.9285577535629272,
-            "reward_var": 0.012602807022631168,
-            "avg_think_len": 20.181705474853516,
-            "reward_format_mean": 0.9992559552192688,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9368501901626587,
-            "step": 873
-        },
-        {
-            "loss": 6.858212873339653e-05,
-            "reward_mean": 0.9171494245529175,
-            "reward_var": 0.015255440026521683,
-            "avg_think_len": 20.134431838989258,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9750000238418579,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9904725551605225,
-            "step": 874
-        },
-        {
-            "loss": 3.248569555580616e-06,
-            "reward_mean": 0.9061734676361084,
-            "reward_var": 0.011332781054079533,
-            "avg_think_len": 19.936784744262695,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9371938705444336,
-            "step": 875
-        },
-        {
-            "loss": 5.240878090262413e-05,
-            "reward_mean": 0.8555601835250854,
-            "reward_var": 0.020491916686296463,
-            "avg_think_len": 20.064390182495117,
-            "reward_format_mean": 0.9997735619544983,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.9068419337272644,
-            "step": 876
-        },
-        {
-            "loss": 3.905268386006355e-05,
-            "reward_mean": 0.9281378984451294,
-            "reward_var": 0.0065059298649430275,
-            "avg_think_len": 20.031089782714844,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9469267129898071,
-            "step": 877
-        },
-        {
-            "loss": 6.269058212637901e-05,
-            "reward_mean": 0.8967553973197937,
-            "reward_var": 0.016076156869530678,
-            "avg_think_len": 19.972009658813477,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.8995215892791748,
-            "step": 878
-        },
-        {
-            "loss": 6.794510409235954e-05,
-            "reward_mean": 0.9294946193695068,
-            "reward_var": 0.03379928693175316,
-            "avg_think_len": 20.15364646911621,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.859375,
-            "reward_response_efficiency_mean": 0.9367287158966064,
-            "step": 879
-        },
-        {
-            "loss": 6.550768739543855e-05,
-            "reward_mean": 0.8969416618347168,
-            "reward_var": 0.028503233566880226,
-            "avg_think_len": 19.9543399810791,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9627666473388672,
-            "step": 880
-        },
-        {
-            "loss": 6.597826723009348e-05,
-            "reward_mean": 0.9486159682273865,
-            "reward_var": 0.0074039241299033165,
-            "avg_think_len": 20.07952117919922,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9975886344909668,
-            "step": 881
-        },
-        {
-            "loss": 0.000158761627972126,
-            "reward_mean": 0.8941953182220459,
-            "reward_var": 0.007128675933927298,
-            "avg_think_len": 19.953378677368164,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.998656153678894,
-            "step": 882
-        },
-        {
-            "loss": 7.82310962677002e-05,
-            "reward_mean": 0.9141378402709961,
-            "reward_var": 0.016229234635829926,
-            "avg_think_len": 20.02638053894043,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9846763014793396,
-            "step": 883
-        },
-        {
-            "loss": -3.514089621603489e-05,
-            "reward_mean": 0.972337007522583,
-            "reward_var": 0.00432569719851017,
-            "avg_think_len": 20.17545509338379,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.890625,
-            "reward_response_efficiency_mean": 0.9987230896949768,
-            "step": 884
-        },
-        {
-            "loss": 9.31612157728523e-05,
-            "reward_mean": 0.9162299633026123,
-            "reward_var": 0.030067000538110733,
-            "avg_think_len": 20.050609588623047,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.90625,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.9149197340011597,
-            "step": 885
-        },
-        {
-            "loss": 6.168009713292122e-05,
-            "reward_mean": 0.9513460397720337,
-            "reward_var": 0.007683983072638512,
-            "avg_think_len": 20.040809631347656,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.9616339206695557,
-            "step": 886
-        },
-        {
-            "loss": 6.794696673750877e-05,
-            "reward_mean": 0.8901491165161133,
-            "reward_var": 0.011703292839229107,
-            "avg_think_len": 20.20701789855957,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9824716448783875,
-            "step": 887
-        },
-        {
-            "loss": 6.271572783589363e-05,
-            "reward_mean": 0.9171789288520813,
-            "reward_var": 0.005152874626219273,
-            "avg_think_len": 20.08086395263672,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9655908346176147,
-            "step": 888
-        },
-        {
-            "loss": 7.0993322879076e-05,
-            "reward_mean": 0.8452761769294739,
-            "reward_var": 0.0065297153778374195,
-            "avg_think_len": 20.122838973999023,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.40625,
-            "reward_response_efficiency_mean": 0.9904796481132507,
-            "step": 889
-        },
-        {
-            "loss": 6.143003702163696e-05,
-            "reward_mean": 0.9405695796012878,
-            "reward_var": 0.005362424999475479,
-            "avg_think_len": 20.178569793701172,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9497783184051514,
-            "step": 890
-        },
-        {
-            "loss": 9.73232090473175e-07,
-            "reward_mean": 0.9038636684417725,
-            "reward_var": 0.006115122698247433,
-            "avg_think_len": 19.984479904174805,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9904545545578003,
-            "step": 891
-        },
-        {
-            "loss": 6.0867052525281906e-05,
-            "reward_mean": 0.899573564529419,
-            "reward_var": 0.017793046310544014,
-            "avg_think_len": 20.11858367919922,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9750000238418579,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9826692342758179,
-            "step": 892
-        },
-        {
-            "loss": 9.322981350123882e-05,
-            "reward_mean": 0.8437494039535522,
-            "reward_var": 0.018845342099666595,
-            "avg_think_len": 20.00446891784668,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.9374978542327881,
-            "step": 893
-        },
-        {
-            "loss": 2.6891939342021942e-05,
-            "reward_mean": 0.9018396139144897,
-            "reward_var": 0.0048783039674162865,
-            "avg_think_len": 20.17405128479004,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9979836344718933,
-            "step": 894
-        },
-        {
-            "loss": 0.00011382834054529667,
-            "reward_mean": 0.868204653263092,
-            "reward_var": 0.017082136124372482,
-            "avg_think_len": 20.08702278137207,
-            "reward_format_mean": 0.9964914917945862,
-            "reward_time_mean": 0.987500011920929,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.9419521689414978,
-            "step": 895
-        },
-        {
-            "loss": 6.365962326526642e-05,
-            "reward_mean": 0.9138029217720032,
-            "reward_var": 0.009134414605796337,
-            "avg_think_len": 20.11430549621582,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9520865678787231,
-            "step": 896
-        },
-        {
-            "loss": 0.00010829977691173553,
-            "reward_mean": 0.8898985981941223,
-            "reward_var": 0.008331329561769962,
-            "avg_think_len": 19.965917587280273,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9814695119857788,
-            "step": 897
-        },
-        {
-            "loss": 1.544831320643425e-05,
-            "reward_mean": 0.894580066204071,
-            "reward_var": 0.01521562784910202,
-            "avg_think_len": 20.05417823791504,
-            "reward_format_mean": 0.999813973903656,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9691311717033386,
-            "step": 898
-        },
-        {
-            "loss": 7.877964526414871e-05,
-            "reward_mean": 0.9156697392463684,
-            "reward_var": 0.01070893369615078,
-            "avg_think_len": 20.164098739624023,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9908040165901184,
-            "step": 899
-        },
-        {
-            "loss": -3.4981523640453815e-05,
-            "reward_mean": 0.9089082479476929,
-            "reward_var": 0.010584583505988121,
-            "avg_think_len": 20.06460189819336,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9481331706047058,
-            "step": 900
-        },
-        {
-            "loss": 4.848511889576912e-05,
-            "reward_mean": 0.876773476600647,
-            "reward_var": 0.007364524062722921,
-            "avg_think_len": 20.098604202270508,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9914689064025879,
-            "step": 901
-        },
-        {
-            "loss": 7.581640966236591e-05,
-            "reward_mean": 0.9161792993545532,
-            "reward_var": 0.018148096278309822,
-            "avg_think_len": 20.041196823120117,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9615920186042786,
-            "step": 902
-        },
-        {
-            "loss": 3.785453736782074e-05,
-            "reward_mean": 0.9370630979537964,
-            "reward_var": 0.009779923595488071,
-            "avg_think_len": 20.12558937072754,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9982525110244751,
-            "step": 903
-        },
-        {
-            "loss": 7.517435005865991e-05,
-            "reward_mean": 0.9409593343734741,
-            "reward_var": 0.007945702411234379,
-            "avg_think_len": 20.0598087310791,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9982121586799622,
-            "step": 904
-        },
-        {
-            "loss": 0.00012561283074319363,
-            "reward_mean": 0.9464272260665894,
-            "reward_var": 0.01259004045277834,
-            "avg_think_len": 20.040374755859375,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.859375,
-            "reward_response_efficiency_mean": 0.9419587254524231,
-            "step": 905
-        },
-        {
-            "loss": 4.477379843592644e-05,
-            "reward_mean": 0.8785226345062256,
-            "reward_var": 0.0009779101237654686,
-            "avg_think_len": 20.109407424926758,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.998465359210968,
-            "step": 906
-        },
-        {
-            "loss": 9.990343824028969e-05,
-            "reward_mean": 0.8677515983581543,
-            "reward_var": 0.005719544366002083,
-            "avg_think_len": 20.09479331970215,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9866315126419067,
-            "step": 907
-        },
-        {
-            "loss": 5.2900053560733795e-05,
-            "reward_mean": 0.9332441091537476,
-            "reward_var": 0.005467966198921204,
-            "avg_think_len": 20.17922592163086,
-            "reward_format_mean": 0.9994611740112305,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9835149645805359,
-            "step": 908
-        },
-        {
-            "loss": 6.490573287010193e-05,
-            "reward_mean": 0.8816057443618774,
-            "reward_var": 0.025148939341306686,
-            "avg_think_len": 19.98822593688965,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.9326728582382202,
-            "step": 909
-        },
-        {
-            "loss": 8.56093829497695e-05,
-            "reward_mean": 0.9248669147491455,
-            "reward_var": 0.0056354752741754055,
-            "avg_think_len": 20.125276565551758,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9400925636291504,
-            "step": 910
-        },
-        {
-            "loss": 2.5996938347816467e-05,
-            "reward_mean": 0.8942037224769592,
-            "reward_var": 0.0076716127805411816,
-            "avg_think_len": 20.22496223449707,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9986898899078369,
-            "step": 911
-        },
-        {
-            "loss": 6.278412183746696e-05,
-            "reward_mean": 0.9187933206558228,
-            "reward_var": 0.011463243514299393,
-            "avg_think_len": 19.997909545898438,
-            "reward_format_mean": 0.9998183250427246,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9597299098968506,
-            "step": 912
-        },
-        {
-            "loss": 5.519110709428787e-05,
-            "reward_mean": 0.8980003595352173,
-            "reward_var": 0.011439851485192776,
-            "avg_think_len": 19.962018966674805,
-            "reward_format_mean": 0.9997971057891846,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9984543919563293,
-            "step": 913
-        },
-        {
-            "loss": 0.00011971888307016343,
-            "reward_mean": 0.9343664646148682,
-            "reward_var": 0.020237253978848457,
-            "avg_think_len": 20.189952850341797,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.9405912160873413,
-            "step": 914
-        },
-        {
-            "loss": 6.266310811042786e-05,
-            "reward_mean": 0.9799759387969971,
-            "reward_var": 0.0030260663479566574,
-            "avg_think_len": 20.176380157470703,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.953125,
-            "reward_response_efficiency_mean": 0.9699038863182068,
-            "step": 915
-        },
-        {
-            "loss": 3.224797546863556e-05,
-            "reward_mean": 0.9311960339546204,
-            "reward_var": 0.009446804411709309,
-            "avg_think_len": 20.0806941986084,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9591591954231262,
-            "step": 916
-        },
-        {
-            "loss": -0.00013088388368487358,
-            "reward_mean": 0.9411358833312988,
-            "reward_var": 0.005718282889574766,
-            "avg_think_len": 20.202159881591797,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9989186525344849,
-            "step": 917
-        },
-        {
-            "loss": 4.781363531947136e-05,
-            "reward_mean": 0.8961913585662842,
-            "reward_var": 0.013484714552760124,
-            "avg_think_len": 20.183748245239258,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9910155534744263,
-            "step": 918
-        },
-        {
-            "loss": 7.433490827679634e-05,
-            "reward_mean": 0.8873787522315979,
-            "reward_var": 0.015615609474480152,
-            "avg_think_len": 20.169553756713867,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9713902473449707,
-            "step": 919
-        },
-        {
-            "loss": 4.465749952942133e-05,
-            "reward_mean": 0.8874850869178772,
-            "reward_var": 0.01744483783841133,
-            "avg_think_len": 20.115245819091797,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.859375,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9718155264854431,
-            "step": 920
-        },
-        {
-            "loss": 7.677963003516197e-05,
-            "reward_mean": 0.9214073419570923,
-            "reward_var": 0.006678182631731033,
-            "avg_think_len": 20.236452102661133,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9981296062469482,
-            "step": 921
-        },
-        {
-            "loss": 4.3990323320031166e-05,
-            "reward_mean": 0.9117329120635986,
-            "reward_var": 0.002191726816818118,
-            "avg_think_len": 20.140161514282227,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.9906814694404602,
-            "step": 922
-        },
-        {
-            "loss": 6.556663720402867e-05,
-            "reward_mean": 0.9216079115867615,
-            "reward_var": 0.007946988567709923,
-            "avg_think_len": 20.043838500976562,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9676812887191772,
-            "step": 923
-        },
-        {
-            "loss": 6.184633821249008e-05,
-            "reward_mean": 0.8798043727874756,
-            "reward_var": 0.011113429442048073,
-            "avg_think_len": 20.081703186035156,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.9879672527313232,
-            "step": 924
-        },
-        {
-            "loss": 0.00010486063547432423,
-            "reward_mean": 0.9463376998901367,
-            "reward_var": 0.009781260043382645,
-            "avg_think_len": 20.022750854492188,
-            "reward_format_mean": 0.9998641014099121,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.9729867577552795,
-            "step": 925
-        },
-        {
-            "loss": 0.00015329339657910168,
-            "reward_mean": 0.8651770949363708,
-            "reward_var": 0.03388100489974022,
-            "avg_think_len": 20.029401779174805,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9138333797454834,
-            "step": 926
-        },
-        {
-            "loss": 0.00012331543257459998,
-            "reward_mean": 0.9079798460006714,
-            "reward_var": 0.009858855977654457,
-            "avg_think_len": 20.140281677246094,
-            "reward_format_mean": 0.9993489980697632,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.935695469379425,
-            "step": 927
-        },
-        {
-            "loss": 2.920767292380333e-05,
-            "reward_mean": 0.9266597032546997,
-            "reward_var": 0.014849048107862473,
-            "avg_think_len": 20.054122924804688,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9566388726234436,
-            "step": 928
-        },
-        {
-            "loss": 7.426703814417124e-05,
-            "reward_mean": 0.8834879398345947,
-            "reward_var": 0.00593556510284543,
-            "avg_think_len": 20.132545471191406,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9245768189430237,
-            "step": 929
-        },
-        {
-            "loss": 8.812663145363331e-05,
-            "reward_mean": 0.8824639916419983,
-            "reward_var": 0.010863803327083588,
-            "avg_think_len": 20.16145896911621,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9986059069633484,
-            "step": 930
-        },
-        {
-            "loss": 5.1842427637893707e-05,
-            "reward_mean": 0.8886590003967285,
-            "reward_var": 0.011977042071521282,
-            "avg_think_len": 20.055940628051758,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9765109419822693,
-            "step": 931
-        },
-        {
-            "loss": 0.00012570060789585114,
-            "reward_mean": 0.9051654934883118,
-            "reward_var": 0.00677904998883605,
-            "avg_think_len": 20.168611526489258,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9800370931625366,
-            "step": 932
-        },
-        {
-            "loss": 6.591994315385818e-05,
-            "reward_mean": 0.9131443500518799,
-            "reward_var": 0.012274019420146942,
-            "avg_think_len": 19.82550621032715,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9557021856307983,
-            "step": 933
-        },
-        {
-            "loss": 3.855861723423004e-05,
-            "reward_mean": 0.9410823583602905,
-            "reward_var": 0.007980532944202423,
-            "avg_think_len": 20.137989044189453,
-            "reward_format_mean": 0.999804675579071,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9988998174667358,
-            "step": 934
-        },
-        {
-            "loss": 6.523076444864273e-05,
-            "reward_mean": 0.9320588707923889,
-            "reward_var": 0.010350645519793034,
-            "avg_think_len": 20.032058715820312,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.987500011920929,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9907355308532715,
-            "step": 935
-        },
-        {
-            "loss": 9.447499178349972e-05,
-            "reward_mean": 0.9324929118156433,
-            "reward_var": 0.0053834207355976105,
-            "avg_think_len": 20.128799438476562,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9799715280532837,
-            "step": 936
-        },
-        {
-            "loss": 6.661657243967056e-05,
-            "reward_mean": 0.9458729028701782,
-            "reward_var": 0.013018155470490456,
-            "avg_think_len": 20.130701065063477,
-            "reward_format_mean": 0.9997971057891846,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.9711945652961731,
-            "step": 937
-        },
-        {
-            "loss": 0.00012072129175066948,
-            "reward_mean": 0.9497510194778442,
-            "reward_var": 0.006484233774244785,
-            "avg_think_len": 20.0802001953125,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9865038990974426,
-            "step": 938
-        },
-        {
-            "loss": 7.340358570218086e-05,
-            "reward_mean": 0.9260127544403076,
-            "reward_var": 0.0154604846611619,
-            "avg_think_len": 20.037328720092773,
-            "reward_format_mean": 0.9997735619544983,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9855273962020874,
-            "step": 939
-        },
-        {
-            "loss": 0.00010722456499934196,
-            "reward_mean": 0.8414551019668579,
-            "reward_var": 0.0252382829785347,
-            "avg_think_len": 19.94192123413086,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9593750238418579,
-            "reward_correctness_mean": 0.453125,
-            "reward_response_efficiency_mean": 0.9533203840255737,
-            "step": 940
-        },
-        {
-            "loss": -6.9141387939453125e-06,
-            "reward_mean": 0.8711571097373962,
-            "reward_var": 0.010548144578933716,
-            "avg_think_len": 19.966426849365234,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9690033793449402,
-            "step": 941
-        },
-        {
-            "loss": 6.639491766691208e-05,
-            "reward_mean": 0.9052495956420898,
-            "reward_var": 0.011378446593880653,
-            "avg_think_len": 20.23850440979004,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.9647483229637146,
-            "step": 942
-        },
-        {
-            "loss": 6.810761988162994e-05,
-            "reward_mean": 0.9012370705604553,
-            "reward_var": 0.011309163644909859,
-            "avg_think_len": 20.099367141723633,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9799481630325317,
-            "step": 943
-        },
-        {
-            "loss": 2.0858366042375565e-05,
-            "reward_mean": 0.9381142854690552,
-            "reward_var": 0.009512949734926224,
-            "avg_think_len": 20.001625061035156,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9868324995040894,
-            "step": 944
-        },
-        {
-            "loss": 7.576402276754379e-05,
-            "reward_mean": 0.889055073261261,
-            "reward_var": 0.009573102928698063,
-            "avg_think_len": 20.115522384643555,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9780951738357544,
-            "step": 945
-        },
-        {
-            "loss": 7.733656093478203e-05,
-            "reward_mean": 0.9038825035095215,
-            "reward_var": 0.007079161703586578,
-            "avg_think_len": 20.010114669799805,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9905297756195068,
-            "step": 946
-        },
-        {
-            "loss": 9.545590728521347e-05,
-            "reward_mean": 0.905026912689209,
-            "reward_var": 0.012278968468308449,
-            "avg_think_len": 20.1237735748291,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9638578295707703,
-            "step": 947
-        },
-        {
-            "loss": 7.028714753687382e-05,
-            "reward_mean": 0.8823882341384888,
-            "reward_var": 0.015080919489264488,
-            "avg_think_len": 20.136878967285156,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9983029365539551,
-            "step": 948
-        },
-        {
-            "loss": 9.870645590126514e-05,
-            "reward_mean": 0.9212931394577026,
-            "reward_var": 0.006946786772459745,
-            "avg_think_len": 20.00238609313965,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9976726770401001,
-            "step": 949
-        },
-        {
-            "loss": 6.723118713125587e-05,
-            "reward_mean": 0.9511741399765015,
-            "reward_var": 0.003783829975873232,
-            "avg_think_len": 20.129600524902344,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.9765716195106506,
-            "step": 950
-        },
-        {
-            "loss": 4.837522283196449e-05,
-            "reward_mean": 0.9253641963005066,
-            "reward_var": 0.009086777456104755,
-            "avg_think_len": 19.920387268066406,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9983320236206055,
-            "step": 951
-        },
-        {
-            "loss": 0.0001919666538015008,
-            "reward_mean": 0.8879696726799011,
-            "reward_var": 0.012302388437092304,
-            "avg_think_len": 20.119186401367188,
-            "reward_format_mean": 0.9998483061790466,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.8645304441452026,
-            "step": 952
-        },
-        {
-            "loss": 0.00012731854803860188,
-            "reward_mean": 0.9372075200080872,
-            "reward_var": 0.005856208968907595,
-            "avg_think_len": 20.058712005615234,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9988300800323486,
-            "step": 953
-        },
-        {
-            "loss": 5.903292912989855e-05,
-            "reward_mean": 0.871398389339447,
-            "reward_var": 0.01912507228553295,
-            "avg_think_len": 20.083681106567383,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9387186169624329,
-            "step": 954
-        },
-        {
-            "loss": 7.792096585035324e-05,
-            "reward_mean": 0.9417597651481628,
-            "reward_var": 0.0065425545908510685,
-            "avg_think_len": 20.189865112304688,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9857887029647827,
-            "step": 955
-        },
-        {
-            "loss": 3.989739343523979e-05,
-            "reward_mean": 0.9131346940994263,
-            "reward_var": 0.0132317328825593,
-            "avg_think_len": 20.11878204345703,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9806638956069946,
-            "step": 956
-        },
-        {
-            "loss": 6.795767694711685e-05,
-            "reward_mean": 0.9227452278137207,
-            "reward_var": 0.010287587530910969,
-            "avg_think_len": 20.17372703552246,
-            "reward_format_mean": 0.9992897510528564,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9416912794113159,
-            "step": 957
-        },
-        {
-            "loss": 9.592948481440544e-05,
-            "reward_mean": 0.9077277779579163,
-            "reward_var": 0.017687970772385597,
-            "avg_think_len": 20.130632400512695,
-            "reward_format_mean": 0.9998282790184021,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9435828328132629,
-            "step": 958
-        },
-        {
-            "loss": 7.429136894643307e-05,
-            "reward_mean": 0.9086214303970337,
-            "reward_var": 0.017647331580519676,
-            "avg_think_len": 20.090198516845703,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9782358407974243,
-            "step": 959
-        },
-        {
-            "loss": 1.3266457244753838e-05,
-            "reward_mean": 0.923674464225769,
-            "reward_var": 0.003905822755768895,
-            "avg_think_len": 19.830230712890625,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9978227615356445,
-            "step": 960
-        },
-        {
-            "loss": 7.383822230622172e-05,
-            "reward_mean": 0.8982182741165161,
-            "reward_var": 0.008854720741510391,
-            "avg_think_len": 20.13201141357422,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.987500011920929,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9803732633590698,
-            "step": 961
-        },
-        {
-            "loss": 3.843102604150772e-05,
-            "reward_mean": 0.9172981381416321,
-            "reward_var": 0.008775007911026478,
-            "avg_think_len": 20.021282196044922,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9973176121711731,
-            "step": 962
-        },
-        {
-            "loss": 4.645209992304444e-05,
-            "reward_mean": 0.9090188145637512,
-            "reward_var": 0.0187884122133255,
-            "avg_think_len": 19.996395111083984,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9485753774642944,
-            "step": 963
-        },
-        {
-            "loss": 5.137873813509941e-05,
-            "reward_mean": 0.8614920377731323,
-            "reward_var": 0.022025398910045624,
-            "avg_think_len": 19.90050506591797,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.8834680318832397,
-            "step": 964
-        },
-        {
-            "loss": 0.00013500871136784554,
-            "reward_mean": 0.9409093856811523,
-            "reward_var": 0.006548662669956684,
-            "avg_think_len": 20.083919525146484,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9980121850967407,
-            "step": 965
-        },
-        {
-            "loss": 2.9084738343954086e-05,
-            "reward_mean": 0.9039049744606018,
-            "reward_var": 0.02635888196527958,
-            "avg_think_len": 20.044509887695312,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9437450766563416,
-            "step": 966
-        },
-        {
-            "loss": 8.228514343500137e-05,
-            "reward_mean": 0.8117830157279968,
-            "reward_var": 0.03687549754977226,
-            "avg_think_len": 20.047256469726562,
-            "reward_format_mean": 0.9998263716697693,
-            "reward_time_mean": 0.903124988079071,
-            "reward_correctness_mean": 0.4375,
-            "reward_response_efficiency_mean": 0.9066804647445679,
-            "step": 967
-        },
-        {
-            "loss": 5.9450045228004456e-05,
-            "reward_mean": 0.86228346824646,
-            "reward_var": 0.013519584201276302,
-            "avg_think_len": 20.142324447631836,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.9491336941719055,
-            "step": 968
-        },
-        {
-            "loss": 2.208910882472992e-05,
-            "reward_mean": 0.848756730556488,
-            "reward_var": 0.017975840717554092,
-            "avg_think_len": 19.94896697998047,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.9419019818305969,
-            "step": 969
-        },
-        {
-            "loss": 0.00014603241288568825,
-            "reward_mean": 0.8747209906578064,
-            "reward_var": 0.026212623342871666,
-            "avg_think_len": 20.092992782592773,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9520085453987122,
-            "step": 970
-        },
-        {
-            "loss": 6.303656846284866e-05,
-            "reward_mean": 0.8653163313865662,
-            "reward_var": 0.029855523258447647,
-            "avg_think_len": 20.11986541748047,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9300152659416199,
-            "step": 971
-        },
-        {
-            "loss": 8.088909089565277e-05,
-            "reward_mean": 0.8580392003059387,
-            "reward_var": 0.019369402900338173,
-            "avg_think_len": 20.074474334716797,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.934374988079071,
-            "reward_correctness_mean": 0.515625,
-            "reward_response_efficiency_mean": 0.9821566939353943,
-            "step": 972
-        },
-        {
-            "loss": 8.040782995522022e-05,
-            "reward_mean": 0.8540419340133667,
-            "reward_var": 0.03601331263780594,
-            "avg_think_len": 20.07175636291504,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.8692927956581116,
-            "step": 973
-        },
-        {
-            "loss": 9.779958054423332e-05,
-            "reward_mean": 0.9761411547660828,
-            "reward_var": 0.011710097081959248,
-            "avg_think_len": 20.068401336669922,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.9375,
-            "reward_response_efficiency_mean": 0.9826894998550415,
-            "step": 974
-        },
-        {
-            "loss": 7.439858745783567e-05,
-            "reward_mean": 0.919464647769928,
-            "reward_var": 0.009605820290744305,
-            "avg_think_len": 20.081642150878906,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9903584718704224,
-            "step": 975
-        },
-        {
-            "loss": 7.148366421461105e-05,
-            "reward_mean": 0.9245588779449463,
-            "reward_var": 0.011349333450198174,
-            "avg_think_len": 20.016971588134766,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9951104521751404,
-            "step": 976
-        },
-        {
-            "loss": 2.1832995116710663e-05,
-            "reward_mean": 0.8733659386634827,
-            "reward_var": 0.008589551784098148,
-            "avg_think_len": 19.94736671447754,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9184634685516357,
-            "step": 977
-        },
-        {
-            "loss": 6.638932973146439e-05,
-            "reward_mean": 0.9514537453651428,
-            "reward_var": 0.01655902899801731,
-            "avg_think_len": 20.090553283691406,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.875,
-            "reward_response_efficiency_mean": 0.9620649814605713,
-            "step": 978
-        },
-        {
-            "loss": 5.0537579227238894e-05,
-            "reward_mean": 0.850516140460968,
-            "reward_var": 0.026491360738873482,
-            "avg_think_len": 20.02608299255371,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.8999999761581421,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.92393958568573,
-            "step": 979
-        },
-        {
-            "loss": 6.699562072753906e-05,
-            "reward_mean": 0.9468706250190735,
-            "reward_var": 0.009871155954897404,
-            "avg_think_len": 20.15345001220703,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9906074404716492,
-            "step": 980
-        },
-        {
-            "loss": 8.421950042247772e-05,
-            "reward_mean": 0.8965283632278442,
-            "reward_var": 0.010682584717869759,
-            "avg_think_len": 20.06294822692871,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9454885125160217,
-            "step": 981
-        },
-        {
-            "loss": 9.949319064617157e-05,
-            "reward_mean": 0.9541015028953552,
-            "reward_var": 0.003300492651760578,
-            "avg_think_len": 20.04775619506836,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.9726557731628418,
-            "step": 982
-        },
-        {
-            "loss": 0.00010150764137506485,
-            "reward_mean": 0.9806154370307922,
-            "reward_var": 0.004648828413337469,
-            "avg_think_len": 20.12729263305664,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.9375,
-            "reward_response_efficiency_mean": 0.9849618673324585,
-            "step": 983
-        },
-        {
-            "loss": 5.920161493122578e-05,
-            "reward_mean": 0.8511520624160767,
-            "reward_var": 0.015279300510883331,
-            "avg_think_len": 20.014522552490234,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.484375,
-            "reward_response_efficiency_mean": 0.9827330112457275,
-            "step": 984
-        },
-        {
-            "loss": 5.977647379040718e-05,
-            "reward_mean": 0.9193294048309326,
-            "reward_var": 0.015744509175419807,
-            "avg_think_len": 20.123531341552734,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9585676789283752,
-            "step": 985
-        },
-        {
-            "loss": 6.92540779709816e-05,
-            "reward_mean": 0.928242027759552,
-            "reward_var": 0.01756146177649498,
-            "avg_think_len": 20.122949600219727,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9473430514335632,
-            "step": 986
-        },
-        {
-            "loss": 2.5495304726064205e-05,
-            "reward_mean": 0.9013698101043701,
-            "reward_var": 0.017068738117814064,
-            "avg_think_len": 19.983415603637695,
-            "reward_format_mean": 0.9998094439506531,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9181694388389587,
-            "step": 987
-        },
-        {
-            "loss": 6.672055314993486e-05,
-            "reward_mean": 0.9332683682441711,
-            "reward_var": 0.01351418811827898,
-            "avg_think_len": 20.025604248046875,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9830738306045532,
-            "step": 988
-        },
-        {
-            "loss": 0.00014813058078289032,
-            "reward_mean": 0.9723920822143555,
-            "reward_var": 0.0026469076983630657,
-            "avg_think_len": 20.127225875854492,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.890625,
-            "reward_response_efficiency_mean": 0.9989433884620667,
-            "step": 989
-        },
-        {
-            "loss": 7.571722380816936e-05,
-            "reward_mean": 0.8941624760627747,
-            "reward_var": 0.005165648180991411,
-            "avg_think_len": 20.116697311401367,
-            "reward_format_mean": 0.9994212985038757,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9991037249565125,
-            "step": 990
-        },
-        {
-            "loss": 7.751071825623512e-05,
-            "reward_mean": 0.9781160354614258,
-            "reward_var": 0.004842731636017561,
-            "avg_think_len": 19.97796630859375,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.921875,
-            "reward_response_efficiency_mean": 0.9905893206596375,
-            "step": 991
-        },
-        {
-            "loss": 7.502594962716103e-05,
-            "reward_mean": 0.9327291250228882,
-            "reward_var": 0.007009811699390411,
-            "avg_think_len": 20.056047439575195,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9184165596961975,
-            "step": 992
-        },
-        {
-            "loss": 6.490555824711919e-05,
-            "reward_mean": 0.9643915295600891,
-            "reward_var": 0.003766875946894288,
-            "avg_think_len": 20.042009353637695,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.859375,
-            "reward_response_efficiency_mean": 0.9981910586357117,
-            "step": 993
-        },
-        {
-            "loss": 0.00011371425352990627,
-            "reward_mean": 0.9645963907241821,
-            "reward_var": 0.00512800645083189,
-            "avg_think_len": 20.141143798828125,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.90625,
-            "reward_response_efficiency_mean": 0.9521353840827942,
-            "step": 994
-        },
-        {
-            "loss": 6.479001604020596e-05,
-            "reward_mean": 0.9629325866699219,
-            "reward_var": 0.01123793888837099,
-            "avg_think_len": 20.000354766845703,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.890625,
-            "reward_response_efficiency_mean": 0.9767301678657532,
-            "step": 995
-        },
-        {
-            "loss": 8.647236973047256e-05,
-            "reward_mean": 0.9175949096679688,
-            "reward_var": 0.004049599636346102,
-            "avg_think_len": 20.0903377532959,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9985045194625854,
-            "step": 996
-        },
-        {
-            "loss": 5.822116509079933e-05,
-            "reward_mean": 0.8075111508369446,
-            "reward_var": 0.010726538486778736,
-            "avg_think_len": 20.188783645629883,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.981249988079071,
-            "reward_correctness_mean": 0.28125,
-            "reward_response_efficiency_mean": 0.967544674873352,
-            "step": 997
-        },
-        {
-            "loss": 5.7516153901815414e-05,
-            "reward_mean": 0.9448530673980713,
-            "reward_var": 0.0036316639743745327,
-            "avg_think_len": 20.08601188659668,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9981623888015747,
-            "step": 998
-        },
-        {
-            "loss": 9.930366650223732e-05,
-            "reward_mean": 0.909821093082428,
-            "reward_var": 0.007405904121696949,
-            "avg_think_len": 20.19225311279297,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9986593723297119,
-            "step": 999
-        },
-        {
-            "loss": 9.206868708133698e-05,
-            "reward_mean": 0.8786209225654602,
-            "reward_var": 0.022705981507897377,
-            "avg_think_len": 20.085649490356445,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9593750238418579,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9457335472106934,
-            "step": 1000
-        },
-        {
-            "loss": -6.742635741829872e-05,
-            "reward_mean": 0.973186731338501,
-            "reward_var": 0.00362033792771399,
-            "avg_think_len": 19.989835739135742,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.90625,
-            "reward_response_efficiency_mean": 0.9864970445632935,
-            "step": 1001
-        },
-        {
-            "loss": 7.259962148964405e-05,
-            "reward_mean": 0.8778827786445618,
-            "reward_var": 0.022815540432929993,
-            "avg_think_len": 20.03228759765625,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.965624988079071,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9365312457084656,
-            "step": 1002
-        },
-        {
-            "loss": 0.0001038314076140523,
-            "reward_mean": 0.8960766792297363,
-            "reward_var": 0.012229128740727901,
-            "avg_think_len": 20.073429107666016,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.974931538105011,
-            "step": 1003
-        },
-        {
-            "loss": 0.00010826729703694582,
-            "reward_mean": 0.8727704882621765,
-            "reward_var": 0.006490251049399376,
-            "avg_think_len": 20.016218185424805,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9598318934440613,
-            "step": 1004
-        },
-        {
-            "loss": 1.0382849723100662e-05,
-            "reward_mean": 0.8532410264015198,
-            "reward_var": 0.017555871978402138,
-            "avg_think_len": 20.112058639526367,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.9754639267921448,
-            "step": 1005
-        },
-        {
-            "loss": 5.7544559240341187e-05,
-            "reward_mean": 0.9995403289794922,
-            "reward_var": 3.3887818062794395e-07,
-            "avg_think_len": 20.194429397583008,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 1.0,
-            "reward_response_efficiency_mean": 0.9981614351272583,
-            "step": 1006
-        },
-        {
-            "loss": 6.091780960559845e-05,
-            "reward_mean": 0.874778151512146,
-            "reward_var": 0.01786975748836994,
-            "avg_think_len": 20.149465560913086,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9522376656532288,
-            "step": 1007
-        },
-        {
-            "loss": 0.0001872330904006958,
-            "reward_mean": 0.9438742399215698,
-            "reward_var": 0.009096447378396988,
-            "avg_think_len": 20.07166862487793,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9629970788955688,
-            "step": 1008
-        },
-        {
-            "loss": 7.869757246226072e-05,
-            "reward_mean": 0.9589192867279053,
-            "reward_var": 0.005587874911725521,
-            "avg_think_len": 20.124549865722656,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.859375,
-            "reward_response_efficiency_mean": 0.9763016700744629,
-            "step": 1009
-        },
-        {
-            "loss": 3.425450995564461e-05,
-            "reward_mean": 0.8994784355163574,
-            "reward_var": 0.008567268028855324,
-            "avg_think_len": 20.0643310546875,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9906250238418579,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.9510385990142822,
-            "step": 1010
-        },
-        {
-            "loss": 3.691950405482203e-05,
-            "reward_mean": 0.8932908177375793,
-            "reward_var": 0.006019410211592913,
-            "avg_think_len": 20.109617233276367,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9012882709503174,
-            "step": 1011
-        },
-        {
-            "loss": 6.0026999562978745e-05,
-            "reward_mean": 0.8549268245697021,
-            "reward_var": 0.013919467106461525,
-            "avg_think_len": 20.179723739624023,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.949999988079071,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.9697073101997375,
-            "step": 1012
-        },
-        {
-            "loss": 9.707128629088402e-05,
-            "reward_mean": 0.9445220828056335,
-            "reward_var": 0.006881106644868851,
-            "avg_think_len": 20.152732849121094,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9906250238418579,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9905885457992554,
-            "step": 1013
-        },
-        {
-            "loss": 4.973472096025944e-05,
-            "reward_mean": 0.8666678071022034,
-            "reward_var": 0.008095178753137589,
-            "avg_think_len": 20.004575729370117,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.46875,
-            "reward_response_efficiency_mean": 0.9979209303855896,
-            "step": 1014
-        },
-        {
-            "loss": 7.061962969601154e-05,
-            "reward_mean": 0.9293416738510132,
-            "reward_var": 0.007261770777404308,
-            "avg_think_len": 20.064592361450195,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9986166954040527,
-            "step": 1015
-        },
-        {
-            "loss": 6.368570029735565e-05,
-            "reward_mean": 0.9422193169593811,
-            "reward_var": 0.010282487608492374,
-            "avg_think_len": 20.173583984375,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9876272678375244,
-            "step": 1016
-        },
-        {
-            "loss": 3.0336203053593636e-05,
-            "reward_mean": 0.8891869783401489,
-            "reward_var": 0.01787523366510868,
-            "avg_think_len": 20.155418395996094,
-            "reward_format_mean": 0.999813973903656,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.8850586414337158,
-            "step": 1017
-        },
-        {
-            "loss": 5.1454000640660524e-05,
-            "reward_mean": 0.8631357550621033,
-            "reward_var": 0.015106471255421638,
-            "avg_think_len": 20.074413299560547,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.5,
-            "reward_response_efficiency_mean": 0.9681681394577026,
-            "step": 1018
-        },
-        {
-            "loss": 8.609145879745483e-05,
-            "reward_mean": 0.9390507936477661,
-            "reward_var": 0.008772043511271477,
-            "avg_think_len": 20.000749588012695,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9749531745910645,
-            "step": 1019
-        },
-        {
-            "loss": 8.879322558641434e-05,
-            "reward_mean": 0.87073814868927,
-            "reward_var": 0.0076975212432444096,
-            "avg_think_len": 20.135826110839844,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.484375,
-            "reward_response_efficiency_mean": 0.998577356338501,
-            "step": 1020
-        },
-        {
-            "loss": 0.00011498574167490005,
-            "reward_mean": 0.8736653923988342,
-            "reward_var": 0.018335163593292236,
-            "avg_think_len": 20.167282104492188,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9477866888046265,
-            "step": 1021
-        },
-        {
-            "loss": 5.8639561757445335e-05,
-            "reward_mean": 0.8822247385978699,
-            "reward_var": 0.012539287097752094,
-            "avg_think_len": 20.10245132446289,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.987500011920929,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9788989424705505,
-            "step": 1022
-        },
-        {
-            "loss": 6.120721809566021e-05,
-            "reward_mean": 0.8958592414855957,
-            "reward_var": 0.023873141035437584,
-            "avg_think_len": 20.065811157226562,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9428123235702515,
-            "step": 1023
-        },
-        {
-            "loss": 5.298620089888573e-05,
-            "reward_mean": 0.8520433306694031,
-            "reward_var": 0.03826681524515152,
-            "avg_think_len": 19.919979095458984,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.90625,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.845673680305481,
-            "step": 1024
-        },
-        {
-            "loss": 0.0001330472296103835,
-            "reward_mean": 0.8773822784423828,
-            "reward_var": 0.01811763271689415,
-            "avg_think_len": 19.938688278198242,
-            "reward_format_mean": 0.9988839626312256,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9012702107429504,
-            "step": 1025
-        },
-        {
-            "loss": 6.857048720121384e-05,
-            "reward_mean": 0.8932690620422363,
-            "reward_var": 0.011669398285448551,
-            "avg_think_len": 20.0119686126709,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9824510216712952,
-            "step": 1026
-        },
-        {
-            "loss": 0.00019719405099749565,
-            "reward_mean": 0.9349023103713989,
-            "reward_var": 0.003909182269126177,
-            "avg_think_len": 20.056777954101562,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9739842414855957,
-            "step": 1027
-        },
-        {
-            "loss": 6.241397932171822e-05,
-            "reward_mean": 0.8395495414733887,
-            "reward_var": 0.03979543223977089,
-            "avg_think_len": 19.88115882873535,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.890625,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9050732851028442,
-            "step": 1028
-        },
-        {
-            "loss": 5.99399209022522e-05,
-            "reward_mean": 0.8903700709342957,
-            "reward_var": 0.00998410303145647,
-            "avg_think_len": 20.151683807373047,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9833550453186035,
-            "step": 1029
-        },
-        {
-            "loss": 1.8228660337626934e-05,
-            "reward_mean": 0.8980523347854614,
-            "reward_var": 0.011166663840413094,
-            "avg_think_len": 20.007354736328125,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9984594583511353,
-            "step": 1030
-        },
-        {
-            "loss": 7.20643438398838e-05,
-            "reward_mean": 0.956523597240448,
-            "reward_var": 0.007273167837411165,
-            "avg_think_len": 20.119386672973633,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.859375,
-            "reward_response_efficiency_mean": 0.9667196273803711,
-            "step": 1031
-        },
-        {
-            "loss": 3.993278369307518e-06,
-            "reward_mean": 0.8569115400314331,
-            "reward_var": 0.011670888401567936,
-            "avg_think_len": 20.21106719970703,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.484375,
-            "reward_response_efficiency_mean": 0.9588961005210876,
-            "step": 1032
-        },
-        {
-            "loss": 8.837785571813583e-05,
-            "reward_mean": 0.9217929244041443,
-            "reward_var": 0.004766055382788181,
-            "avg_think_len": 20.121231079101562,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9840463995933533,
-            "step": 1033
-        },
-        {
-            "loss": 0.0001078255008906126,
-            "reward_mean": 0.8903660774230957,
-            "reward_var": 0.027603117749094963,
-            "avg_think_len": 20.084278106689453,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9520896077156067,
-            "step": 1034
-        },
-        {
-            "loss": 0.0001148078590631485,
-            "reward_mean": 0.9192661643028259,
-            "reward_var": 0.006777808535844088,
-            "avg_think_len": 19.937496185302734,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9583150148391724,
-            "step": 1035
-        },
-        {
-            "loss": 9.711843449622393e-05,
-            "reward_mean": 0.9156438112258911,
-            "reward_var": 0.019786417484283447,
-            "avg_think_len": 20.01498031616211,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9468749761581421,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9344503283500671,
-            "step": 1036
-        },
-        {
-            "loss": 7.28280283510685e-05,
-            "reward_mean": 0.9597249627113342,
-            "reward_var": 0.012979448772966862,
-            "avg_think_len": 20.094589233398438,
-            "reward_format_mean": 0.9998579621315002,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.90625,
-            "reward_response_efficiency_mean": 0.9640420079231262,
-            "step": 1037
-        },
-        {
-            "loss": 7.990930316736922e-05,
-            "reward_mean": 0.8824003338813782,
-            "reward_var": 0.007822108455002308,
-            "avg_think_len": 20.158552169799805,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9983512759208679,
-            "step": 1038
-        },
-        {
-            "loss": 9.645288810133934e-05,
-            "reward_mean": 0.9331419467926025,
-            "reward_var": 0.010164549574255943,
-            "avg_think_len": 20.104867935180664,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9981927871704102,
-            "step": 1039
-        },
-        {
-            "loss": 6.724311970174313e-05,
-            "reward_mean": 0.8873449563980103,
-            "reward_var": 0.008069973438978195,
-            "avg_think_len": 20.013566970825195,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9868797063827515,
-            "step": 1040
-        },
-        {
-            "loss": 9.29730013012886e-05,
-            "reward_mean": 0.9189478754997253,
-            "reward_var": 0.0026609981432557106,
-            "avg_think_len": 20.10206413269043,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9882913827896118,
-            "step": 1041
-        },
-        {
-            "loss": 0.00011680068564601243,
-            "reward_mean": 0.8912940621376038,
-            "reward_var": 0.013171905651688576,
-            "avg_think_len": 20.001203536987305,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9714263677597046,
-            "step": 1042
-        },
-        {
-            "loss": 7.191486656665802e-05,
-            "reward_mean": 0.8674586415290833,
-            "reward_var": 0.004847819451242685,
-            "avg_think_len": 19.89548683166504,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.484375,
-            "reward_response_efficiency_mean": 0.9885845184326172,
-            "step": 1043
-        },
-        {
-            "loss": 6.642984226346016e-05,
-            "reward_mean": 0.9652840495109558,
-            "reward_var": 0.00694854324683547,
-            "avg_think_len": 20.00360870361328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.875,
-            "reward_response_efficiency_mean": 0.9861364364624023,
-            "step": 1044
-        },
-        {
-            "loss": 0.00013408425729721785,
-            "reward_mean": 0.9606115221977234,
-            "reward_var": 0.005043823271989822,
-            "avg_think_len": 20.128883361816406,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.998695969581604,
-            "step": 1045
-        },
-        {
-            "loss": 5.518784746527672e-05,
-            "reward_mean": 0.9057198762893677,
-            "reward_var": 0.0183790884912014,
-            "avg_think_len": 20.155012130737305,
-            "reward_format_mean": 0.9994419813156128,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9359373450279236,
-            "step": 1046
-        },
-        {
-            "loss": 1.4914781786501408e-05,
-            "reward_mean": 0.9446878433227539,
-            "reward_var": 0.005303243640810251,
-            "avg_think_len": 20.05095863342285,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9975014925003052,
-            "step": 1047
-        },
-        {
-            "loss": -1.1090392945334315e-05,
-            "reward_mean": 0.9067456722259521,
-            "reward_var": 0.013651715591549873,
-            "avg_think_len": 20.187301635742188,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9863576889038086,
-            "step": 1048
-        },
-        {
-            "loss": 9.652925655245781e-05,
-            "reward_mean": 0.9034680724143982,
-            "reward_var": 0.013444880954921246,
-            "avg_think_len": 20.150203704833984,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9576224088668823,
-            "step": 1049
-        },
-        {
-            "loss": 8.664233610033989e-05,
-            "reward_mean": 0.9350509643554688,
-            "reward_var": 0.01975957117974758,
-            "avg_think_len": 19.896745681762695,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.9433289766311646,
-            "step": 1050
-        },
-        {
-            "loss": -5.878973752260208e-06,
-            "reward_mean": 0.9138145446777344,
-            "reward_var": 0.00939953327178955,
-            "avg_think_len": 20.019638061523438,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.967758059501648,
-            "step": 1051
-        },
-        {
-            "loss": 4.466119207791053e-05,
-            "reward_mean": 0.8799877166748047,
-            "reward_var": 0.011859310790896416,
-            "avg_think_len": 20.122970581054688,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.9730760455131531,
-            "step": 1052
-        },
-        {
-            "loss": 4.864227958023548e-05,
-            "reward_mean": 0.9215013980865479,
-            "reward_var": 0.008623761124908924,
-            "avg_think_len": 20.12749481201172,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9985055327415466,
-            "step": 1053
-        },
-        {
-            "loss": 0.00013815797865390778,
-            "reward_mean": 0.8942774534225464,
-            "reward_var": 0.004228018224239349,
-            "avg_think_len": 19.968114852905273,
-            "reward_format_mean": 0.9988839626312256,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.990725576877594,
-            "step": 1054
-        },
-        {
-            "loss": 0.00013297656551003456,
-            "reward_mean": 0.9449135661125183,
-            "reward_var": 0.012820211239159107,
-            "avg_think_len": 20.221525192260742,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.9827790260314941,
-            "step": 1055
-        },
-        {
-            "loss": 5.710613913834095e-05,
-            "reward_mean": 0.8905631303787231,
-            "reward_var": 0.011971830390393734,
-            "avg_think_len": 20.059606552124023,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9685025215148926,
-            "step": 1056
-        },
-        {
-            "loss": 7.078051567077637e-08,
-            "reward_mean": 0.9846125841140747,
-            "reward_var": 0.0032756009604781866,
-            "avg_think_len": 20.041845321655273,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.953125,
-            "reward_response_efficiency_mean": 0.9853257536888123,
-            "step": 1057
-        },
-        {
-            "loss": 9.685999248176813e-05,
-            "reward_mean": 0.9211685657501221,
-            "reward_var": 0.011201734654605389,
-            "avg_think_len": 19.948131561279297,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.987500011920929,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9784241318702698,
-            "step": 1058
-        },
-        {
-            "loss": 5.505490116775036e-05,
-            "reward_mean": 0.896432638168335,
-            "reward_var": 0.0071395523846149445,
-            "avg_think_len": 20.14211082458496,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.9763554930686951,
-            "step": 1059
-        },
-        {
-            "loss": 5.522719584405422e-05,
-            "reward_mean": 0.927122175693512,
-            "reward_var": 0.011788052506744862,
-            "avg_think_len": 20.08714485168457,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9781249761581421,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9491137862205505,
-            "step": 1060
-        },
-        {
-            "loss": 8.38235137052834e-05,
-            "reward_mean": 0.9156026840209961,
-            "reward_var": 0.006940394639968872,
-            "avg_think_len": 20.169214248657227,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9905357360839844,
-            "step": 1061
-        },
-        {
-            "loss": 8.429400622844696e-05,
-            "reward_mean": 0.9118078947067261,
-            "reward_var": 0.012665778398513794,
-            "avg_think_len": 20.178861618041992,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.9909815192222595,
-            "step": 1062
-        },
-        {
-            "loss": 4.6111177653074265e-05,
-            "reward_mean": 0.8979832530021667,
-            "reward_var": 0.014138195663690567,
-            "avg_think_len": 19.978986740112305,
-            "reward_format_mean": 0.9986979365348816,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.625,
-            "reward_response_efficiency_mean": 0.983860194683075,
-            "step": 1063
-        },
-        {
-            "loss": 4.212046042084694e-05,
-            "reward_mean": 0.9859179258346558,
-            "reward_var": 0.0029019408393651247,
-            "avg_think_len": 20.19460678100586,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.953125,
-            "reward_response_efficiency_mean": 0.990546703338623,
-            "step": 1064
-        },
-        {
-            "loss": 5.208933725953102e-05,
-            "reward_mean": 0.8863972425460815,
-            "reward_var": 0.00655408576130867,
-            "avg_think_len": 20.051366806030273,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.9987142086029053,
-            "step": 1065
-        },
-        {
-            "loss": 3.592902794480324e-05,
-            "reward_mean": 0.8252991437911987,
-            "reward_var": 0.01735294610261917,
-            "avg_think_len": 20.114246368408203,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.375,
-            "reward_response_efficiency_mean": 0.9730712175369263,
-            "step": 1066
-        },
-        {
-            "loss": 0.0001028524711728096,
-            "reward_mean": 0.8687523603439331,
-            "reward_var": 0.01508943922817707,
-            "avg_think_len": 20.085256576538086,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.53125,
-            "reward_response_efficiency_mean": 0.9593843221664429,
-            "step": 1067
-        },
-        {
-            "loss": 5.968223558738828e-05,
-            "reward_mean": 0.9158999919891357,
-            "reward_var": 0.011639446020126343,
-            "avg_think_len": 19.996170043945312,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.960474967956543,
-            "step": 1068
-        },
-        {
-            "loss": 2.442067489027977e-05,
-            "reward_mean": 0.9313775897026062,
-            "reward_var": 0.011818774044513702,
-            "avg_think_len": 20.069355010986328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9755104780197144,
-            "step": 1069
-        },
-        {
-            "loss": 6.746011786162853e-05,
-            "reward_mean": 0.9956735372543335,
-            "reward_var": 0.0009780998807400465,
-            "avg_think_len": 20.05174446105957,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.984375,
-            "reward_response_efficiency_mean": 0.9983192682266235,
-            "step": 1070
-        },
-        {
-            "loss": 6.78338692523539e-05,
-            "reward_mean": 0.9118170738220215,
-            "reward_var": 0.008472174406051636,
-            "avg_think_len": 20.075807571411133,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.991018533706665,
-            "step": 1071
-        },
-        {
-            "loss": 6.567081436514854e-05,
-            "reward_mean": 0.8588194251060486,
-            "reward_var": 0.011442205868661404,
-            "avg_think_len": 20.184980392456055,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.4375,
-            "reward_response_efficiency_mean": 0.9977777600288391,
-            "step": 1072
-        },
-        {
-            "loss": 6.706826388835907e-05,
-            "reward_mean": 0.9429631233215332,
-            "reward_var": 0.012618438340723515,
-            "avg_think_len": 20.161544799804688,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.8125,
-            "reward_response_efficiency_mean": 0.974977433681488,
-            "step": 1073
-        },
-        {
-            "loss": 2.622627653181553e-05,
-            "reward_mean": 0.9110780954360962,
-            "reward_var": 0.009625871665775776,
-            "avg_think_len": 20.238327026367188,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.9880622625350952,
-            "step": 1074
-        },
-        {
-            "loss": 7.137656211853027e-05,
-            "reward_mean": 0.8969138264656067,
-            "reward_var": 0.010999594815075397,
-            "avg_think_len": 20.01692771911621,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9782801866531372,
-            "step": 1075
-        },
-        {
-            "loss": 7.615890353918076e-05,
-            "reward_mean": 0.9404295682907104,
-            "reward_var": 0.005609145853668451,
-            "avg_think_len": 20.112375259399414,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.996874988079071,
-            "reward_correctness_mean": 0.78125,
-            "reward_response_efficiency_mean": 0.9835931658744812,
-            "step": 1076
-        },
-        {
-            "loss": 3.3284304663538933e-06,
-            "reward_mean": 0.8935518860816956,
-            "reward_var": 0.022394046187400818,
-            "avg_think_len": 20.1846923828125,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9375,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9492074847221375,
-            "step": 1077
-        },
-        {
-            "loss": 6.697035860270262e-05,
-            "reward_mean": 0.8510587215423584,
-            "reward_var": 0.007258884143084288,
-            "avg_think_len": 20.039325714111328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.40625,
-            "reward_response_efficiency_mean": 0.9979851841926575,
-            "step": 1078
-        },
-        {
-            "loss": 4.340754821896553e-05,
-            "reward_mean": 0.9215344190597534,
-            "reward_var": 0.006981523707509041,
-            "avg_think_len": 20.142589569091797,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.6875,
-            "reward_response_efficiency_mean": 0.9986376762390137,
-            "step": 1079
-        },
-        {
-            "loss": 7.536751218140125e-05,
-            "reward_mean": 0.924257218837738,
-            "reward_var": 0.015043997205793858,
-            "avg_think_len": 20.076505661010742,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9782789349555969,
-            "step": 1080
-        },
-        {
-            "loss": 9.675580076873302e-05,
-            "reward_mean": 0.9018901586532593,
-            "reward_var": 0.006841922644525766,
-            "avg_think_len": 19.97674560546875,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9981853365898132,
-            "step": 1081
-        },
-        {
-            "loss": 9.6077099442482e-05,
-            "reward_mean": 0.936855137348175,
-            "reward_var": 0.007503790780901909,
-            "avg_think_len": 19.944652557373047,
-            "reward_format_mean": 0.9989583492279053,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9984621405601501,
-            "step": 1082
-        },
-        {
-            "loss": 6.626127287745476e-05,
-            "reward_mean": 0.8244688510894775,
-            "reward_var": 0.010914594866335392,
-            "avg_think_len": 20.164796829223633,
-            "reward_format_mean": 0.9998301863670349,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.390625,
-            "reward_response_efficiency_mean": 0.9074203372001648,
-            "step": 1083
-        },
-        {
-            "loss": 0.00010631527402438223,
-            "reward_mean": 0.9007306098937988,
-            "reward_var": 0.01361012365669012,
-            "avg_think_len": 20.07556915283203,
-            "reward_format_mean": 0.9997479915618896,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9781743288040161,
-            "step": 1084
-        },
-        {
-            "loss": 5.037570372223854e-05,
-            "reward_mean": 0.9268187284469604,
-            "reward_var": 0.008625069633126259,
-            "avg_think_len": 20.004858016967773,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9729001522064209,
-            "step": 1085
-        },
-        {
-            "loss": 5.737552419304848e-05,
-            "reward_mean": 0.9324633479118347,
-            "reward_var": 0.02030906453728676,
-            "avg_think_len": 19.875057220458984,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.890625,
-            "reward_response_efficiency_mean": 0.8704783320426941,
-            "step": 1086
-        },
-        {
-            "loss": 9.732041507959366e-05,
-            "reward_mean": 0.8980600237846375,
-            "reward_var": 0.004743323195725679,
-            "avg_think_len": 20.1710262298584,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9984901547431946,
-            "step": 1087
-        },
-        {
-            "loss": 7.750885561108589e-05,
-            "reward_mean": 0.8307944536209106,
-            "reward_var": 0.013675727881491184,
-            "avg_think_len": 19.96842384338379,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.453125,
-            "reward_response_efficiency_mean": 0.9013028144836426,
-            "step": 1088
-        },
-        {
-            "loss": 6.820494309067726e-05,
-            "reward_mean": 0.9062903523445129,
-            "reward_var": 0.015440428629517555,
-            "avg_think_len": 19.926654815673828,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9781249761581421,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.9439114332199097,
-            "step": 1089
-        },
-        {
-            "loss": 7.186317816376686e-05,
-            "reward_mean": 0.9024631977081299,
-            "reward_var": 0.018436934798955917,
-            "avg_think_len": 19.927745819091797,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9848527312278748,
-            "step": 1090
-        },
-        {
-            "loss": 5.991430953145027e-05,
-            "reward_mean": 0.9487693309783936,
-            "reward_var": 0.008501382544636726,
-            "avg_think_len": 19.986879348754883,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.796875,
-            "reward_response_efficiency_mean": 0.9982020258903503,
-            "step": 1091
-        },
-        {
-            "loss": 7.587112486362457e-05,
-            "reward_mean": 0.9281933307647705,
-            "reward_var": 0.006404022220522165,
-            "avg_think_len": 20.154884338378906,
-            "reward_format_mean": 0.9982638955116272,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9801341891288757,
-            "step": 1092
-        },
-        {
-            "loss": 6.366102024912834e-05,
-            "reward_mean": 0.9163224697113037,
-            "reward_var": 0.016915172338485718,
-            "avg_think_len": 20.117582321166992,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9309147596359253,
-            "step": 1093
-        },
-        {
-            "loss": 6.361998384818435e-05,
-            "reward_mean": 0.9164298176765442,
-            "reward_var": 0.013933326117694378,
-            "avg_think_len": 19.996944427490234,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9313443303108215,
-            "step": 1094
-        },
-        {
-            "loss": 3.8959900848567486e-05,
-            "reward_mean": 0.9158923625946045,
-            "reward_var": 0.012923058122396469,
-            "avg_think_len": 19.97260856628418,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9291946887969971,
-            "step": 1095
-        },
-        {
-            "loss": 6.536440923810005e-05,
-            "reward_mean": 0.8771044611930847,
-            "reward_var": 0.020970191806554794,
-            "avg_think_len": 19.854373931884766,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.971875011920929,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9427928924560547,
-            "step": 1096
-        },
-        {
-            "loss": 5.1849521696567535e-05,
-            "reward_mean": 0.8837690353393555,
-            "reward_var": 0.010006858967244625,
-            "avg_think_len": 20.21588134765625,
-            "reward_format_mean": 0.9998405575752258,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.578125,
-            "reward_response_efficiency_mean": 0.9571105241775513,
-            "step": 1097
-        },
-        {
-            "loss": 6.630178540945053e-05,
-            "reward_mean": 0.9179639220237732,
-            "reward_var": 0.008481341414153576,
-            "avg_think_len": 19.984878540039062,
-            "reward_format_mean": 0.9988839626312256,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9542217254638672,
-            "step": 1098
-        },
-        {
-            "loss": 0.00012659869389608502,
-            "reward_mean": 0.963996171951294,
-            "reward_var": 0.013669038191437721,
-            "avg_think_len": 20.103361129760742,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.921875,
-            "reward_response_efficiency_mean": 0.965359628200531,
-            "step": 1099
-        },
-        {
-            "loss": 4.444411024451256e-05,
-            "reward_mean": 0.9147030711174011,
-            "reward_var": 0.012544477358460426,
-            "avg_think_len": 20.069440841674805,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.703125,
-            "reward_response_efficiency_mean": 0.971312403678894,
-            "step": 1100
-        },
-        {
-            "loss": 9.952345862984657e-05,
-            "reward_mean": 0.9605699777603149,
-            "reward_var": 0.007807284127920866,
-            "avg_think_len": 20.059446334838867,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.998529851436615,
-            "step": 1101
-        },
-        {
-            "loss": 6.577512249350548e-05,
-            "reward_mean": 0.8841226100921631,
-            "reward_var": 0.006669867783784866,
-            "avg_think_len": 20.165891647338867,
-            "reward_format_mean": 0.9997438192367554,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9273715615272522,
-            "step": 1102
-        },
-        {
-            "loss": 8.555315434932709e-05,
-            "reward_mean": 0.9409540891647339,
-            "reward_var": 0.006833150051534176,
-            "avg_think_len": 19.99193000793457,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.765625,
-            "reward_response_efficiency_mean": 0.9981913566589355,
-            "step": 1103
-        },
-        {
-            "loss": 2.628238871693611e-05,
-            "reward_mean": 0.9062013030052185,
-            "reward_var": 0.009930919855833054,
-            "avg_think_len": 20.067352294921875,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.984180212020874,
-            "step": 1104
-        },
-        {
-            "loss": 5.8687757700681686e-05,
-            "reward_mean": 0.909653902053833,
-            "reward_var": 0.009622524492442608,
-            "avg_think_len": 20.20421600341797,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.65625,
-            "reward_response_efficiency_mean": 0.9979904890060425,
-            "step": 1105
-        },
-        {
-            "loss": 6.067659705877304e-05,
-            "reward_mean": 0.9138195514678955,
-            "reward_var": 0.01131872832775116,
-            "avg_think_len": 19.928564071655273,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9209031462669373,
-            "step": 1106
-        },
-        {
-            "loss": 0.00013951119035482407,
-            "reward_mean": 0.9566980600357056,
-            "reward_var": 0.0032106356229633093,
-            "avg_think_len": 20.203144073486328,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.828125,
-            "reward_response_efficiency_mean": 0.9986671805381775,
-            "step": 1107
-        },
-        {
-            "loss": 6.309570744633675e-05,
-            "reward_mean": 0.8774133920669556,
-            "reward_var": 0.013934307731688023,
-            "avg_think_len": 20.165956497192383,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.546875,
-            "reward_response_efficiency_mean": 0.9784035682678223,
-            "step": 1108
-        },
-        {
-            "loss": 7.913773879408836e-05,
-            "reward_mean": 0.901995837688446,
-            "reward_var": 0.007116986438632011,
-            "avg_think_len": 19.96878433227539,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9986086487770081,
-            "step": 1109
-        },
-        {
-            "loss": 8.176872506737709e-05,
-            "reward_mean": 0.8815920352935791,
-            "reward_var": 0.015379766002297401,
-            "avg_think_len": 20.201929092407227,
-            "reward_format_mean": 0.9998319745063782,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.640625,
-            "reward_response_efficiency_mean": 0.9171609282493591,
-            "step": 1110
-        },
-        {
-            "loss": 8.895708742784336e-05,
-            "reward_mean": 0.9132699966430664,
-            "reward_var": 0.018106067553162575,
-            "avg_think_len": 20.092435836791992,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.734375,
-            "reward_response_efficiency_mean": 0.9499548077583313,
-            "step": 1111
-        },
-        {
-            "loss": 6.361445412039757e-05,
-            "reward_mean": 0.8441603183746338,
-            "reward_var": 0.010484856553375721,
-            "avg_think_len": 19.927776336669922,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.9937499761581421,
-            "reward_correctness_mean": 0.453125,
-            "reward_response_efficiency_mean": 0.9297664165496826,
-            "step": 1112
-        },
-        {
-            "loss": 7.904693484306335e-05,
-            "reward_mean": 0.8600031137466431,
-            "reward_var": 0.008763367310166359,
-            "avg_think_len": 20.031497955322266,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.859375,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.9868872165679932,
-            "step": 1113
-        },
-        {
-            "loss": 3.318674862384796e-05,
-            "reward_mean": 0.9722337126731873,
-            "reward_var": 0.0048748222179710865,
-            "avg_think_len": 19.99652862548828,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.890625,
-            "reward_response_efficiency_mean": 0.9983097910881042,
-            "step": 1114
-        },
-        {
-            "loss": 2.9318267479538918e-05,
-            "reward_mean": 0.8939862847328186,
-            "reward_var": 0.02448812685906887,
-            "avg_think_len": 20.07280731201172,
-            "reward_format_mean": 0.9997209906578064,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.71875,
-            "reward_response_efficiency_mean": 0.9043492078781128,
-            "step": 1115
-        },
-        {
-            "loss": 8.37469706311822e-05,
-            "reward_mean": 0.9391821622848511,
-            "reward_var": 0.02490559220314026,
-            "avg_think_len": 20.066036224365234,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.953125,
-            "reward_correctness_mean": 0.84375,
-            "reward_response_efficiency_mean": 0.9598535299301147,
-            "step": 1116
-        },
-        {
-            "loss": 5.83187211304903e-05,
-            "reward_mean": 0.8491774797439575,
-            "reward_var": 0.015018519014120102,
-            "avg_think_len": 20.14603042602539,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.421875,
-            "reward_response_efficiency_mean": 0.990459680557251,
-            "step": 1117
-        },
-        {
-            "loss": 7.57463276386261e-05,
-            "reward_mean": 0.8917096853256226,
-            "reward_var": 0.016513783484697342,
-            "avg_think_len": 19.921512603759766,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.609375,
-            "reward_response_efficiency_mean": 0.9730886816978455,
-            "step": 1118
-        },
-        {
-            "loss": 5.993945524096489e-05,
-            "reward_mean": 0.9096800088882446,
-            "reward_var": 0.015198338776826859,
-            "avg_think_len": 20.13595962524414,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.671875,
-            "reward_response_efficiency_mean": 0.9824697375297546,
-            "step": 1119
-        },
-        {
-            "loss": 6.062840111553669e-05,
-            "reward_mean": 0.9209223985671997,
-            "reward_var": 0.009702391922473907,
-            "avg_think_len": 20.0801944732666,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.75,
-            "reward_response_efficiency_mean": 0.9493144750595093,
-            "step": 1120
-        },
-        {
-            "loss": 4.3173786252737045e-05,
-            "reward_mean": 0.9586657881736755,
-            "reward_var": 0.00791262835264206,
-            "avg_think_len": 19.960800170898438,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.984375,
-            "reward_correctness_mean": 0.859375,
-            "reward_response_efficiency_mean": 0.990912675857544,
-            "step": 1121
-        },
-        {
-            "loss": 3.558024764060974e-05,
-            "reward_mean": 0.881234884262085,
-            "reward_var": 0.01561199314892292,
-            "avg_think_len": 19.97524070739746,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.96875,
-            "reward_correctness_mean": 0.59375,
-            "reward_response_efficiency_mean": 0.962439775466919,
-            "step": 1122
-        },
-        {
-            "loss": -2.7467191102914512e-05,
-            "reward_mean": 0.8567469120025635,
-            "reward_var": 0.012564903125166893,
-            "avg_think_len": 20.11345863342285,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 0.921875,
-            "reward_correctness_mean": 0.5625,
-            "reward_response_efficiency_mean": 0.9426124691963196,
-            "step": 1123
-        },
-        {
-            "loss": 8.306419476866722e-05,
-            "reward_mean": 0.9731650352478027,
-            "reward_var": 0.005696384701877832,
-            "avg_think_len": 20.207298278808594,
-            "reward_format_mean": 1.0,
-            "reward_time_mean": 1.0,
-            "reward_correctness_mean": 0.90625,
-            "reward_response_efficiency_mean": 0.9864100813865662,
-            "step": 1124
-        }
-    ],
-    "global_step": 1125,
-    "epoch_idx": 0,
-    "epoch": 0.0
-}
\ No newline at end of file