[
  {
    "iteration": 0,
    "elapsed_s": 283.72707653045654,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.6677320075757578,
    "avg_questions": 0.71875,
    "avg_turns": 1.6875,
    "ppo_loss": -0.0012642779620364308,
    "value_loss": 0.9520214434180941,
    "approx_kl": 0.009626129126575376,
    "clip_frac": 0.18303571481789863,
    "entropy": 0.25815688552600996,
    "kl_penalty": 0.011675153459821428,
    "kl_per_seq": 0.011675153459821428,
    "kl_seq_max": 0.18973214285714285
  },
  {
    "iteration": 1,
    "elapsed_s": 265.6153476238251,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.6511522331834831,
    "avg_questions": 0.75,
    "avg_turns": 1.71875,
    "ppo_loss": -0.02305267165814127,
    "value_loss": 0.3821369180721896,
    "approx_kl": 0.006722210083223347,
    "clip_frac": 0.09438775533011981,
    "entropy": 0.22343750271414006,
    "kl_penalty": 0.010663713727678572,
    "kl_per_seq": 0.010663713727678572,
    "kl_seq_max": 0.14174107142857142
  },
  {
    "iteration": 2,
    "elapsed_s": 283.4707496166229,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.6145833333333333,
    "avg_questions": 0.71875,
    "avg_turns": 1.6875,
    "ppo_loss": 0.0003246570039274437,
    "value_loss": 0.5281469119446618,
    "approx_kl": 0.005086135942422386,
    "clip_frac": 0.0595238097012043,
    "entropy": 0.20671099424362183,
    "kl_penalty": 0.022153581891741072,
    "kl_per_seq": 0.022153581891741072,
    "kl_seq_max": 0.15959821428571427
  },
  {
    "iteration": 3,
    "elapsed_s": 277.0273263454437,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.8185267857142857,
    "avg_questions": 0.875,
    "avg_turns": 1.84375,
    "ppo_loss": 0.017307979986071587,
    "value_loss": 0.2391173227224499,
    "approx_kl": 0.006241402057639789,
    "clip_frac": 0.0859375,
    "entropy": 0.24670861288905144,
    "kl_penalty": 0.0052337646484375,
    "kl_per_seq": 0.0052337646484375,
    "kl_seq_max": 0.130126953125
  },
  {
    "iteration": 4,
    "elapsed_s": 286.0855031013489,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.6283482142857142,
    "avg_questions": 0.875,
    "avg_turns": 1.875,
    "ppo_loss": -0.023693807190284133,
    "value_loss": 0.23578157555311918,
    "approx_kl": 0.006290083052590489,
    "clip_frac": 0.08984375,
    "entropy": 0.2442768611945212,
    "kl_penalty": 0.013458251953125,
    "kl_per_seq": 0.013458251953125,
    "kl_seq_max": 0.13818359375
  },
  {
    "iteration": 5,
    "elapsed_s": 242.1764998435974,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.7465401785714285,
    "avg_questions": 0.78125,
    "avg_turns": 1.78125,
    "ppo_loss": -0.0412354568640391,
    "value_loss": 0.36194167503466207,
    "approx_kl": 0.02360554877668619,
    "clip_frac": 0.21354166666666666,
    "entropy": 0.28217028516034287,
    "kl_penalty": 0.044108072916666664,
    "kl_per_seq": 0.044108072916666664,
    "kl_seq_max": 0.22916666666666666
  },
  {
    "iteration": 6,
    "elapsed_s": 297.9436032772064,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.6431051587301587,
    "avg_questions": 0.84375,
    "avg_turns": 1.84375,
    "ppo_loss": -0.01692520675715059,
    "value_loss": 0.28041012212634087,
    "approx_kl": 0.006847011674835812,
    "clip_frac": 0.11197916697710752,
    "entropy": 0.25802480173297226,
    "kl_penalty": 0.01638031005859375,
    "kl_per_seq": 0.01638031005859375,
    "kl_seq_max": 0.14501953125
  },
  {
    "iteration": 7,
    "elapsed_s": 330.66188168525696,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.6246899801587302,
    "avg_questions": 0.8125,
    "avg_turns": 1.78125,
    "ppo_loss": 0.004835611267480999,
    "value_loss": 0.42471681741881184,
    "approx_kl": 0.006292660720646381,
    "clip_frac": 0.078125,
    "entropy": 0.22681239678058773,
    "kl_penalty": 0.0085601806640625,
    "kl_per_seq": 0.0085601806640625,
    "kl_seq_max": 0.14404296875
  },
  {
    "iteration": 8,
    "elapsed_s": 300.30717039108276,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.6534733495670996,
    "avg_questions": 0.78125,
    "avg_turns": 1.75,
    "ppo_loss": -0.017197133707148687,
    "value_loss": 0.3457076509616205,
    "approx_kl": 0.006416547777397292,
    "clip_frac": 0.07589285714285714,
    "entropy": 0.22316500225237437,
    "kl_penalty": 0.004743303571428571,
    "kl_per_seq": 0.004743303571428571,
    "kl_seq_max": 0.197265625
  },
  {
    "iteration": 9,
    "elapsed_s": 286.7236270904541,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.7079376352813853,
    "avg_questions": 0.78125,
    "avg_turns": 1.75,
    "ppo_loss": -0.011897004608597075,
    "value_loss": 0.411878263843911,
    "approx_kl": 0.007929061938609396,
    "clip_frac": 0.11607142857142858,
    "entropy": 0.2569970648203577,
    "kl_penalty": 0.00020926339285714285,
    "kl_per_seq": 0.00020926339285714285,
    "kl_seq_max": 0.16936383928571427
  },
  {
    "iteration": 10,
    "elapsed_s": 276.631822347641,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.8028939810189811,
    "avg_questions": 0.84375,
    "avg_turns": 1.78125,
    "ppo_loss": -0.033312247483991086,
    "value_loss": 0.2869396444875747,
    "approx_kl": 0.008976982207968831,
    "clip_frac": 0.12109375,
    "entropy": 0.25112444773549214,
    "kl_penalty": 0.01202392578125,
    "kl_per_seq": 0.01202392578125,
    "kl_seq_max": 0.203857421875
  },
  {
    "iteration": 11,
    "elapsed_s": 255.2351098060608,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.5704326923076923,
    "avg_questions": 0.78125,
    "avg_turns": 1.75,
    "ppo_loss": -0.005331723640362422,
    "value_loss": 0.42935604602098465,
    "approx_kl": 0.016193445358011458,
    "clip_frac": 0.2222222222222222,
    "entropy": 0.23103458185990652,
    "kl_penalty": 0.0,
    "kl_per_seq": 0.0,
    "kl_seq_max": 0.2294921875
  },
  {
    "iteration": 12,
    "elapsed_s": 254.100581407547,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.5869724025974025,
    "avg_questions": 0.71875,
    "avg_turns": 1.6875,
    "ppo_loss": -0.03959607765344637,
    "value_loss": 0.34272722474166323,
    "approx_kl": 0.012407979046526765,
    "clip_frac": 0.20684523933700152,
    "entropy": 0.24096926195280893,
    "kl_penalty": 0.007428850446428571,
    "kl_per_seq": 0.007428850446428571,
    "kl_seq_max": 0.2572544642857143
  },
  {
    "iteration": 13,
    "elapsed_s": 335.7170238494873,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.6823145604395605,
    "avg_questions": 0.84375,
    "avg_turns": 1.84375,
    "ppo_loss": -0.03688290709008773,
    "value_loss": 0.21781166903674604,
    "approx_kl": 0.013189310727951428,
    "clip_frac": 0.19583333432674407,
    "entropy": 0.2323363353808721,
    "kl_penalty": 0.024055989583333333,
    "kl_per_seq": 0.024055989583333333,
    "kl_seq_max": 0.3768229166666667
  },
  {
    "iteration": 14,
    "elapsed_s": 264.29171442985535,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.7472098214285713,
    "avg_questions": 0.875,
    "avg_turns": 1.875,
    "ppo_loss": -0.027799080746869247,
    "value_loss": 0.4135706809659799,
    "approx_kl": 0.012586234758297602,
    "clip_frac": 0.15625,
    "entropy": 0.2517068100472291,
    "kl_penalty": 0.0135498046875,
    "kl_per_seq": 0.0135498046875,
    "kl_seq_max": 0.3229166666666667
  },
  {
    "iteration": 15,
    "elapsed_s": 241.14811420440674,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.6659226190476191,
    "avg_questions": 0.65625,
    "avg_turns": 1.65625,
    "ppo_loss": -0.03369090032009851,
    "value_loss": 0.34638335804144543,
    "approx_kl": 0.013905501381183663,
    "clip_frac": 0.19047619189534867,
    "entropy": 0.22407633740277516,
    "kl_penalty": 0.036342075892857144,
    "kl_per_seq": 0.036342075892857144,
    "kl_seq_max": 0.45982142857142855
  },
  {
    "iteration": 16,
    "elapsed_s": 238.80150389671326,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.6585475288600288,
    "avg_questions": 0.90625,
    "avg_turns": 1.84375,
    "ppo_loss": -0.056268499715736615,
    "value_loss": 0.3362144205581259,
    "approx_kl": 0.019216162325055512,
    "clip_frac": 0.3024691376421187,
    "entropy": 0.2751925124062432,
    "kl_penalty": 0.022162543402777776,
    "kl_per_seq": 0.022162543402777776,
    "kl_seq_max": 0.5263310185185185
  },
  {
    "iteration": 17,
    "elapsed_s": 232.22505855560303,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.7668154761904762,
    "avg_questions": 0.90625,
    "avg_turns": 1.875,
    "ppo_loss": -0.0331775728199217,
    "value_loss": 0.3092526584449742,
    "approx_kl": 0.01231207988328404,
    "clip_frac": 0.20833333333333334,
    "entropy": 0.25865906063053345,
    "kl_penalty": 0.07600911458333333,
    "kl_per_seq": 0.07600911458333333,
    "kl_seq_max": 0.65625
  },
  {
    "iteration": 18,
    "elapsed_s": 311.01340675354004,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.6972830988455989,
    "avg_questions": 0.8125,
    "avg_turns": 1.78125,
    "ppo_loss": 0.009420987218618393,
    "value_loss": 0.4465955500304699,
    "approx_kl": 0.015329486131668091,
    "clip_frac": 0.26,
    "entropy": 0.21777834206819535,
    "kl_penalty": 0.04673828125,
    "kl_per_seq": 0.04673828125,
    "kl_seq_max": 0.56625
  },
  {
    "iteration": 19,
    "elapsed_s": 249.58963227272034,
    "lambda1": 0.0,
    "lambda2": 0.0,
    "n_episodes": 32,
    "avg_reward": 0.7551739926739928,
    "avg_questions": 0.6875,
    "avg_turns": 1.6875,
    "ppo_loss": -0.037938538430766625,
    "value_loss": 0.36645138856362214,
    "approx_kl": 0.018135146047411996,
    "clip_frac": 0.2746212130243128,
    "entropy": 0.24184788357127796,
    "kl_penalty": 0.002840909090909091,
    "kl_per_seq": 0.002840909090909091,
    "kl_seq_max": 0.6051136363636364
  }
]