diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2568/trainer_state.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2568/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e3c5d040995086992b3f684c39d4ee97e09ac67
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2568/trainer_state.json
@@ -0,0 +1,610 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.0,
+  "eval_steps": 500,
+  "global_step": 2568,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8341716039180755,
+      "epoch": 0.11689070718877849,
+      "grad_norm": 1.1992322206497192,
+      "learning_rate": 5.366728168873013e-05,
+      "loss": 1.6918942260742187,
+      "mean_token_accuracy": 0.631798365265131,
+      "num_tokens": 153689.0,
+      "step": 50
+    },
+    {
+      "entropy": 0.9513179230690002,
+      "epoch": 0.23378141437755698,
+      "grad_norm": 1.1290454864501953,
+      "learning_rate": 0.00010842981402416902,
+      "loss": 0.8867578887939453,
+      "mean_token_accuracy": 0.7587438315153122,
+      "num_tokens": 310548.0,
+      "step": 100
+    },
+    {
+      "entropy": 0.8510265004634857,
+      "epoch": 0.3506721215663355,
+      "grad_norm": 0.7732136845588684,
+      "learning_rate": 0.00016319234635960792,
+      "loss": 0.7856448364257812,
+      "mean_token_accuracy": 0.7769897204637527,
+      "num_tokens": 468560.0,
+      "step": 150
+    },
+    {
+      "entropy": 0.8142367601394653,
+      "epoch": 0.46756282875511396,
+      "grad_norm": 0.7580806016921997,
+      "learning_rate": 0.00021795487869504684,
+      "loss": 0.7461666870117187,
+      "mean_token_accuracy": 0.7888966089487076,
+      "num_tokens": 619552.0,
+      "step": 200
+    },
+    {
+      "entropy": 0.772578022480011,
+      "epoch": 0.5844535359438925,
+      "grad_norm": 0.5330358147621155,
+      "learning_rate": 0.00027271741103048575,
+      "loss": 0.7142900848388671,
+      "mean_token_accuracy": 0.7959607627987861,
+      "num_tokens": 771972.0,
+      "step": 250
+    },
+    {
+      "entropy": 0.7481484657526016,
+      "epoch": 0.701344243132671,
+      "grad_norm": 0.8242517709732056,
+      "learning_rate": 0.00032747994336592464,
+      "loss": 0.6911422729492187,
+      "mean_token_accuracy": 0.8028205358982086,
+      "num_tokens": 928088.0,
+      "step": 300
+    },
+    {
+      "entropy": 0.741372903585434,
+      "epoch": 0.8182349503214494,
+      "grad_norm": 0.7722981572151184,
+      "learning_rate": 0.00038224247570136353,
+      "loss": 0.6908904266357422,
+      "mean_token_accuracy": 0.8018302822113037,
+      "num_tokens": 1086836.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.7257227802276611,
+      "epoch": 0.9351256575102279,
+      "grad_norm": 0.7969573140144348,
+      "learning_rate": 0.0004370050080368024,
+      "loss": 0.6714310455322265,
+      "mean_token_accuracy": 0.8057891410589219,
+      "num_tokens": 1241227.0,
+      "step": 400
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.8317563653766334,
+      "eval_loss": 0.8135058283805847,
+      "eval_mean_token_accuracy": 0.7755894215850087,
+      "eval_num_tokens": 1325643.0,
+      "eval_runtime": 37.4951,
+      "eval_samples_per_second": 32.831,
+      "eval_steps_per_second": 4.107,
+      "step": 428
+    },
+    {
+      "entropy": 0.7179872094087265,
+      "epoch": 1.0514319111630626,
+      "grad_norm": 0.6224768161773682,
+      "learning_rate": 0.00046873290101955413,
+      "loss": 0.6683222961425781,
+      "mean_token_accuracy": 0.8054787133207273,
+      "num_tokens": 1393922.0,
+      "step": 450
+    },
+    {
+      "entropy": 0.6991294291615486,
+      "epoch": 1.1683226183518411,
+      "grad_norm": 0.6813111901283264,
+      "learning_rate": 0.00046837443306086947,
+      "loss": 0.6493977355957031,
+      "mean_token_accuracy": 0.8089349576830864,
+      "num_tokens": 1551442.0,
+      "step": 500
+    },
+    {
+      "entropy": 0.6951853144168854,
+      "epoch": 1.2852133255406195,
+      "grad_norm": 0.617091953754425,
+      "learning_rate": 0.0004676269147738558,
+      "loss": 0.6484123229980469,
+      "mean_token_accuracy": 0.8088837671279907,
+      "num_tokens": 1706697.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.6852494943141937,
+      "epoch": 1.4021040327293979,
+      "grad_norm": 0.4677598774433136,
+      "learning_rate": 0.0004664915890374708,
+      "loss": 0.6416233062744141,
+      "mean_token_accuracy": 0.8106312158703805,
+      "num_tokens": 1865246.0,
+      "step": 600
+    },
+    {
+      "entropy": 0.6703594943881035,
+      "epoch": 1.5189947399181765,
+      "grad_norm": 0.5708025693893433,
+      "learning_rate": 0.0004649703435278991,
+      "loss": 0.6273183441162109,
+      "mean_token_accuracy": 0.8155005398392677,
+      "num_tokens": 2022060.0,
+      "step": 650
+    },
+    {
+      "entropy": 0.6676286320388317,
+      "epoch": 1.635885447106955,
+      "grad_norm": 0.5555347800254822,
+      "learning_rate": 0.00046306570757996264,
+      "loss": 0.6276700973510743,
+      "mean_token_accuracy": 0.8154828292131424,
+      "num_tokens": 2178909.0,
+      "step": 700
+    },
+    {
+      "entropy": 0.6669242936372757,
+      "epoch": 1.7527761542957334,
+      "grad_norm": 0.5260112285614014,
+      "learning_rate": 0.0004607808479816624,
+      "loss": 0.6228141784667969,
+      "mean_token_accuracy": 0.8164840793609619,
+      "num_tokens": 2329314.0,
+      "step": 750
+    },
+    {
+      "entropy": 0.6598560312390327,
+      "epoch": 1.869666861484512,
+      "grad_norm": 0.5677736401557922,
+      "learning_rate": 0.0004581195637088436,
+      "loss": 0.6214292907714843,
+      "mean_token_accuracy": 0.8171795177459716,
+      "num_tokens": 2478559.0,
+      "step": 800
+    },
+    {
+      "entropy": 0.6687870016694069,
+      "epoch": 1.9865575686732906,
+      "grad_norm": 0.6449615955352783,
+      "learning_rate": 0.00045508627960873823,
+      "loss": 0.6243909454345703,
+      "mean_token_accuracy": 0.8147780740261078,
+      "num_tokens": 2631937.0,
+      "step": 850
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.7786144471013701,
+      "eval_loss": 0.7599140405654907,
+      "eval_mean_token_accuracy": 0.7935443106409791,
+      "eval_num_tokens": 2651286.0,
+      "eval_runtime": 39.0048,
+      "eval_samples_per_second": 31.56,
+      "eval_steps_per_second": 3.948,
+      "step": 856
+    },
+    {
+      "entropy": 0.5821921329701966,
+      "epoch": 2.102863822326125,
+      "grad_norm": 0.6116960048675537,
+      "learning_rate": 0.00045168603904288863,
+      "loss": 0.5406004714965821,
+      "mean_token_accuracy": 0.8334921918921734,
+      "num_tokens": 2795200.0,
+      "step": 900
+    },
+    {
+      "entropy": 0.5937331764400006,
+      "epoch": 2.2197545295149035,
+      "grad_norm": 0.5440122485160828,
+      "learning_rate": 0.00044792449550168286,
+      "loss": 0.5515737533569336,
+      "mean_token_accuracy": 0.8314691257476806,
+      "num_tokens": 2950534.0,
+      "step": 950
+    },
+    {
+      "entropy": 0.5877273553609847,
+      "epoch": 2.3366452367036823,
+      "grad_norm": 0.6505260467529297,
+      "learning_rate": 0.0004438079032044453,
+      "loss": 0.5507744979858399,
+      "mean_token_accuracy": 0.8314063146710395,
+      "num_tokens": 3101471.0,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5897100016474723,
+      "epoch": 2.4535359438924607,
+      "grad_norm": 0.54567551612854,
+      "learning_rate": 0.0004393431067007111,
+      "loss": 0.5528768157958984,
+      "mean_token_accuracy": 0.8315245220065117,
+      "num_tokens": 3258554.0,
+      "step": 1050
+    },
+    {
+      "entropy": 0.5831590622663498,
+      "epoch": 2.570426651081239,
+      "grad_norm": 0.44826796650886536,
+      "learning_rate": 0.00043453752948997376,
+      "loss": 0.5442767333984375,
+      "mean_token_accuracy": 0.8327079233527184,
+      "num_tokens": 3416088.0,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5839906217157841,
+      "epoch": 2.6873173582700174,
+      "grad_norm": 0.5067696571350098,
+      "learning_rate": 0.0004293991616788285,
+      "loss": 0.5485663604736328,
+      "mean_token_accuracy": 0.8324281191825866,
+      "num_tokens": 3567251.0,
+      "step": 1150
+    },
+    {
+      "entropy": 0.5870274990797043,
+      "epoch": 2.8042080654587958,
+      "grad_norm": 0.44032806158065796,
+      "learning_rate": 0.00042393654669603217,
+      "loss": 0.5524474716186524,
+      "mean_token_accuracy": 0.8313613015413285,
+      "num_tokens": 3719464.0,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5864466108381748,
+      "epoch": 2.9210987726475746,
+      "grad_norm": 0.5520864129066467,
+      "learning_rate": 0.00041815876708756964,
+      "loss": 0.5490786361694336,
+      "mean_token_accuracy": 0.8314764249324799,
+      "num_tokens": 3874423.0,
+      "step": 1250
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6819627613990338,
+      "eval_loss": 0.7593190670013428,
+      "eval_mean_token_accuracy": 0.7857603395914102,
+      "eval_num_tokens": 3976929.0,
+      "eval_runtime": 39.3654,
+      "eval_samples_per_second": 31.271,
+      "eval_steps_per_second": 3.912,
+      "step": 1284
+    },
+    {
+      "entropy": 0.5587641523411525,
+      "epoch": 3.037405026300409,
+      "grad_norm": 0.622401773929596,
+      "learning_rate": 0.0004120754294153441,
+      "loss": 0.518932762145996,
+      "mean_token_accuracy": 0.8399260457436643,
+      "num_tokens": 4027609.0,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5028628017008304,
+      "epoch": 3.1542957334891875,
+      "grad_norm": 0.577396810054779,
+      "learning_rate": 0.00040569664828459917,
+      "loss": 0.4589382171630859,
+      "mean_token_accuracy": 0.8538418188691139,
+      "num_tokens": 4180560.0,
+      "step": 1350
+    },
+    {
+      "entropy": 0.5227407096326351,
+      "epoch": 3.2711864406779663,
+      "grad_norm": 0.5113071203231812,
+      "learning_rate": 0.00039903302952663176,
+      "loss": 0.47863777160644533,
+      "mean_token_accuracy": 0.847364938557148,
+      "num_tokens": 4329954.0,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5050077450275421,
+      "epoch": 3.3880771478667446,
+      "grad_norm": 0.48977184295654297,
+      "learning_rate": 0.0003920956525647558,
+      "loss": 0.46841480255126955,
+      "mean_token_accuracy": 0.8519425508379936,
+      "num_tokens": 4484354.0,
+      "step": 1450
+    },
+    {
+      "entropy": 0.5163291451334954,
+      "epoch": 3.504967855055523,
+      "grad_norm": 0.509864330291748,
+      "learning_rate": 0.000384896051992837,
+      "loss": 0.47587432861328127,
+      "mean_token_accuracy": 0.8486990982294083,
+      "num_tokens": 4638919.0,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5134807989001274,
+      "epoch": 3.6218585622443014,
+      "grad_norm": 0.46544620394706726,
+      "learning_rate": 0.00037744619839702735,
+      "loss": 0.47692710876464844,
+      "mean_token_accuracy": 0.8489899519085884,
+      "num_tokens": 4795540.0,
+      "step": 1550
+    },
+    {
+      "entropy": 0.5216251534223556,
+      "epoch": 3.73874926943308,
+      "grad_norm": 0.5167004466056824,
+      "learning_rate": 0.0003697584784525874,
+      "loss": 0.4837848663330078,
+      "mean_token_accuracy": 0.847066233754158,
+      "num_tokens": 4951177.0,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5254101701080799,
+      "epoch": 3.8556399766218585,
+      "grad_norm": 0.5504006743431091,
+      "learning_rate": 0.00036184567432888745,
+      "loss": 0.48759506225585936,
+      "mean_token_accuracy": 0.8457375919818878,
+      "num_tokens": 5104814.0,
+      "step": 1650
+    },
+    {
+      "entropy": 0.5107753933966159,
+      "epoch": 3.972530683810637,
+      "grad_norm": 0.421975314617157,
+      "learning_rate": 0.0003537209424368311,
+      "loss": 0.4743759536743164,
+      "mean_token_accuracy": 0.8505110186338425,
+      "num_tokens": 5265185.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.665543915002377,
+      "eval_loss": 0.7487082481384277,
+      "eval_mean_token_accuracy": 0.7999678904359991,
+      "eval_num_tokens": 5302572.0,
+      "eval_runtime": 38.5439,
+      "eval_samples_per_second": 31.938,
+      "eval_steps_per_second": 3.995,
+      "step": 1712
+    },
+    {
+      "entropy": 0.4414465431891494,
+      "epoch": 4.088836937463472,
+      "grad_norm": 0.5308493971824646,
+      "learning_rate": 0.0003453977915540383,
+      "loss": 0.39650299072265627,
+      "mean_token_accuracy": 0.8703725772287378,
+      "num_tokens": 5419002.0,
+      "step": 1750
+    },
+    {
+      "entropy": 0.42591195791959763,
+      "epoch": 4.20572764465225,
+      "grad_norm": 0.6111563444137573,
+      "learning_rate": 0.00033689006036415585,
+      "loss": 0.37838283538818357,
+      "mean_token_accuracy": 0.874235480427742,
+      "num_tokens": 5572969.0,
+      "step": 1800
+    },
+    {
+      "entropy": 0.44086351931095125,
+      "epoch": 4.322618351841029,
+      "grad_norm": 0.5524880290031433,
+      "learning_rate": 0.0003282118944476435,
+      "loss": 0.3920352554321289,
+      "mean_token_accuracy": 0.8694414687156677,
+      "num_tokens": 5733805.0,
+      "step": 1850
+    },
+    {
+      "entropy": 0.4380896310508251,
+      "epoch": 4.439509059029807,
+      "grad_norm": 0.5023863315582275,
+      "learning_rate": 0.0003193777227622898,
+      "loss": 0.3916081237792969,
+      "mean_token_accuracy": 0.8697139009833336,
+      "num_tokens": 5888694.0,
+      "step": 1900
+    },
+    {
+      "entropy": 0.44005885019898416,
+      "epoch": 4.556399766218585,
+      "grad_norm": 0.455997109413147,
+      "learning_rate": 0.000310402233652564,
+      "loss": 0.3955466842651367,
+      "mean_token_accuracy": 0.8693471103906631,
+      "num_tokens": 6042655.0,
+      "step": 1950
+    },
+    {
+      "entropy": 0.444269048422575,
+      "epoch": 4.673290473407365,
+      "grad_norm": 0.4542011320590973,
+      "learning_rate": 0.00030130035042769316,
+      "loss": 0.3988466262817383,
+      "mean_token_accuracy": 0.8680117425322532,
+      "num_tokens": 6203980.0,
+      "step": 2000
+    },
+    {
+      "entropy": 0.43520899042487143,
+      "epoch": 4.790181180596143,
+      "grad_norm": 0.49051445722579956,
+      "learning_rate": 0.0002920872065490688,
+      "loss": 0.39819797515869143,
+      "mean_token_accuracy": 0.8678940117359162,
+      "num_tokens": 6354604.0,
+      "step": 2050
+    },
+    {
+      "entropy": 0.429456724524498,
+      "epoch": 4.907071887784921,
+      "grad_norm": 0.5267980098724365,
+      "learning_rate": 0.0002827781204682396,
+      "loss": 0.39413917541503907,
+      "mean_token_accuracy": 0.870051506459713,
+      "num_tokens": 6507484.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.0,
+      "eval_entropy": 0.5826305385146823,
+      "eval_loss": 0.7694042921066284,
+      "eval_mean_token_accuracy": 0.7950637824349589,
+      "eval_num_tokens": 6628215.0,
+      "eval_runtime": 38.5641,
+      "eval_samples_per_second": 31.921,
+      "eval_steps_per_second": 3.993,
+      "step": 2140
+    },
+    {
+      "entropy": 0.41423329688496324,
+      "epoch": 5.023378141437756,
+      "grad_norm": 0.6393507719039917,
+      "learning_rate": 0.0002733885701573256,
+      "loss": 0.3733938598632813,
+      "mean_token_accuracy": 0.8762160557598325,
+      "num_tokens": 6662015.0,
+      "step": 2150
+    },
+    {
+      "entropy": 0.32116023637354374,
+      "epoch": 5.140268848626534,
+      "grad_norm": 0.5128306150436401,
+      "learning_rate": 0.00026393416737420213,
+      "loss": 0.2751211166381836,
+      "mean_token_accuracy": 0.9052674892544746,
+      "num_tokens": 6817989.0,
+      "step": 2200
+    },
+    {
+      "entropy": 0.33186347484588624,
+      "epoch": 5.257159555815313,
+      "grad_norm": 0.6007080078125,
+      "learning_rate": 0.0002544306317052408,
+      "loss": 0.2850212097167969,
+      "mean_token_accuracy": 0.9016379952430725,
+      "num_tokens": 6977021.0,
+      "step": 2250
+    },
+    {
+      "entropy": 0.3474555689096451,
+      "epoch": 5.374050263004091,
+      "grad_norm": 0.6231185793876648,
+      "learning_rate": 0.0002448937644287679,
+      "loss": 0.29555959701538087,
+      "mean_token_accuracy": 0.8966628012061119,
+      "num_tokens": 7133360.0,
+      "step": 2300
+    },
+    {
+      "entropy": 0.346554354429245,
+      "epoch": 5.490940970192869,
+      "grad_norm": 0.5146846175193787,
+      "learning_rate": 0.0002353394222426952,
+      "loss": 0.29524572372436525,
+      "mean_token_accuracy": 0.8978548383712769,
+      "num_tokens": 7282372.0,
+      "step": 2350
+    },
+    {
+      "entropy": 0.3467242659628391,
+      "epoch": 5.607831677381649,
+      "grad_norm": 0.6118131875991821,
+      "learning_rate": 0.00022578349090000624,
+      "loss": 0.2970552635192871,
+      "mean_token_accuracy": 0.8985717830061912,
+      "num_tokens": 7435766.0,
+      "step": 2400
+    },
+    {
+      "entropy": 0.34968974225223065,
+      "epoch": 5.724722384570427,
+      "grad_norm": 0.6346496939659119,
+      "learning_rate": 0.0002162418587959333,
+      "loss": 0.29940626144409177,
+      "mean_token_accuracy": 0.8972864350676537,
+      "num_tokens": 7578408.0,
+      "step": 2450
+    },
+    {
+      "entropy": 0.35084108062088487,
+      "epoch": 5.841613091759205,
+      "grad_norm": 0.6566210389137268,
+      "learning_rate": 0.00020673039055074148,
+      "loss": 0.302426700592041,
+      "mean_token_accuracy": 0.8953999072313309,
+      "num_tokens": 7734599.0,
+      "step": 2500
+    },
+    {
+      "entropy": 0.33895430400967597,
+      "epoch": 5.958503798947984,
+      "grad_norm": 0.517574667930603,
+      "learning_rate": 0.00019726490063204247,
+      "loss": 0.292838020324707,
+      "mean_token_accuracy": 0.8984868070483207,
+      "num_tokens": 7898028.0,
+      "step": 2550
+    },
+    {
+      "epoch": 6.0,
+      "eval_entropy": 0.5211211553254684,
+      "eval_loss": 0.8269560933113098,
+      "eval_mean_token_accuracy": 0.7899065927251593,
+      "eval_num_tokens": 7953858.0,
+      "eval_runtime": 39.369,
+      "eval_samples_per_second": 31.268,
+      "eval_steps_per_second": 3.912,
+      "step": 2568
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 4280,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.1913679730121984e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2996/README.md b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2996/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2996/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2996/adapter_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2996/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c2a7efe7af0ceadd86d0f851dc87d465191c777
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2996/adapter_config.json
@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.04180832058159435,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "v_proj",
+    "o_proj",
+    "q_proj",
+    "gate_proj",
+    "k_proj",
+    "down_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2996/chat_template.jinja b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2996/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2996/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is true %}
+        {{- '<think>\n' }}
+    {%- else %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2996/tokenizer_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2996/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2996/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2996/trainer_state.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2996/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d1c6e15c5e11dd2da4259f069c11ea4d81fd4156
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-2996/trainer_state.json
@@ -0,0 +1,701 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.0,
+  "eval_steps": 500,
+  "global_step": 2996,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8341716039180755,
+      "epoch": 0.11689070718877849,
+      "grad_norm": 1.1992322206497192,
+      "learning_rate": 5.366728168873013e-05,
+      "loss": 1.6918942260742187,
+      "mean_token_accuracy": 0.631798365265131,
+      "num_tokens": 153689.0,
+      "step": 50
+    },
+    {
+      "entropy": 0.9513179230690002,
+      "epoch": 0.23378141437755698,
+      "grad_norm": 1.1290454864501953,
+      "learning_rate": 0.00010842981402416902,
+      "loss": 0.8867578887939453,
+      "mean_token_accuracy": 0.7587438315153122,
+      "num_tokens": 310548.0,
+      "step": 100
+    },
+    {
+      "entropy": 0.8510265004634857,
+      "epoch": 0.3506721215663355,
+      "grad_norm": 0.7732136845588684,
+      "learning_rate": 0.00016319234635960792,
+      "loss": 0.7856448364257812,
+      "mean_token_accuracy": 0.7769897204637527,
+      "num_tokens": 468560.0,
+      "step": 150
+    },
+    {
+      "entropy": 0.8142367601394653,
+      "epoch": 0.46756282875511396,
+      "grad_norm": 0.7580806016921997,
+      "learning_rate": 0.00021795487869504684,
+      "loss": 0.7461666870117187,
+      "mean_token_accuracy": 0.7888966089487076,
+      "num_tokens": 619552.0,
+      "step": 200
+    },
+    {
+      "entropy": 0.772578022480011,
+      "epoch": 0.5844535359438925,
+      "grad_norm": 0.5330358147621155,
+      "learning_rate": 0.00027271741103048575,
+      "loss": 0.7142900848388671,
+      "mean_token_accuracy": 0.7959607627987861,
+      "num_tokens": 771972.0,
+      "step": 250
+    },
+    {
+      "entropy": 0.7481484657526016,
+      "epoch": 0.701344243132671,
+      "grad_norm": 0.8242517709732056,
+      "learning_rate": 0.00032747994336592464,
+      "loss": 0.6911422729492187,
+      "mean_token_accuracy": 0.8028205358982086,
+      "num_tokens": 928088.0,
+      "step": 300
+    },
+    {
+      "entropy": 0.741372903585434,
+      "epoch": 0.8182349503214494,
+      "grad_norm": 0.7722981572151184,
+      "learning_rate": 0.00038224247570136353,
+      "loss": 0.6908904266357422,
+      "mean_token_accuracy": 0.8018302822113037,
+      "num_tokens": 1086836.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.7257227802276611,
+      "epoch": 0.9351256575102279,
+      "grad_norm": 0.7969573140144348,
+      "learning_rate": 0.0004370050080368024,
+      "loss": 0.6714310455322265,
+      "mean_token_accuracy": 0.8057891410589219,
+      "num_tokens": 1241227.0,
+      "step": 400
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.8317563653766334,
+      "eval_loss": 0.8135058283805847,
+      "eval_mean_token_accuracy": 0.7755894215850087,
+      "eval_num_tokens": 1325643.0,
+      "eval_runtime": 37.4951,
+      "eval_samples_per_second": 32.831,
+      "eval_steps_per_second": 4.107,
+      "step": 428
+    },
+    {
+      "entropy": 0.7179872094087265,
+      "epoch": 1.0514319111630626,
+      "grad_norm": 0.6224768161773682,
+      "learning_rate": 0.00046873290101955413,
+      "loss": 0.6683222961425781,
+      "mean_token_accuracy": 0.8054787133207273,
+      "num_tokens": 1393922.0,
+      "step": 450
+    },
+    {
+      "entropy": 0.6991294291615486,
+      "epoch": 1.1683226183518411,
+      "grad_norm": 0.6813111901283264,
+      "learning_rate": 0.00046837443306086947,
+      "loss": 0.6493977355957031,
+      "mean_token_accuracy": 0.8089349576830864,
+      "num_tokens": 1551442.0,
+      "step": 500
+    },
+    {
+      "entropy": 0.6951853144168854,
+      "epoch": 1.2852133255406195,
+      "grad_norm": 0.617091953754425,
+      "learning_rate": 0.0004676269147738558,
+      "loss": 0.6484123229980469,
+      "mean_token_accuracy": 0.8088837671279907,
+      "num_tokens": 1706697.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.6852494943141937,
+      "epoch": 1.4021040327293979,
+      "grad_norm": 0.4677598774433136,
+      "learning_rate": 0.0004664915890374708,
+      "loss": 0.6416233062744141,
+      "mean_token_accuracy": 0.8106312158703805,
+      "num_tokens": 1865246.0,
+      "step": 600
+    },
+    {
+      "entropy": 0.6703594943881035,
+      "epoch": 1.5189947399181765,
+      "grad_norm": 0.5708025693893433,
+      "learning_rate": 0.0004649703435278991,
+      "loss": 0.6273183441162109,
+      "mean_token_accuracy": 0.8155005398392677,
+      "num_tokens": 2022060.0,
+      "step": 650
+    },
+    {
+      "entropy": 0.6676286320388317,
+      "epoch": 1.635885447106955,
+      "grad_norm": 0.5555347800254822,
+      "learning_rate": 0.00046306570757996264,
+      "loss": 0.6276700973510743,
+      "mean_token_accuracy": 0.8154828292131424,
+      "num_tokens": 2178909.0,
+      "step": 700
+    },
+    {
+      "entropy": 0.6669242936372757,
+      "epoch": 1.7527761542957334,
+      "grad_norm": 0.5260112285614014,
+      "learning_rate": 0.0004607808479816624,
+      "loss": 0.6228141784667969,
+      "mean_token_accuracy": 0.8164840793609619,
+      "num_tokens": 2329314.0,
+      "step": 750
+    },
+    {
+      "entropy": 0.6598560312390327,
+      "epoch": 1.869666861484512,
+      "grad_norm": 0.5677736401557922,
+      "learning_rate": 0.0004581195637088436,
+      "loss": 0.6214292907714843,
+      "mean_token_accuracy": 0.8171795177459716,
+      "num_tokens": 2478559.0,
+      "step": 800
+    },
+    {
+      "entropy": 0.6687870016694069,
+      "epoch": 1.9865575686732906,
+      "grad_norm": 0.6449615955352783,
+      "learning_rate": 0.00045508627960873823,
+      "loss": 0.6243909454345703,
+      "mean_token_accuracy": 0.8147780740261078,
+      "num_tokens": 2631937.0,
+      "step": 850
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.7786144471013701,
+      "eval_loss": 0.7599140405654907,
+      "eval_mean_token_accuracy": 0.7935443106409791,
+      "eval_num_tokens": 2651286.0,
+      "eval_runtime": 39.0048,
+      "eval_samples_per_second": 31.56,
+      "eval_steps_per_second": 3.948,
+      "step": 856
+    },
+    {
+      "entropy": 0.5821921329701966,
+      "epoch": 2.102863822326125,
+      "grad_norm": 0.6116960048675537,
+      "learning_rate": 0.00045168603904288863,
+      "loss": 0.5406004714965821,
+      "mean_token_accuracy": 0.8334921918921734,
+      "num_tokens": 2795200.0,
+      "step": 900
+    },
+    {
+      "entropy": 0.5937331764400006,
+      "epoch": 2.2197545295149035,
+      "grad_norm": 0.5440122485160828,
+      "learning_rate": 0.00044792449550168286,
+      "loss": 0.5515737533569336,
+      "mean_token_accuracy": 0.8314691257476806,
+      "num_tokens": 2950534.0,
+      "step": 950
+    },
+    {
+      "entropy": 0.5877273553609847,
+      "epoch": 2.3366452367036823,
+      "grad_norm": 0.6505260467529297,
+      "learning_rate": 0.0004438079032044453,
+      "loss": 0.5507744979858399,
+      "mean_token_accuracy": 0.8314063146710395,
+      "num_tokens": 3101471.0,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5897100016474723,
+      "epoch": 2.4535359438924607,
+      "grad_norm": 0.54567551612854,
+      "learning_rate": 0.0004393431067007111,
+      "loss": 0.5528768157958984,
+      "mean_token_accuracy": 0.8315245220065117,
+      "num_tokens": 3258554.0,
+      "step": 1050
+    },
+    {
+      "entropy": 0.5831590622663498,
+      "epoch": 2.570426651081239,
+      "grad_norm": 0.44826796650886536,
+      "learning_rate": 0.00043453752948997376,
+      "loss": 0.5442767333984375,
+      "mean_token_accuracy": 0.8327079233527184,
+      "num_tokens": 3416088.0,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5839906217157841,
+      "epoch": 2.6873173582700174,
+      "grad_norm": 0.5067696571350098,
+      "learning_rate": 0.0004293991616788285,
+      "loss": 0.5485663604736328,
+      "mean_token_accuracy": 0.8324281191825866,
+      "num_tokens": 3567251.0,
+      "step": 1150
+    },
+    {
+      "entropy": 0.5870274990797043,
+      "epoch": 2.8042080654587958,
+      "grad_norm": 0.44032806158065796,
+      "learning_rate": 0.00042393654669603217,
+      "loss": 0.5524474716186524,
+      "mean_token_accuracy": 0.8313613015413285,
+      "num_tokens": 3719464.0,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5864466108381748,
+      "epoch": 2.9210987726475746,
+      "grad_norm": 0.5520864129066467,
+      "learning_rate": 0.00041815876708756964,
+      "loss": 0.5490786361694336,
+      "mean_token_accuracy": 0.8314764249324799,
+      "num_tokens": 3874423.0,
+      "step": 1250
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6819627613990338,
+      "eval_loss": 0.7593190670013428,
+      "eval_mean_token_accuracy": 0.7857603395914102,
+      "eval_num_tokens": 3976929.0,
+      "eval_runtime": 39.3654,
+      "eval_samples_per_second": 31.271,
+      "eval_steps_per_second": 3.912,
+      "step": 1284
+    },
+    {
+      "entropy": 0.5587641523411525,
+      "epoch": 3.037405026300409,
+      "grad_norm": 0.622401773929596,
+      "learning_rate": 0.0004120754294153441,
+      "loss": 0.518932762145996,
+      "mean_token_accuracy": 0.8399260457436643,
+      "num_tokens": 4027609.0,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5028628017008304,
+      "epoch": 3.1542957334891875,
+      "grad_norm": 0.577396810054779,
+      "learning_rate": 0.00040569664828459917,
+      "loss": 0.4589382171630859,
+      "mean_token_accuracy": 0.8538418188691139,
+      "num_tokens": 4180560.0,
+      "step": 1350
+    },
+    {
+      "entropy": 0.5227407096326351,
+      "epoch": 3.2711864406779663,
+      "grad_norm": 0.5113071203231812,
+      "learning_rate": 0.00039903302952663176,
+      "loss": 0.47863777160644533,
+      "mean_token_accuracy": 0.847364938557148,
+      "num_tokens": 4329954.0,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5050077450275421,
+      "epoch": 3.3880771478667446,
+      "grad_norm": 0.48977184295654297,
+      "learning_rate": 0.0003920956525647558,
+      "loss": 0.46841480255126955,
+      "mean_token_accuracy": 0.8519425508379936,
+      "num_tokens": 4484354.0,
+      "step": 1450
+    },
+    {
+      "entropy": 0.5163291451334954,
+      "epoch": 3.504967855055523,
+      "grad_norm": 0.509864330291748,
+      "learning_rate": 0.000384896051992837,
+      "loss": 0.47587432861328127,
+      "mean_token_accuracy": 0.8486990982294083,
+      "num_tokens": 4638919.0,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5134807989001274,
+      "epoch": 3.6218585622443014,
+      "grad_norm": 0.46544620394706726,
+      "learning_rate": 0.00037744619839702735,
+      "loss": 0.47692710876464844,
+      "mean_token_accuracy": 0.8489899519085884,
+      "num_tokens": 4795540.0,
+      "step": 1550
+    },
+    {
+      "entropy": 0.5216251534223556,
+      "epoch": 3.73874926943308,
+      "grad_norm": 0.5167004466056824,
+      "learning_rate": 0.0003697584784525874,
+      "loss": 0.4837848663330078,
+      "mean_token_accuracy": 0.847066233754158,
+      "num_tokens": 4951177.0,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5254101701080799,
+      "epoch": 3.8556399766218585,
+      "grad_norm": 0.5504006743431091,
+      "learning_rate": 0.00036184567432888745,
+      "loss": 0.48759506225585936,
+      "mean_token_accuracy": 0.8457375919818878,
+      "num_tokens": 5104814.0,
+      "step": 1650
+    },
+    {
+      "entropy": 0.5107753933966159,
+      "epoch": 3.972530683810637,
+      "grad_norm": 0.421975314617157,
+      "learning_rate": 0.0003537209424368311,
+      "loss": 0.4743759536743164,
+      "mean_token_accuracy": 0.8505110186338425,
+      "num_tokens": 5265185.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.665543915002377,
+      "eval_loss": 0.7487082481384277,
+      "eval_mean_token_accuracy": 0.7999678904359991,
+      "eval_num_tokens": 5302572.0,
+      "eval_runtime": 38.5439,
+      "eval_samples_per_second": 31.938,
+      "eval_steps_per_second": 3.995,
+      "step": 1712
+    },
+    {
+      "entropy": 0.4414465431891494,
+      "epoch": 4.088836937463472,
+      "grad_norm": 0.5308493971824646,
+      "learning_rate": 0.0003453977915540383,
+      "loss": 0.39650299072265627,
+      "mean_token_accuracy": 0.8703725772287378,
+      "num_tokens": 5419002.0,
+      "step": 1750
+    },
+    {
+      "entropy": 0.42591195791959763,
+      "epoch": 4.20572764465225,
+      "grad_norm": 0.6111563444137573,
+      "learning_rate": 0.00033689006036415585,
+      "loss": 0.37838283538818357,
+      "mean_token_accuracy": 0.874235480427742,
+      "num_tokens": 5572969.0,
+      "step": 1800
+    },
+    {
+      "entropy": 0.44086351931095125,
+      "epoch": 4.322618351841029,
+      "grad_norm": 0.5524880290031433,
+      "learning_rate": 0.0003282118944476435,
+      "loss": 0.3920352554321289,
+      "mean_token_accuracy": 0.8694414687156677,
+      "num_tokens": 5733805.0,
+      "step": 1850
+    },
+    {
+      "entropy": 0.4380896310508251,
+      "epoch": 4.439509059029807,
+      "grad_norm": 0.5023863315582275,
+      "learning_rate": 0.0003193777227622898,
+      "loss": 0.3916081237792969,
+      "mean_token_accuracy": 0.8697139009833336,
+      "num_tokens": 5888694.0,
+      "step": 1900
+    },
+    {
+      "entropy": 0.44005885019898416,
+      "epoch": 4.556399766218585,
+      "grad_norm": 0.455997109413147,
+      "learning_rate": 0.000310402233652564,
+      "loss": 0.3955466842651367,
+      "mean_token_accuracy": 0.8693471103906631,
+      "num_tokens": 6042655.0,
+      "step": 1950
+    },
+    {
+      "entropy": 0.444269048422575,
+      "epoch": 4.673290473407365,
+      "grad_norm": 0.4542011320590973,
+      "learning_rate": 0.00030130035042769316,
+      "loss": 0.3988466262817383,
+      "mean_token_accuracy": 0.8680117425322532,
+      "num_tokens": 6203980.0,
+      "step": 2000
+    },
+    {
+      "entropy": 0.43520899042487143,
+      "epoch": 4.790181180596143,
+      "grad_norm": 0.49051445722579956,
+      "learning_rate": 0.0002920872065490688,
+      "loss": 0.39819797515869143,
+      "mean_token_accuracy": 0.8678940117359162,
+      "num_tokens": 6354604.0,
+      "step": 2050
+    },
+    {
+      "entropy": 0.429456724524498,
+      "epoch": 4.907071887784921,
+      "grad_norm": 0.5267980098724365,
+      "learning_rate": 0.0002827781204682396,
+      "loss": 0.39413917541503907,
+      "mean_token_accuracy": 0.870051506459713,
+      "num_tokens": 6507484.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.0,
+      "eval_entropy": 0.5826305385146823,
+      "eval_loss": 0.7694042921066284,
+      "eval_mean_token_accuracy": 0.7950637824349589,
+      "eval_num_tokens": 6628215.0,
+      "eval_runtime": 38.5641,
+      "eval_samples_per_second": 31.921,
+      "eval_steps_per_second": 3.993,
+      "step": 2140
+    },
+    {
+      "entropy": 0.41423329688496324,
+      "epoch": 5.023378141437756,
+      "grad_norm": 0.6393507719039917,
+      "learning_rate": 0.0002733885701573256,
+      "loss": 0.3733938598632813,
+      "mean_token_accuracy": 0.8762160557598325,
+      "num_tokens": 6662015.0,
+      "step": 2150
+    },
+    {
+      "entropy": 0.32116023637354374,
+      "epoch": 5.140268848626534,
+      "grad_norm": 0.5128306150436401,
+      "learning_rate": 0.00026393416737420213,
+      "loss": 0.2751211166381836,
+      "mean_token_accuracy": 0.9052674892544746,
+      "num_tokens": 6817989.0,
+      "step": 2200
+    },
+    {
+      "entropy": 0.33186347484588624,
+      "epoch": 5.257159555815313,
+      "grad_norm": 0.6007080078125,
+      "learning_rate": 0.0002544306317052408,
+      "loss": 0.2850212097167969,
+      "mean_token_accuracy": 0.9016379952430725,
+      "num_tokens": 6977021.0,
+      "step": 2250
+    },
+    {
+      "entropy": 0.3474555689096451,
+      "epoch": 5.374050263004091,
+      "grad_norm": 0.6231185793876648,
+      "learning_rate": 0.0002448937644287679,
+      "loss": 0.29555959701538087,
+      "mean_token_accuracy": 0.8966628012061119,
+      "num_tokens": 7133360.0,
+      "step": 2300
+    },
+    {
+      "entropy": 0.346554354429245,
+      "epoch": 5.490940970192869,
+      "grad_norm": 0.5146846175193787,
+      "learning_rate": 0.0002353394222426952,
+      "loss": 0.29524572372436525,
+      "mean_token_accuracy": 0.8978548383712769,
+      "num_tokens": 7282372.0,
+      "step": 2350
+    },
+    {
+      "entropy": 0.3467242659628391,
+      "epoch": 5.607831677381649,
+      "grad_norm": 0.6118131875991821,
+      "learning_rate": 0.00022578349090000624,
+      "loss": 0.2970552635192871,
+      "mean_token_accuracy": 0.8985717830061912,
+      "num_tokens": 7435766.0,
+      "step": 2400
+    },
+    {
+      "entropy": 0.34968974225223065,
+      "epoch": 5.724722384570427,
+      "grad_norm": 0.6346496939659119,
+      "learning_rate": 0.0002162418587959333,
+      "loss": 0.29940626144409177,
+      "mean_token_accuracy": 0.8972864350676537,
+      "num_tokens": 7578408.0,
+      "step": 2450
+    },
+    {
+      "entropy": 0.35084108062088487,
+      "epoch": 5.841613091759205,
+      "grad_norm": 0.6566210389137268,
+      "learning_rate": 0.00020673039055074148,
+      "loss": 0.302426700592041,
+      "mean_token_accuracy": 0.8953999072313309,
+      "num_tokens": 7734599.0,
+      "step": 2500
+    },
+    {
+      "entropy": 0.33895430400967597,
+      "epoch": 5.958503798947984,
+      "grad_norm": 0.517574667930603,
+      "learning_rate": 0.00019726490063204247,
+      "loss": 0.292838020324707,
+      "mean_token_accuracy": 0.8984868070483207,
+      "num_tokens": 7898028.0,
+      "step": 2550
+    },
+    {
+      "epoch": 6.0,
+      "eval_entropy": 0.5211211553254684,
+      "eval_loss": 0.8269560933113098,
+      "eval_mean_token_accuracy": 0.7899065927251593,
+      "eval_num_tokens": 7953858.0,
+      "eval_runtime": 39.369,
+      "eval_samples_per_second": 31.268,
+      "eval_steps_per_second": 3.912,
+      "step": 2568
+    },
+    {
+      "entropy": 0.2763742347009218,
+      "epoch": 6.074810052600818,
+      "grad_norm": 0.5116816163063049,
+      "learning_rate": 0.00018786112706049623,
+      "loss": 0.22524795532226563,
+      "mean_token_accuracy": 0.9218554844209297,
+      "num_tokens": 8052646.0,
+      "step": 2600
+    },
+    {
+      "entropy": 0.23341606348752975,
+      "epoch": 6.1917007597895966,
+      "grad_norm": 0.6335962414741516,
+      "learning_rate": 0.00017853470524261842,
+      "loss": 0.18153846740722657,
+      "mean_token_accuracy": 0.935965863764286,
+      "num_tokens": 8212323.0,
+      "step": 2650
+    },
+    {
+      "entropy": 0.23515729174017908,
+      "epoch": 6.308591466978375,
+      "grad_norm": 0.6312636733055115,
+      "learning_rate": 0.0001693011419742025,
+      "loss": 0.18562854766845704,
+      "mean_token_accuracy": 0.9342144966125489,
+      "num_tokens": 8366715.0,
+      "step": 2700
+    },
+    {
+      "entropy": 0.23566059060394765,
+      "epoch": 6.425482174167154,
+      "grad_norm": 0.5869485139846802,
+      "learning_rate": 0.0001601757896575784,
+      "loss": 0.1867989158630371,
+      "mean_token_accuracy": 0.934665755033493,
+      "num_tokens": 8515955.0,
+      "step": 2750
+    },
+    {
+      "entropy": 0.2446490554511547,
+      "epoch": 6.5423728813559325,
+      "grad_norm": 0.5957017540931702,
+      "learning_rate": 0.00015117382077557817,
+      "loss": 0.19312274932861329,
+      "mean_token_accuracy": 0.9308532625436783,
+      "num_tokens": 8672361.0,
+      "step": 2800
+    },
+    {
+      "entropy": 0.24066772796213626,
+      "epoch": 6.659263588544711,
+      "grad_norm": 0.6069761514663696,
+      "learning_rate": 0.0001423102026646483,
+      "loss": 0.19540096282958985,
+      "mean_token_accuracy": 0.9300534284114838,
+      "num_tokens": 8827819.0,
+      "step": 2850
+    },
+    {
+      "entropy": 0.2378876845538616,
+      "epoch": 6.776154295733489,
+      "grad_norm": 0.6069548726081848,
+      "learning_rate": 0.00013359967262905405,
+      "loss": 0.1942250633239746,
+      "mean_token_accuracy": 0.9310428243875504,
+      "num_tokens": 8979535.0,
+      "step": 2900
+    },
+    {
+      "entropy": 0.23810458809137344,
+      "epoch": 6.893045002922268,
+      "grad_norm": 0.5807433128356934,
+      "learning_rate": 0.00012505671343755173,
+      "loss": 0.1908321762084961,
+      "mean_token_accuracy": 0.9318096882104874,
+      "num_tokens": 9134134.0,
+      "step": 2950
+    },
+    {
+      "epoch": 7.0,
+      "eval_entropy": 0.4395190301266583,
+      "eval_loss": 0.9185124635696411,
+      "eval_mean_token_accuracy": 0.7936263409527865,
+      "eval_num_tokens": 9279501.0,
+      "eval_runtime": 37.9924,
+      "eval_samples_per_second": 32.401,
+      "eval_steps_per_second": 4.053,
+      "step": 2996
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 4280,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.3912420998060749e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3424/README.md b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3424/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3424/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3424/adapter_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3424/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c2a7efe7af0ceadd86d0f851dc87d465191c777
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3424/adapter_config.json
@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.04180832058159435,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "v_proj",
+    "o_proj",
+    "q_proj",
+    "gate_proj",
+    "k_proj",
+    "down_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3424/chat_template.jinja b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3424/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3424/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is true %}
+        {{- '<think>\n' }}
+    {%- else %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3424/tokenizer_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3424/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3424/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3424/trainer_state.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3424/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..fba14d8240b2820ddf6423b6d01f4d958a787940
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3424/trainer_state.json
@@ -0,0 +1,802 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.0,
+  "eval_steps": 500,
+  "global_step": 3424,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8341716039180755,
+      "epoch": 0.11689070718877849,
+      "grad_norm": 1.1992322206497192,
+      "learning_rate": 5.366728168873013e-05,
+      "loss": 1.6918942260742187,
+      "mean_token_accuracy": 0.631798365265131,
+      "num_tokens": 153689.0,
+      "step": 50
+    },
+    {
+      "entropy": 0.9513179230690002,
+      "epoch": 0.23378141437755698,
+      "grad_norm": 1.1290454864501953,
+      "learning_rate": 0.00010842981402416902,
+      "loss": 0.8867578887939453,
+      "mean_token_accuracy": 0.7587438315153122,
+      "num_tokens": 310548.0,
+      "step": 100
+    },
+    {
+      "entropy": 0.8510265004634857,
+      "epoch": 0.3506721215663355,
+      "grad_norm": 0.7732136845588684,
+      "learning_rate": 0.00016319234635960792,
+      "loss": 0.7856448364257812,
+      "mean_token_accuracy": 0.7769897204637527,
+      "num_tokens": 468560.0,
+      "step": 150
+    },
+    {
+      "entropy": 0.8142367601394653,
+      "epoch": 0.46756282875511396,
+      "grad_norm": 0.7580806016921997,
+      "learning_rate": 0.00021795487869504684,
+      "loss": 0.7461666870117187,
+      "mean_token_accuracy": 0.7888966089487076,
+      "num_tokens": 619552.0,
+      "step": 200
+    },
+    {
+      "entropy": 0.772578022480011,
+      "epoch": 0.5844535359438925,
+      "grad_norm": 0.5330358147621155,
+      "learning_rate": 0.00027271741103048575,
+      "loss": 0.7142900848388671,
+      "mean_token_accuracy": 0.7959607627987861,
+      "num_tokens": 771972.0,
+      "step": 250
+    },
+    {
+      "entropy": 0.7481484657526016,
+      "epoch": 0.701344243132671,
+      "grad_norm": 0.8242517709732056,
+      "learning_rate": 0.00032747994336592464,
+      "loss": 0.6911422729492187,
+      "mean_token_accuracy": 0.8028205358982086,
+      "num_tokens": 928088.0,
+      "step": 300
+    },
+    {
+      "entropy": 0.741372903585434,
+      "epoch": 0.8182349503214494,
+      "grad_norm": 0.7722981572151184,
+      "learning_rate": 0.00038224247570136353,
+      "loss": 0.6908904266357422,
+      "mean_token_accuracy": 0.8018302822113037,
+      "num_tokens": 1086836.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.7257227802276611,
+      "epoch": 0.9351256575102279,
+      "grad_norm": 0.7969573140144348,
+      "learning_rate": 0.0004370050080368024,
+      "loss": 0.6714310455322265,
+      "mean_token_accuracy": 0.8057891410589219,
+      "num_tokens": 1241227.0,
+      "step": 400
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.8317563653766334,
+      "eval_loss": 0.8135058283805847,
+      "eval_mean_token_accuracy": 0.7755894215850087,
+      "eval_num_tokens": 1325643.0,
+      "eval_runtime": 37.4951,
+      "eval_samples_per_second": 32.831,
+      "eval_steps_per_second": 4.107,
+      "step": 428
+    },
+    {
+      "entropy": 0.7179872094087265,
+      "epoch": 1.0514319111630626,
+      "grad_norm": 0.6224768161773682,
+      "learning_rate": 0.00046873290101955413,
+      "loss": 0.6683222961425781,
+      "mean_token_accuracy": 0.8054787133207273,
+      "num_tokens": 1393922.0,
+      "step": 450
+    },
+    {
+      "entropy": 0.6991294291615486,
+      "epoch": 1.1683226183518411,
+      "grad_norm": 0.6813111901283264,
+      "learning_rate": 0.00046837443306086947,
+      "loss": 0.6493977355957031,
+      "mean_token_accuracy": 0.8089349576830864,
+      "num_tokens": 1551442.0,
+      "step": 500
+    },
+    {
+      "entropy": 0.6951853144168854,
+      "epoch": 1.2852133255406195,
+      "grad_norm": 0.617091953754425,
+      "learning_rate": 0.0004676269147738558,
+      "loss": 0.6484123229980469,
+      "mean_token_accuracy": 0.8088837671279907,
+      "num_tokens": 1706697.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.6852494943141937,
+      "epoch": 1.4021040327293979,
+      "grad_norm": 0.4677598774433136,
+      "learning_rate": 0.0004664915890374708,
+      "loss": 0.6416233062744141,
+      "mean_token_accuracy": 0.8106312158703805,
+      "num_tokens": 1865246.0,
+      "step": 600
+    },
+    {
+      "entropy": 0.6703594943881035,
+      "epoch": 1.5189947399181765,
+      "grad_norm": 0.5708025693893433,
+      "learning_rate": 0.0004649703435278991,
+      "loss": 0.6273183441162109,
+      "mean_token_accuracy": 0.8155005398392677,
+      "num_tokens": 2022060.0,
+      "step": 650
+    },
+    {
+      "entropy": 0.6676286320388317,
+      "epoch": 1.635885447106955,
+      "grad_norm": 0.5555347800254822,
+      "learning_rate": 0.00046306570757996264,
+      "loss": 0.6276700973510743,
+      "mean_token_accuracy": 0.8154828292131424,
+      "num_tokens": 2178909.0,
+      "step": 700
+    },
+    {
+      "entropy": 0.6669242936372757,
+      "epoch": 1.7527761542957334,
+      "grad_norm": 0.5260112285614014,
+      "learning_rate": 0.0004607808479816624,
+      "loss": 0.6228141784667969,
+      "mean_token_accuracy": 0.8164840793609619,
+      "num_tokens": 2329314.0,
+      "step": 750
+    },
+    {
+      "entropy": 0.6598560312390327,
+      "epoch": 1.869666861484512,
+      "grad_norm": 0.5677736401557922,
+      "learning_rate": 0.0004581195637088436,
+      "loss": 0.6214292907714843,
+      "mean_token_accuracy": 0.8171795177459716,
+      "num_tokens": 2478559.0,
+      "step": 800
+    },
+    {
+      "entropy": 0.6687870016694069,
+      "epoch": 1.9865575686732906,
+      "grad_norm": 0.6449615955352783,
+      "learning_rate": 0.00045508627960873823,
+      "loss": 0.6243909454345703,
+      "mean_token_accuracy": 0.8147780740261078,
+      "num_tokens": 2631937.0,
+      "step": 850
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.7786144471013701,
+      "eval_loss": 0.7599140405654907,
+      "eval_mean_token_accuracy": 0.7935443106409791,
+      "eval_num_tokens": 2651286.0,
+      "eval_runtime": 39.0048,
+      "eval_samples_per_second": 31.56,
+      "eval_steps_per_second": 3.948,
+      "step": 856
+    },
+    {
+      "entropy": 0.5821921329701966,
+      "epoch": 2.102863822326125,
+      "grad_norm": 0.6116960048675537,
+      "learning_rate": 0.00045168603904288863,
+      "loss": 0.5406004714965821,
+      "mean_token_accuracy": 0.8334921918921734,
+      "num_tokens": 2795200.0,
+      "step": 900
+    },
+    {
+      "entropy": 0.5937331764400006,
+      "epoch": 2.2197545295149035,
+      "grad_norm": 0.5440122485160828,
+      "learning_rate": 0.00044792449550168286,
+      "loss": 0.5515737533569336,
+      "mean_token_accuracy": 0.8314691257476806,
+      "num_tokens": 2950534.0,
+      "step": 950
+    },
+    {
+      "entropy": 0.5877273553609847,
+      "epoch": 2.3366452367036823,
+      "grad_norm": 0.6505260467529297,
+      "learning_rate": 0.0004438079032044453,
+      "loss": 0.5507744979858399,
+      "mean_token_accuracy": 0.8314063146710395,
+      "num_tokens": 3101471.0,
+      "step": 1000
+    },
+    {
+      "entropy": 0.5897100016474723,
+      "epoch": 2.4535359438924607,
+      "grad_norm": 0.54567551612854,
+      "learning_rate": 0.0004393431067007111,
+      "loss": 0.5528768157958984,
+      "mean_token_accuracy": 0.8315245220065117,
+      "num_tokens": 3258554.0,
+      "step": 1050
+    },
+    {
+      "entropy": 0.5831590622663498,
+      "epoch": 2.570426651081239,
+      "grad_norm": 0.44826796650886536,
+      "learning_rate": 0.00043453752948997376,
+      "loss": 0.5442767333984375,
+      "mean_token_accuracy": 0.8327079233527184,
+      "num_tokens": 3416088.0,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5839906217157841,
+      "epoch": 2.6873173582700174,
+      "grad_norm": 0.5067696571350098,
+      "learning_rate": 0.0004293991616788285,
+      "loss": 0.5485663604736328,
+      "mean_token_accuracy": 0.8324281191825866,
+      "num_tokens": 3567251.0,
+      "step": 1150
+    },
+    {
+      "entropy": 0.5870274990797043,
+      "epoch": 2.8042080654587958,
+      "grad_norm": 0.44032806158065796,
+      "learning_rate": 0.00042393654669603217,
+      "loss": 0.5524474716186524,
+      "mean_token_accuracy": 0.8313613015413285,
+      "num_tokens": 3719464.0,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5864466108381748,
+      "epoch": 2.9210987726475746,
+      "grad_norm": 0.5520864129066467,
+      "learning_rate": 0.00041815876708756964,
+      "loss": 0.5490786361694336,
+      "mean_token_accuracy": 0.8314764249324799,
+      "num_tokens": 3874423.0,
+      "step": 1250
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6819627613990338,
+      "eval_loss": 0.7593190670013428,
+      "eval_mean_token_accuracy": 0.7857603395914102,
+      "eval_num_tokens": 3976929.0,
+      "eval_runtime": 39.3654,
+      "eval_samples_per_second": 31.271,
+      "eval_steps_per_second": 3.912,
+      "step": 1284
+    },
+    {
+      "entropy": 0.5587641523411525,
+      "epoch": 3.037405026300409,
+      "grad_norm": 0.622401773929596,
+      "learning_rate": 0.0004120754294153441,
+      "loss": 0.518932762145996,
+      "mean_token_accuracy": 0.8399260457436643,
+      "num_tokens": 4027609.0,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5028628017008304,
+      "epoch": 3.1542957334891875,
+      "grad_norm": 0.577396810054779,
+      "learning_rate": 0.00040569664828459917,
+      "loss": 0.4589382171630859,
+      "mean_token_accuracy": 0.8538418188691139,
+      "num_tokens": 4180560.0,
+      "step": 1350
+    },
+    {
+      "entropy": 0.5227407096326351,
+      "epoch": 3.2711864406779663,
+      "grad_norm": 0.5113071203231812,
+      "learning_rate": 0.00039903302952663176,
+      "loss": 0.47863777160644533,
+      "mean_token_accuracy": 0.847364938557148,
+      "num_tokens": 4329954.0,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5050077450275421,
+      "epoch": 3.3880771478667446,
+      "grad_norm": 0.48977184295654297,
+      "learning_rate": 0.0003920956525647558,
+      "loss": 0.46841480255126955,
+      "mean_token_accuracy": 0.8519425508379936,
+      "num_tokens": 4484354.0,
+      "step": 1450
+    },
+    {
+      "entropy": 0.5163291451334954,
+      "epoch": 3.504967855055523,
+      "grad_norm": 0.509864330291748,
+      "learning_rate": 0.000384896051992837,
+      "loss": 0.47587432861328127,
+      "mean_token_accuracy": 0.8486990982294083,
+      "num_tokens": 4638919.0,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5134807989001274,
+      "epoch": 3.6218585622443014,
+      "grad_norm": 0.46544620394706726,
+      "learning_rate": 0.00037744619839702735,
+      "loss": 0.47692710876464844,
+      "mean_token_accuracy": 0.8489899519085884,
+      "num_tokens": 4795540.0,
+      "step": 1550
+    },
+    {
+      "entropy": 0.5216251534223556,
+      "epoch": 3.73874926943308,
+      "grad_norm": 0.5167004466056824,
+      "learning_rate": 0.0003697584784525874,
+      "loss": 0.4837848663330078,
+      "mean_token_accuracy": 0.847066233754158,
+      "num_tokens": 4951177.0,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5254101701080799,
+      "epoch": 3.8556399766218585,
+      "grad_norm": 0.5504006743431091,
+      "learning_rate": 0.00036184567432888745,
+      "loss": 0.48759506225585936,
+      "mean_token_accuracy": 0.8457375919818878,
+      "num_tokens": 5104814.0,
+      "step": 1650
+    },
+    {
+      "entropy": 0.5107753933966159,
+      "epoch": 3.972530683810637,
+      "grad_norm": 0.421975314617157,
+      "learning_rate": 0.0003537209424368311,
+      "loss": 0.4743759536743164,
+      "mean_token_accuracy": 0.8505110186338425,
+      "num_tokens": 5265185.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.665543915002377,
+      "eval_loss": 0.7487082481384277,
+      "eval_mean_token_accuracy": 0.7999678904359991,
+      "eval_num_tokens": 5302572.0,
+      "eval_runtime": 38.5439,
+      "eval_samples_per_second": 31.938,
+      "eval_steps_per_second": 3.995,
+      "step": 1712
+    },
+    {
+      "entropy": 0.4414465431891494,
+      "epoch": 4.088836937463472,
+      "grad_norm": 0.5308493971824646,
+      "learning_rate": 0.0003453977915540383,
+      "loss": 0.39650299072265627,
+      "mean_token_accuracy": 0.8703725772287378,
+      "num_tokens": 5419002.0,
+      "step": 1750
+    },
+    {
+      "entropy": 0.42591195791959763,
+      "epoch": 4.20572764465225,
+      "grad_norm": 0.6111563444137573,
+      "learning_rate": 0.00033689006036415585,
+      "loss": 0.37838283538818357,
+      "mean_token_accuracy": 0.874235480427742,
+      "num_tokens": 5572969.0,
+      "step": 1800
+    },
+    {
+      "entropy": 0.44086351931095125,
+      "epoch": 4.322618351841029,
+      "grad_norm": 0.5524880290031433,
+      "learning_rate": 0.0003282118944476435,
+      "loss": 0.3920352554321289,
+      "mean_token_accuracy": 0.8694414687156677,
+      "num_tokens": 5733805.0,
+      "step": 1850
+    },
+    {
+      "entropy": 0.4380896310508251,
+      "epoch": 4.439509059029807,
+      "grad_norm": 0.5023863315582275,
+      "learning_rate": 0.0003193777227622898,
+      "loss": 0.3916081237792969,
+      "mean_token_accuracy": 0.8697139009833336,
+      "num_tokens": 5888694.0,
+      "step": 1900
+    },
+    {
+      "entropy": 0.44005885019898416,
+      "epoch": 4.556399766218585,
+      "grad_norm": 0.455997109413147,
+      "learning_rate": 0.000310402233652564,
+      "loss": 0.3955466842651367,
+      "mean_token_accuracy": 0.8693471103906631,
+      "num_tokens": 6042655.0,
+      "step": 1950
+    },
+    {
+      "entropy": 0.444269048422575,
+      "epoch": 4.673290473407365,
+      "grad_norm": 0.4542011320590973,
+      "learning_rate": 0.00030130035042769316,
+      "loss": 0.3988466262817383,
+      "mean_token_accuracy": 0.8680117425322532,
+      "num_tokens": 6203980.0,
+      "step": 2000
+    },
+    {
+      "entropy": 0.43520899042487143,
+      "epoch": 4.790181180596143,
+      "grad_norm": 0.49051445722579956,
+      "learning_rate": 0.0002920872065490688,
+      "loss": 0.39819797515869143,
+      "mean_token_accuracy": 0.8678940117359162,
+      "num_tokens": 6354604.0,
+      "step": 2050
+    },
+    {
+      "entropy": 0.429456724524498,
+      "epoch": 4.907071887784921,
+      "grad_norm": 0.5267980098724365,
+      "learning_rate": 0.0002827781204682396,
+      "loss": 0.39413917541503907,
+      "mean_token_accuracy": 0.870051506459713,
+      "num_tokens": 6507484.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.0,
+      "eval_entropy": 0.5826305385146823,
+      "eval_loss": 0.7694042921066284,
+      "eval_mean_token_accuracy": 0.7950637824349589,
+      "eval_num_tokens": 6628215.0,
+      "eval_runtime": 38.5641,
+      "eval_samples_per_second": 31.921,
+      "eval_steps_per_second": 3.993,
+      "step": 2140
+    },
+    {
+      "entropy": 0.41423329688496324,
+      "epoch": 5.023378141437756,
+      "grad_norm": 0.6393507719039917,
+      "learning_rate": 0.0002733885701573256,
+      "loss": 0.3733938598632813,
+      "mean_token_accuracy": 0.8762160557598325,
+      "num_tokens": 6662015.0,
+      "step": 2150
+    },
+    {
+      "entropy": 0.32116023637354374,
+      "epoch": 5.140268848626534,
+      "grad_norm": 0.5128306150436401,
+      "learning_rate": 0.00026393416737420213,
+      "loss": 0.2751211166381836,
+      "mean_token_accuracy": 0.9052674892544746,
+      "num_tokens": 6817989.0,
+      "step": 2200
+    },
+    {
+      "entropy": 0.33186347484588624,
+      "epoch": 5.257159555815313,
+      "grad_norm": 0.6007080078125,
+      "learning_rate": 0.0002544306317052408,
+      "loss": 0.2850212097167969,
+      "mean_token_accuracy": 0.9016379952430725,
+      "num_tokens": 6977021.0,
+      "step": 2250
+    },
+    {
+      "entropy": 0.3474555689096451,
+      "epoch": 5.374050263004091,
+      "grad_norm": 0.6231185793876648,
+      "learning_rate": 0.0002448937644287679,
+      "loss": 0.29555959701538087,
+      "mean_token_accuracy": 0.8966628012061119,
+      "num_tokens": 7133360.0,
+      "step": 2300
+    },
+    {
+      "entropy": 0.346554354429245,
+      "epoch": 5.490940970192869,
+      "grad_norm": 0.5146846175193787,
+      "learning_rate": 0.0002353394222426952,
+      "loss": 0.29524572372436525,
+      "mean_token_accuracy": 0.8978548383712769,
+      "num_tokens": 7282372.0,
+      "step": 2350
+    },
+    {
+      "entropy": 0.3467242659628391,
+      "epoch": 5.607831677381649,
+      "grad_norm": 0.6118131875991821,
+      "learning_rate": 0.00022578349090000624,
+      "loss": 0.2970552635192871,
+      "mean_token_accuracy": 0.8985717830061912,
+      "num_tokens": 7435766.0,
+      "step": 2400
+    },
+    {
+      "entropy": 0.34968974225223065,
+      "epoch": 5.724722384570427,
+      "grad_norm": 0.6346496939659119,
+      "learning_rate": 0.0002162418587959333,
+      "loss": 0.29940626144409177,
+      "mean_token_accuracy": 0.8972864350676537,
+      "num_tokens": 7578408.0,
+      "step": 2450
+    },
+    {
+      "entropy": 0.35084108062088487,
+      "epoch": 5.841613091759205,
+      "grad_norm": 0.6566210389137268,
+      "learning_rate": 0.00020673039055074148,
+      "loss": 0.302426700592041,
+      "mean_token_accuracy": 0.8953999072313309,
+      "num_tokens": 7734599.0,
+      "step": 2500
+    },
+    {
+      "entropy": 0.33895430400967597,
+      "epoch": 5.958503798947984,
+      "grad_norm": 0.517574667930603,
+      "learning_rate": 0.00019726490063204247,
+      "loss": 0.292838020324707,
+      "mean_token_accuracy": 0.8984868070483207,
+      "num_tokens": 7898028.0,
+      "step": 2550
+    },
+    {
+      "epoch": 6.0,
+      "eval_entropy": 0.5211211553254684,
+      "eval_loss": 0.8269560933113098,
+      "eval_mean_token_accuracy": 0.7899065927251593,
+      "eval_num_tokens": 7953858.0,
+      "eval_runtime": 39.369,
+      "eval_samples_per_second": 31.268,
+      "eval_steps_per_second": 3.912,
+      "step": 2568
+    },
+    {
+      "entropy": 0.2763742347009218,
+      "epoch": 6.074810052600818,
+      "grad_norm": 0.5116816163063049,
+      "learning_rate": 0.00018786112706049623,
+      "loss": 0.22524795532226563,
+      "mean_token_accuracy": 0.9218554844209297,
+      "num_tokens": 8052646.0,
+      "step": 2600
+    },
+    {
+      "entropy": 0.23341606348752975,
+      "epoch": 6.1917007597895966,
+      "grad_norm": 0.6335962414741516,
+      "learning_rate": 0.00017853470524261842,
+      "loss": 0.18153846740722657,
+      "mean_token_accuracy": 0.935965863764286,
+      "num_tokens": 8212323.0,
+      "step": 2650
+    },
+    {
+      "entropy": 0.23515729174017908,
+      "epoch": 6.308591466978375,
+      "grad_norm": 0.6312636733055115,
+      "learning_rate": 0.0001693011419742025,
+      "loss": 0.18562854766845704,
+      "mean_token_accuracy": 0.9342144966125489,
+      "num_tokens": 8366715.0,
+      "step": 2700
+    },
+    {
+      "entropy": 0.23566059060394765,
+      "epoch": 6.425482174167154,
+      "grad_norm": 0.5869485139846802,
+      "learning_rate": 0.0001601757896575784,
+      "loss": 0.1867989158630371,
+      "mean_token_accuracy": 0.934665755033493,
+      "num_tokens": 8515955.0,
+      "step": 2750
+    },
+    {
+      "entropy": 0.2446490554511547,
+      "epoch": 6.5423728813559325,
+      "grad_norm": 0.5957017540931702,
+      "learning_rate": 0.00015117382077557817,
+      "loss": 0.19312274932861329,
+      "mean_token_accuracy": 0.9308532625436783,
+      "num_tokens": 8672361.0,
+      "step": 2800
+    },
+    {
+      "entropy": 0.24066772796213626,
+      "epoch": 6.659263588544711,
+      "grad_norm": 0.6069761514663696,
+      "learning_rate": 0.0001423102026646483,
+      "loss": 0.19540096282958985,
+      "mean_token_accuracy": 0.9300534284114838,
+      "num_tokens": 8827819.0,
+      "step": 2850
+    },
+    {
+      "entropy": 0.2378876845538616,
+      "epoch": 6.776154295733489,
+      "grad_norm": 0.6069548726081848,
+      "learning_rate": 0.00013359967262905405,
+      "loss": 0.1942250633239746,
+      "mean_token_accuracy": 0.9310428243875504,
+      "num_tokens": 8979535.0,
+      "step": 2900
+    },
+    {
+      "entropy": 0.23810458809137344,
+      "epoch": 6.893045002922268,
+      "grad_norm": 0.5807433128356934,
+      "learning_rate": 0.00012505671343755173,
+      "loss": 0.1908321762084961,
+      "mean_token_accuracy": 0.9318096882104874,
+      "num_tokens": 9134134.0,
+      "step": 2950
+    },
+    {
+      "epoch": 7.0,
+      "eval_entropy": 0.4395190301266583,
+      "eval_loss": 0.9185124635696411,
+      "eval_mean_token_accuracy": 0.7936263409527865,
+      "eval_num_tokens": 9279501.0,
+      "eval_runtime": 37.9924,
+      "eval_samples_per_second": 32.401,
+      "eval_steps_per_second": 4.053,
+      "step": 2996
+    },
+    {
+      "entropy": 0.22885651856511083,
+      "epoch": 7.009351256575102,
+      "grad_norm": 0.49437037110328674,
+      "learning_rate": 0.00011669552924327112,
+      "loss": 0.18458213806152343,
+      "mean_token_accuracy": 0.9342338658457425,
+      "num_tokens": 9290416.0,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1516158328205347,
+      "epoch": 7.1262419637638805,
+      "grad_norm": 0.5319057703018188,
+      "learning_rate": 0.00010853002196684399,
+      "loss": 0.10484781265258789,
+      "mean_token_accuracy": 0.9638857007026672,
+      "num_tokens": 9444705.0,
+      "step": 3050
+    },
+    {
+      "entropy": 0.14421835087239743,
+      "epoch": 7.243132670952659,
+      "grad_norm": 0.7959536910057068,
+      "learning_rate": 0.00010057376818204711,
+      "loss": 0.10359338760375976,
+      "mean_token_accuracy": 0.964658388197422,
+      "num_tokens": 9601476.0,
+      "step": 3100
+    },
+    {
+      "entropy": 0.14138914909213782,
+      "epoch": 7.360023378141438,
+      "grad_norm": 0.5152567028999329,
+      "learning_rate": 9.283999654239098e-05,
+      "loss": 0.10241118431091309,
+      "mean_token_accuracy": 0.9644565668702125,
+      "num_tokens": 9757913.0,
+      "step": 3150
+    },
+    {
+      "entropy": 0.14779060527682306,
+      "epoch": 7.4769140853302165,
+      "grad_norm": 0.5124489068984985,
+      "learning_rate": 8.534156578618579e-05,
+      "loss": 0.10744266510009766,
+      "mean_token_accuracy": 0.9628658410906792,
+      "num_tokens": 9905790.0,
+      "step": 3200
+    },
+    {
+      "entropy": 0.1453312272951007,
+      "epoch": 7.593804792518995,
+      "grad_norm": 0.4871680438518524,
+      "learning_rate": 7.809094335665658e-05,
+      "loss": 0.10566988945007325,
+      "mean_token_accuracy": 0.9635729387402534,
+      "num_tokens": 10064234.0,
+      "step": 3250
+    },
+    {
+      "entropy": 0.14601945489645005,
+      "epoch": 7.710695499707773,
+      "grad_norm": 0.5538403987884521,
+      "learning_rate": 7.110018467265364e-05,
+      "loss": 0.10567939758300782,
+      "mean_token_accuracy": 0.9625948050618172,
+      "num_tokens": 10221089.0,
+      "step": 3300
+    },
+    {
+      "entropy": 0.14106836922466756,
+      "epoch": 7.827586206896552,
+      "grad_norm": 0.4848695695400238,
+      "learning_rate": 6.438091308442527e-05,
+      "loss": 0.10246496200561524,
+      "mean_token_accuracy": 0.9643726572394371,
+      "num_tokens": 10379655.0,
+      "step": 3350
+    },
+    {
+      "entropy": 0.14235207289457322,
+      "epoch": 7.94447691408533,
+      "grad_norm": 0.5310181379318237,
+      "learning_rate": 5.7944300547779826e-05,
+      "loss": 0.10354788780212403,
+      "mean_token_accuracy": 0.9633960571885108,
+      "num_tokens": 10531301.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.3818505619253431,
+      "eval_loss": 1.0742957592010498,
+      "eval_mean_token_accuracy": 0.7910223340059256,
+      "eval_num_tokens": 10605144.0,
+      "eval_runtime": 38.0708,
+      "eval_samples_per_second": 32.335,
+      "eval_steps_per_second": 4.045,
+      "step": 3424
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 4280,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.589156466300795e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3852/README.md b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3852/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3852/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3852/adapter_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3852/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c2a7efe7af0ceadd86d0f851dc87d465191c777
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-3852/adapter_config.json
@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.04180832058159435,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "v_proj",
+    "o_proj",
+    "q_proj",
+    "gate_proj",
+    "k_proj",
+    "down_proj",
+    "up_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-4280/tokenizer_config.json b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-4280/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/DBCA_original_Estonian/Qwen3.5-2B-Base_original_features_structural_train_original_features_structural_test1/checkpoint-4280/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/README.md b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..dd5a19f4b6265236f518255cb787811724b3becf
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/README.md
@@ -0,0 +1,58 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: transformers
+model_name: Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1
+tags:
+- generated_from_trainer
+- sft
+- trl
+licence: license
+---
+
+# Model Card for Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1
+
+This model is a fine-tuned version of [Qwen/Qwen3.5-2B-Base](https://huggingface.co/Qwen/Qwen3.5-2B-Base).
+It has been trained using [TRL](https://github.com/huggingface/trl).
+
+## Quick start
+
+```python
+from transformers import pipeline
+
+question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
+generator = pipeline("text-generation", model="None", device="cuda")
+output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
+print(output["generated_text"])
+```
+
+## Training procedure
+
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/katriin-kukk/Cross_lingual_morphological_generalization/runs/9asazldh) 
+
+
+
+This model was trained with SFT.
+
+### Framework versions
+
+- TRL: 0.29.0
+- Transformers: 5.5.4
+- Pytorch: 2.10.0
+- Datasets: 4.6.1
+- Tokenizers: 0.22.2
+
+## Citations
+
+
+
+Cite TRL as:
+    
+```bibtex
+@software{vonwerra2020trl,
+  title   = {{TRL: Transformers Reinforcement Learning}},
+  author  = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin},
+  license = {Apache-2.0},
+  url     = {https://github.com/huggingface/trl},
+  year    = {2020}
+}
+```
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1224/README.md b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1224/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1224/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1224/adapter_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1224/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b9596096c71017d49955136a4e23d9f97138d305
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1224/adapter_config.json
@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_bias": false,
+  "lora_dropout": 0.08602048083239339,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 16,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "gate_proj",
+    "up_proj",
+    "o_proj",
+    "down_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1224/chat_template.jinja b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1224/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1224/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is true %}
+        {{- '<think>\n' }}
+    {%- else %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1224/tokenizer_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1224/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1224/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1224/trainer_state.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1224/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..9106ebb115e5e5b8f52f77133ccdd55d333f4788
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1224/trainer_state.json
@@ -0,0 +1,340 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 1224,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.2408065378665922,
+      "epoch": 0.12269938650306748,
+      "grad_norm": 1.6718658208847046,
+      "learning_rate": 4.487804970912036e-05,
+      "loss": 2.129624786376953,
+      "mean_token_accuracy": 0.5743644836544991,
+      "num_tokens": 134329.0,
+      "step": 50
+    },
+    {
+      "entropy": 1.1078504550457,
+      "epoch": 0.24539877300613497,
+      "grad_norm": 2.4091813564300537,
+      "learning_rate": 9.067197798373296e-05,
+      "loss": 1.0346390533447265,
+      "mean_token_accuracy": 0.7085381114482879,
+      "num_tokens": 262407.0,
+      "step": 100
+    },
+    {
+      "entropy": 0.9421598988771439,
+      "epoch": 0.36809815950920244,
+      "grad_norm": 1.3913646936416626,
+      "learning_rate": 0.00013646590625834558,
+      "loss": 0.8716551208496094,
+      "mean_token_accuracy": 0.7449040985107422,
+      "num_tokens": 401136.0,
+      "step": 150
+    },
+    {
+      "entropy": 0.8723550146818161,
+      "epoch": 0.49079754601226994,
+      "grad_norm": 1.1664468050003052,
+      "learning_rate": 0.0001822598345329582,
+      "loss": 0.8095977783203125,
+      "mean_token_accuracy": 0.7585240352153778,
+      "num_tokens": 531647.0,
+      "step": 200
+    },
+    {
+      "entropy": 0.8466519421339035,
+      "epoch": 0.6134969325153374,
+      "grad_norm": 1.0145421028137207,
+      "learning_rate": 0.00022805376280757083,
+      "loss": 0.7830857849121093,
+      "mean_token_accuracy": 0.7661705195903779,
+      "num_tokens": 666127.0,
+      "step": 250
+    },
+    {
+      "entropy": 0.8180295366048813,
+      "epoch": 0.7361963190184049,
+      "grad_norm": 0.9731705784797668,
+      "learning_rate": 0.00027384769108218336,
+      "loss": 0.7573603820800782,
+      "mean_token_accuracy": 0.7704453605413437,
+      "num_tokens": 799292.0,
+      "step": 300
+    },
+    {
+      "entropy": 0.7990227991342544,
+      "epoch": 0.8588957055214724,
+      "grad_norm": 1.0642374753952026,
+      "learning_rate": 0.000319641619356796,
+      "loss": 0.73709228515625,
+      "mean_token_accuracy": 0.7752733880281448,
+      "num_tokens": 930614.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.7899087685346603,
+      "epoch": 0.9815950920245399,
+      "grad_norm": 0.880613386631012,
+      "learning_rate": 0.0003654355476314086,
+      "loss": 0.7347319793701171,
+      "mean_token_accuracy": 0.7766731631755829,
+      "num_tokens": 1061654.0,
+      "step": 400
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.83522472347532,
+      "eval_mean_token_accuracy": 0.7484098076820374,
+      "eval_not_syn_loss": 0.7872886061668396,
+      "eval_not_syn_runtime": 54.8697,
+      "eval_not_syn_samples_per_second": 25.424,
+      "eval_not_syn_steps_per_second": 3.189,
+      "eval_num_tokens": 1080196.0,
+      "step": 408
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.7784442271505083,
+      "eval_mean_token_accuracy": 0.8021791185651507,
+      "eval_num_tokens": 1080196.0,
+      "eval_syn_loss": 0.6980098485946655,
+      "eval_syn_runtime": 56.6342,
+      "eval_syn_samples_per_second": 24.632,
+      "eval_syn_steps_per_second": 3.09,
+      "step": 408
+    },
+    {
+      "entropy": 0.75672573694075,
+      "epoch": 1.1030674846625768,
+      "grad_norm": 0.9244198203086853,
+      "learning_rate": 0.00037356351883456205,
+      "loss": 0.7048745727539063,
+      "mean_token_accuracy": 0.783016896609104,
+      "num_tokens": 1199170.0,
+      "step": 450
+    },
+    {
+      "entropy": 0.7406797379255294,
+      "epoch": 1.2257668711656442,
+      "grad_norm": 0.6744896769523621,
+      "learning_rate": 0.00037311248151765587,
+      "loss": 0.6860730743408203,
+      "mean_token_accuracy": 0.787369327545166,
+      "num_tokens": 1334298.0,
+      "step": 500
+    },
+    {
+      "entropy": 0.7424973237514496,
+      "epoch": 1.3484662576687116,
+      "grad_norm": 0.92305588722229,
+      "learning_rate": 0.000372320629215228,
+      "loss": 0.6855724334716797,
+      "mean_token_accuracy": 0.7871134179830551,
+      "num_tokens": 1463189.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.7502243828773498,
+      "epoch": 1.471165644171779,
+      "grad_norm": 0.8259047269821167,
+      "learning_rate": 0.00037118941074037944,
+      "loss": 0.6949834442138672,
+      "mean_token_accuracy": 0.7851914083957672,
+      "num_tokens": 1590310.0,
+      "step": 600
+    },
+    {
+      "entropy": 0.7233260518312454,
+      "epoch": 1.5938650306748468,
+      "grad_norm": 0.882722795009613,
+      "learning_rate": 0.00036972089582775814,
+      "loss": 0.6649005889892579,
+      "mean_token_accuracy": 0.7898879665136337,
+      "num_tokens": 1721836.0,
+      "step": 650
+    },
+    {
+      "entropy": 0.7162708967924118,
+      "epoch": 1.716564417177914,
+      "grad_norm": 0.6593027114868164,
+      "learning_rate": 0.0003679177713466678,
+      "loss": 0.6653710174560546,
+      "mean_token_accuracy": 0.7905678844451904,
+      "num_tokens": 1853507.0,
+      "step": 700
+    },
+    {
+      "entropy": 0.7150520020723343,
+      "epoch": 1.8392638036809816,
+      "grad_norm": 0.665340781211853,
+      "learning_rate": 0.0003657833363850354,
+      "loss": 0.6600938415527344,
+      "mean_token_accuracy": 0.7918649202585221,
+      "num_tokens": 1988113.0,
+      "step": 750
+    },
+    {
+      "entropy": 0.6993386310338974,
+      "epoch": 1.961963190184049,
+      "grad_norm": 0.6502187252044678,
+      "learning_rate": 0.00036332149621323294,
+      "loss": 0.6469607543945313,
+      "mean_token_accuracy": 0.7950778317451477,
+      "num_tokens": 2121314.0,
+      "step": 800
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.7315139893123082,
+      "eval_mean_token_accuracy": 0.7713504004478454,
+      "eval_not_syn_loss": 0.699893593788147,
+      "eval_not_syn_runtime": 55.4135,
+      "eval_not_syn_samples_per_second": 25.174,
+      "eval_not_syn_steps_per_second": 3.158,
+      "eval_num_tokens": 2160392.0,
+      "step": 816
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.6848451798302787,
+      "eval_mean_token_accuracy": 0.814895657471248,
+      "eval_num_tokens": 2160392.0,
+      "eval_syn_loss": 0.6415576934814453,
+      "eval_syn_runtime": 56.4735,
+      "eval_syn_samples_per_second": 24.702,
+      "eval_syn_steps_per_second": 3.099,
+      "step": 816
+    },
+    {
+      "entropy": 0.6596363644407253,
+      "epoch": 2.083435582822086,
+      "grad_norm": 0.6647700071334839,
+      "learning_rate": 0.00036053675513879666,
+      "loss": 0.599151611328125,
+      "mean_token_accuracy": 0.8071986864311527,
+      "num_tokens": 2253179.0,
+      "step": 850
+    },
+    {
+      "entropy": 0.6626429110765457,
+      "epoch": 2.2061349693251535,
+      "grad_norm": 0.7356868386268616,
+      "learning_rate": 0.00035743420826511783,
+      "loss": 0.6004195404052735,
+      "mean_token_accuracy": 0.805362731218338,
+      "num_tokens": 2379965.0,
+      "step": 900
+    },
+    {
+      "entropy": 0.6503470712900161,
+      "epoch": 2.3288343558282207,
+      "grad_norm": 1.095608115196228,
+      "learning_rate": 0.0003540195321691835,
+      "loss": 0.5951333999633789,
+      "mean_token_accuracy": 0.807324025630951,
+      "num_tokens": 2515593.0,
+      "step": 950
+    },
+    {
+      "entropy": 0.660577870607376,
+      "epoch": 2.4515337423312884,
+      "grad_norm": 0.6602179408073425,
+      "learning_rate": 0.00035029897451542325,
+      "loss": 0.6056365585327148,
+      "mean_token_accuracy": 0.8057438349723816,
+      "num_tokens": 2642790.0,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6555853113532066,
+      "epoch": 2.574233128834356,
+      "grad_norm": 0.7459151148796082,
+      "learning_rate": 0.00034627934262466606,
+      "loss": 0.5963204193115235,
+      "mean_token_accuracy": 0.8060598099231719,
+      "num_tokens": 2774396.0,
+      "step": 1050
+    },
+    {
+      "entropy": 0.6505853188037872,
+      "epoch": 2.6969325153374233,
+      "grad_norm": 0.6359792947769165,
+      "learning_rate": 0.00034196799101912103,
+      "loss": 0.5993065643310547,
+      "mean_token_accuracy": 0.806280387043953,
+      "num_tokens": 2912102.0,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6509970253705979,
+      "epoch": 2.819631901840491,
+      "grad_norm": 0.8086174726486206,
+      "learning_rate": 0.0003373728079661707,
+      "loss": 0.6001268005371094,
+      "mean_token_accuracy": 0.806866540312767,
+      "num_tokens": 3042987.0,
+      "step": 1150
+    },
+    {
+      "entropy": 0.6404844307899475,
+      "epoch": 2.942331288343558,
+      "grad_norm": 0.7038407325744629,
+      "learning_rate": 0.0003325022010455975,
+      "loss": 0.5870526123046875,
+      "mean_token_accuracy": 0.8104586935043335,
+      "num_tokens": 3178044.0,
+      "step": 1200
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6645996727262224,
+      "eval_mean_token_accuracy": 0.798654237134116,
+      "eval_not_syn_loss": 0.6586771011352539,
+      "eval_not_syn_runtime": 55.1088,
+      "eval_not_syn_samples_per_second": 25.314,
+      "eval_not_syn_steps_per_second": 3.176,
+      "eval_num_tokens": 3240588.0,
+      "step": 1224
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6233607809884207,
+      "eval_mean_token_accuracy": 0.7977249867575509,
+      "eval_num_tokens": 3240588.0,
+      "eval_syn_loss": 0.6288298964500427,
+      "eval_syn_runtime": 56.6081,
+      "eval_syn_samples_per_second": 24.643,
+      "eval_syn_steps_per_second": 3.091,
+      "step": 1224
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 4080,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.458126379206016e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1632/README.md b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1632/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1632/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1632/adapter_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1632/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b9596096c71017d49955136a4e23d9f97138d305
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test1/checkpoint-1632/adapter_config.json
@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_bias": false,
+  "lora_dropout": 0.08602048083239339,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 16,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "gate_proj",
+    "up_proj",
+    "o_proj",
+    "down_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/README.md b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..8e55572239b6c32b5b9b6754aefce409039ac9ab
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/README.md
@@ -0,0 +1,58 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: transformers
+model_name: Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2
+tags:
+- generated_from_trainer
+- sft
+- trl
+licence: license
+---
+
+# Model Card for Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2
+
+This model is a fine-tuned version of [Qwen/Qwen3.5-2B-Base](https://huggingface.co/Qwen/Qwen3.5-2B-Base).
+It has been trained using [TRL](https://github.com/huggingface/trl).
+
+## Quick start
+
+```python
+from transformers import pipeline
+
+question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
+generator = pipeline("text-generation", model="None", device="cuda")
+output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
+print(output["generated_text"])
+```
+
+## Training procedure
+
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/katriin-kukk/Cross_lingual_morphological_generalization/runs/ypap1lsm) 
+
+
+
+This model was trained with SFT.
+
+### Framework versions
+
+- TRL: 0.29.0
+- Transformers: 5.5.4
+- Pytorch: 2.10.0
+- Datasets: 4.6.1
+- Tokenizers: 0.22.2
+
+## Citations
+
+
+
+Cite TRL as:
+    
+```bibtex
+@software{vonwerra2020trl,
+  title   = {{TRL: Transformers Reinforcement Learning}},
+  author  = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin},
+  license = {Apache-2.0},
+  url     = {https://github.com/huggingface/trl},
+  year    = {2020}
+}
+```
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1224/README.md b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1224/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1224/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1224/adapter_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1224/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..100cd582d71fb39c74fbba0e56e6d3415622454e
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1224/adapter_config.json
@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.05635041589438875,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "gate_proj",
+    "up_proj",
+    "o_proj",
+    "down_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1224/chat_template.jinja b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1224/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1224/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is true %}
+        {{- '<think>\n' }}
+    {%- else %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1224/tokenizer_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1224/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1224/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1224/trainer_state.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1224/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c75693c13b7b0250de0b2dc43ad401b8b00c19fc
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1224/trainer_state.json
@@ -0,0 +1,340 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 1224,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8153233402967452,
+      "epoch": 0.12269938650306748,
+      "grad_norm": 1.8510568141937256,
+      "learning_rate": 4.8469594667113335e-05,
+      "loss": 1.7217323303222656,
+      "mean_token_accuracy": 0.6236635231971741,
+      "num_tokens": 134329.0,
+      "step": 50
+    },
+    {
+      "entropy": 0.97831438601017,
+      "epoch": 0.24539877300613497,
+      "grad_norm": 1.889868974685669,
+      "learning_rate": 9.792836473559631e-05,
+      "loss": 0.9050540924072266,
+      "mean_token_accuracy": 0.7377463465929032,
+      "num_tokens": 262407.0,
+      "step": 100
+    },
+    {
+      "entropy": 0.8743131357431412,
+      "epoch": 0.36809815950920244,
+      "grad_norm": 1.1766732931137085,
+      "learning_rate": 0.0001473871348040793,
+      "loss": 0.8129763031005859,
+      "mean_token_accuracy": 0.7583417356014251,
+      "num_tokens": 401136.0,
+      "step": 150
+    },
+    {
+      "entropy": 0.8273966419696808,
+      "epoch": 0.49079754601226994,
+      "grad_norm": 1.1311014890670776,
+      "learning_rate": 0.00019684590487256229,
+      "loss": 0.7711280822753906,
+      "mean_token_accuracy": 0.7668228060007095,
+      "num_tokens": 531647.0,
+      "step": 200
+    },
+    {
+      "entropy": 0.8101438617706299,
+      "epoch": 0.6134969325153374,
+      "grad_norm": 0.8905950784683228,
+      "learning_rate": 0.0002463046749410453,
+      "loss": 0.7609222412109375,
+      "mean_token_accuracy": 0.7717174577713013,
+      "num_tokens": 666127.0,
+      "step": 250
+    },
+    {
+      "entropy": 0.7885055243968964,
+      "epoch": 0.7361963190184049,
+      "grad_norm": 1.1944245100021362,
+      "learning_rate": 0.00029576344500952824,
+      "loss": 0.7477743530273437,
+      "mean_token_accuracy": 0.773262197971344,
+      "num_tokens": 799292.0,
+      "step": 300
+    },
+    {
+      "entropy": 0.7858527797460556,
+      "epoch": 0.8588957055214724,
+      "grad_norm": 1.1311068534851074,
+      "learning_rate": 0.00034522221507801124,
+      "loss": 0.7411377716064453,
+      "mean_token_accuracy": 0.7745680212974548,
+      "num_tokens": 930614.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.7879797518253326,
+      "epoch": 0.9815950920245399,
+      "grad_norm": 0.9370436072349548,
+      "learning_rate": 0.00039468098514649423,
+      "loss": 0.7507785034179687,
+      "mean_token_accuracy": 0.7747444450855255,
+      "num_tokens": 1061654.0,
+      "step": 400
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.8527693561145238,
+      "eval_mean_token_accuracy": 0.7730878329277039,
+      "eval_not_syn_loss": 0.781921923160553,
+      "eval_not_syn_runtime": 56.9157,
+      "eval_not_syn_samples_per_second": 24.527,
+      "eval_not_syn_steps_per_second": 3.075,
+      "eval_num_tokens": 1080196.0,
+      "step": 408
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.7952005583899362,
+      "eval_mean_token_accuracy": 0.7705784467288426,
+      "eval_num_tokens": 1080196.0,
+      "eval_syn_loss": 0.7461121082305908,
+      "eval_syn_runtime": 59.5758,
+      "eval_syn_samples_per_second": 23.432,
+      "eval_syn_steps_per_second": 2.937,
+      "step": 408
+    },
+    {
+      "entropy": 0.7564875923021875,
+      "epoch": 1.1030674846625768,
+      "grad_norm": 1.1334599256515503,
+      "learning_rate": 0.00040345942966973637,
+      "loss": 0.7208955383300781,
+      "mean_token_accuracy": 0.7788144470465304,
+      "num_tokens": 1199170.0,
+      "step": 450
+    },
+    {
+      "entropy": 0.7427570706605912,
+      "epoch": 1.2257668711656442,
+      "grad_norm": 1.0777021646499634,
+      "learning_rate": 0.00040297229629224153,
+      "loss": 0.7049168395996094,
+      "mean_token_accuracy": 0.7830300116539002,
+      "num_tokens": 1334298.0,
+      "step": 500
+    },
+    {
+      "entropy": 0.7533353108167649,
+      "epoch": 1.3484662576687116,
+      "grad_norm": 1.3439561128616333,
+      "learning_rate": 0.00040211707285041045,
+      "loss": 0.7102288818359375,
+      "mean_token_accuracy": 0.7818449640274048,
+      "num_tokens": 1463189.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.7666342407464981,
+      "epoch": 1.471165644171779,
+      "grad_norm": 1.1329729557037354,
+      "learning_rate": 0.00040089532410439227,
+      "loss": 0.7270996856689453,
+      "mean_token_accuracy": 0.778404277563095,
+      "num_tokens": 1590310.0,
+      "step": 600
+    },
+    {
+      "entropy": 0.7359934949874878,
+      "epoch": 1.5938650306748468,
+      "grad_norm": 1.2227331399917603,
+      "learning_rate": 0.0003993092854276068,
+      "loss": 0.6996555328369141,
+      "mean_token_accuracy": 0.7834958267211914,
+      "num_tokens": 1721836.0,
+      "step": 650
+    },
+    {
+      "entropy": 0.7278932738304138,
+      "epoch": 1.716564417177914,
+      "grad_norm": 0.8753179311752319,
+      "learning_rate": 0.0003973618587167924,
+      "loss": 0.694002456665039,
+      "mean_token_accuracy": 0.7849677371978759,
+      "num_tokens": 1853507.0,
+      "step": 700
+    },
+    {
+      "entropy": 0.7190143716335297,
+      "epoch": 1.8392638036809816,
+      "grad_norm": 0.8524742126464844,
+      "learning_rate": 0.0003950566070825483,
+      "loss": 0.6890143585205079,
+      "mean_token_accuracy": 0.7865350896120071,
+      "num_tokens": 1988113.0,
+      "step": 750
+    },
+    {
+      "entropy": 0.7072037667036056,
+      "epoch": 1.961963190184049,
+      "grad_norm": 0.877348005771637,
+      "learning_rate": 0.00039239774833008714,
+      "loss": 0.6795162200927735,
+      "mean_token_accuracy": 0.7885233855247498,
+      "num_tokens": 2121314.0,
+      "step": 800
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.818787784235818,
+      "eval_mean_token_accuracy": 0.7573730857031686,
+      "eval_not_syn_loss": 0.7486392855644226,
+      "eval_not_syn_runtime": 56.7426,
+      "eval_not_syn_samples_per_second": 24.602,
+      "eval_not_syn_steps_per_second": 3.084,
+      "eval_num_tokens": 2160392.0,
+      "step": 816
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.7632522354807173,
+      "eval_mean_token_accuracy": 0.8101163625717163,
+      "eval_num_tokens": 2160392.0,
+      "eval_syn_loss": 0.6829464435577393,
+      "eval_syn_runtime": 58.5255,
+      "eval_syn_samples_per_second": 23.853,
+      "eval_syn_steps_per_second": 2.99,
+      "step": 816
+    },
+    {
+      "entropy": 0.6619202801675508,
+      "epoch": 2.083435582822086,
+      "grad_norm": 0.8124546408653259,
+      "learning_rate": 0.00038939014724212456,
+      "loss": 0.6311377716064454,
+      "mean_token_accuracy": 0.7991988044796567,
+      "num_tokens": 2253179.0,
+      "step": 850
+    },
+    {
+      "entropy": 0.6668815296888352,
+      "epoch": 2.2061349693251535,
+      "grad_norm": 0.8693709969520569,
+      "learning_rate": 0.0003860393066780265,
+      "loss": 0.6279015731811524,
+      "mean_token_accuracy": 0.7983763587474823,
+      "num_tokens": 2379965.0,
+      "step": 900
+    },
+    {
+      "entropy": 0.6557431703805924,
+      "epoch": 2.3288343558282207,
+      "grad_norm": 0.8323768377304077,
+      "learning_rate": 0.0003823513575054986,
+      "loss": 0.6214478683471679,
+      "mean_token_accuracy": 0.8012632429599762,
+      "num_tokens": 2515593.0,
+      "step": 950
+    },
+    {
+      "entropy": 0.6633540654182434,
+      "epoch": 2.4515337423312884,
+      "grad_norm": 0.8328827023506165,
+      "learning_rate": 0.0003783330473832399,
+      "loss": 0.6310849761962891,
+      "mean_token_accuracy": 0.7995503985881806,
+      "num_tokens": 2642790.0,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6542247200012207,
+      "epoch": 2.574233128834356,
+      "grad_norm": 0.8204748630523682,
+      "learning_rate": 0.000373991728415085,
+      "loss": 0.6195787048339844,
+      "mean_token_accuracy": 0.8006457507610321,
+      "num_tokens": 2774396.0,
+      "step": 1050
+    },
+    {
+      "entropy": 0.6526653742790223,
+      "epoch": 2.6969325153374233,
+      "grad_norm": 1.065635323524475,
+      "learning_rate": 0.0003693353436982218,
+      "loss": 0.6206568145751953,
+      "mean_token_accuracy": 0.8002620726823807,
+      "num_tokens": 2912102.0,
+      "step": 1100
+    },
+    {
+      "entropy": 0.654406590461731,
+      "epoch": 2.819631901840491,
+      "grad_norm": 0.8653210401535034,
+      "learning_rate": 0.00036437241279009834,
+      "loss": 0.6259954452514649,
+      "mean_token_accuracy": 0.8008036535978317,
+      "num_tokens": 3042987.0,
+      "step": 1150
+    },
+    {
+      "entropy": 0.6310120761394501,
+      "epoch": 2.942331288343558,
+      "grad_norm": 0.7713245749473572,
+      "learning_rate": 0.0003591120161206093,
+      "loss": 0.6115898132324219,
+      "mean_token_accuracy": 0.8043569129705429,
+      "num_tokens": 3178044.0,
+      "step": 1200
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6591688336644854,
+      "eval_mean_token_accuracy": 0.7992405581474304,
+      "eval_not_syn_loss": 0.6962469220161438,
+      "eval_not_syn_runtime": 56.161,
+      "eval_not_syn_samples_per_second": 24.857,
+      "eval_not_syn_steps_per_second": 3.116,
+      "eval_num_tokens": 3240588.0,
+      "step": 1224
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6191406740461077,
+      "eval_mean_token_accuracy": 0.7810143651281084,
+      "eval_num_tokens": 3240588.0,
+      "eval_syn_loss": 0.6663697361946106,
+      "eval_syn_runtime": 58.4102,
+      "eval_syn_samples_per_second": 23.9,
+      "eval_syn_steps_per_second": 2.996,
+      "step": 1224
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 4080,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.610858430973312e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1632/adapter_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1632/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..100cd582d71fb39c74fbba0e56e6d3415622454e
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1632/adapter_config.json
@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.05635041589438875,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "gate_proj",
+    "up_proj",
+    "o_proj",
+    "down_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1632/chat_template.jinja b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1632/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1632/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is true %}
+        {{- '<think>\n' }}
+    {%- else %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1632/tokenizer_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1632/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1632/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1632/trainer_state.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1632/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6dc70f3599903b9f071a73ade442f5211ddadc5a
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-1632/trainer_state.json
@@ -0,0 +1,442 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.0,
+  "eval_steps": 500,
+  "global_step": 1632,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8153233402967452,
+      "epoch": 0.12269938650306748,
+      "grad_norm": 1.8510568141937256,
+      "learning_rate": 4.8469594667113335e-05,
+      "loss": 1.7217323303222656,
+      "mean_token_accuracy": 0.6236635231971741,
+      "num_tokens": 134329.0,
+      "step": 50
+    },
+    {
+      "entropy": 0.97831438601017,
+      "epoch": 0.24539877300613497,
+      "grad_norm": 1.889868974685669,
+      "learning_rate": 9.792836473559631e-05,
+      "loss": 0.9050540924072266,
+      "mean_token_accuracy": 0.7377463465929032,
+      "num_tokens": 262407.0,
+      "step": 100
+    },
+    {
+      "entropy": 0.8743131357431412,
+      "epoch": 0.36809815950920244,
+      "grad_norm": 1.1766732931137085,
+      "learning_rate": 0.0001473871348040793,
+      "loss": 0.8129763031005859,
+      "mean_token_accuracy": 0.7583417356014251,
+      "num_tokens": 401136.0,
+      "step": 150
+    },
+    {
+      "entropy": 0.8273966419696808,
+      "epoch": 0.49079754601226994,
+      "grad_norm": 1.1311014890670776,
+      "learning_rate": 0.00019684590487256229,
+      "loss": 0.7711280822753906,
+      "mean_token_accuracy": 0.7668228060007095,
+      "num_tokens": 531647.0,
+      "step": 200
+    },
+    {
+      "entropy": 0.8101438617706299,
+      "epoch": 0.6134969325153374,
+      "grad_norm": 0.8905950784683228,
+      "learning_rate": 0.0002463046749410453,
+      "loss": 0.7609222412109375,
+      "mean_token_accuracy": 0.7717174577713013,
+      "num_tokens": 666127.0,
+      "step": 250
+    },
+    {
+      "entropy": 0.7885055243968964,
+      "epoch": 0.7361963190184049,
+      "grad_norm": 1.1944245100021362,
+      "learning_rate": 0.00029576344500952824,
+      "loss": 0.7477743530273437,
+      "mean_token_accuracy": 0.773262197971344,
+      "num_tokens": 799292.0,
+      "step": 300
+    },
+    {
+      "entropy": 0.7858527797460556,
+      "epoch": 0.8588957055214724,
+      "grad_norm": 1.1311068534851074,
+      "learning_rate": 0.00034522221507801124,
+      "loss": 0.7411377716064453,
+      "mean_token_accuracy": 0.7745680212974548,
+      "num_tokens": 930614.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.7879797518253326,
+      "epoch": 0.9815950920245399,
+      "grad_norm": 0.9370436072349548,
+      "learning_rate": 0.00039468098514649423,
+      "loss": 0.7507785034179687,
+      "mean_token_accuracy": 0.7747444450855255,
+      "num_tokens": 1061654.0,
+      "step": 400
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.8527693561145238,
+      "eval_mean_token_accuracy": 0.7730878329277039,
+      "eval_not_syn_loss": 0.781921923160553,
+      "eval_not_syn_runtime": 56.9157,
+      "eval_not_syn_samples_per_second": 24.527,
+      "eval_not_syn_steps_per_second": 3.075,
+      "eval_num_tokens": 1080196.0,
+      "step": 408
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.7952005583899362,
+      "eval_mean_token_accuracy": 0.7705784467288426,
+      "eval_num_tokens": 1080196.0,
+      "eval_syn_loss": 0.7461121082305908,
+      "eval_syn_runtime": 59.5758,
+      "eval_syn_samples_per_second": 23.432,
+      "eval_syn_steps_per_second": 2.937,
+      "step": 408
+    },
+    {
+      "entropy": 0.7564875923021875,
+      "epoch": 1.1030674846625768,
+      "grad_norm": 1.1334599256515503,
+      "learning_rate": 0.00040345942966973637,
+      "loss": 0.7208955383300781,
+      "mean_token_accuracy": 0.7788144470465304,
+      "num_tokens": 1199170.0,
+      "step": 450
+    },
+    {
+      "entropy": 0.7427570706605912,
+      "epoch": 1.2257668711656442,
+      "grad_norm": 1.0777021646499634,
+      "learning_rate": 0.00040297229629224153,
+      "loss": 0.7049168395996094,
+      "mean_token_accuracy": 0.7830300116539002,
+      "num_tokens": 1334298.0,
+      "step": 500
+    },
+    {
+      "entropy": 0.7533353108167649,
+      "epoch": 1.3484662576687116,
+      "grad_norm": 1.3439561128616333,
+      "learning_rate": 0.00040211707285041045,
+      "loss": 0.7102288818359375,
+      "mean_token_accuracy": 0.7818449640274048,
+      "num_tokens": 1463189.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.7666342407464981,
+      "epoch": 1.471165644171779,
+      "grad_norm": 1.1329729557037354,
+      "learning_rate": 0.00040089532410439227,
+      "loss": 0.7270996856689453,
+      "mean_token_accuracy": 0.778404277563095,
+      "num_tokens": 1590310.0,
+      "step": 600
+    },
+    {
+      "entropy": 0.7359934949874878,
+      "epoch": 1.5938650306748468,
+      "grad_norm": 1.2227331399917603,
+      "learning_rate": 0.0003993092854276068,
+      "loss": 0.6996555328369141,
+      "mean_token_accuracy": 0.7834958267211914,
+      "num_tokens": 1721836.0,
+      "step": 650
+    },
+    {
+      "entropy": 0.7278932738304138,
+      "epoch": 1.716564417177914,
+      "grad_norm": 0.8753179311752319,
+      "learning_rate": 0.0003973618587167924,
+      "loss": 0.694002456665039,
+      "mean_token_accuracy": 0.7849677371978759,
+      "num_tokens": 1853507.0,
+      "step": 700
+    },
+    {
+      "entropy": 0.7190143716335297,
+      "epoch": 1.8392638036809816,
+      "grad_norm": 0.8524742126464844,
+      "learning_rate": 0.0003950566070825483,
+      "loss": 0.6890143585205079,
+      "mean_token_accuracy": 0.7865350896120071,
+      "num_tokens": 1988113.0,
+      "step": 750
+    },
+    {
+      "entropy": 0.7072037667036056,
+      "epoch": 1.961963190184049,
+      "grad_norm": 0.877348005771637,
+      "learning_rate": 0.00039239774833008714,
+      "loss": 0.6795162200927735,
+      "mean_token_accuracy": 0.7885233855247498,
+      "num_tokens": 2121314.0,
+      "step": 800
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.818787784235818,
+      "eval_mean_token_accuracy": 0.7573730857031686,
+      "eval_not_syn_loss": 0.7486392855644226,
+      "eval_not_syn_runtime": 56.7426,
+      "eval_not_syn_samples_per_second": 24.602,
+      "eval_not_syn_steps_per_second": 3.084,
+      "eval_num_tokens": 2160392.0,
+      "step": 816
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.7632522354807173,
+      "eval_mean_token_accuracy": 0.8101163625717163,
+      "eval_num_tokens": 2160392.0,
+      "eval_syn_loss": 0.6829464435577393,
+      "eval_syn_runtime": 58.5255,
+      "eval_syn_samples_per_second": 23.853,
+      "eval_syn_steps_per_second": 2.99,
+      "step": 816
+    },
+    {
+      "entropy": 0.6619202801675508,
+      "epoch": 2.083435582822086,
+      "grad_norm": 0.8124546408653259,
+      "learning_rate": 0.00038939014724212456,
+      "loss": 0.6311377716064454,
+      "mean_token_accuracy": 0.7991988044796567,
+      "num_tokens": 2253179.0,
+      "step": 850
+    },
+    {
+      "entropy": 0.6668815296888352,
+      "epoch": 2.2061349693251535,
+      "grad_norm": 0.8693709969520569,
+      "learning_rate": 0.0003860393066780265,
+      "loss": 0.6279015731811524,
+      "mean_token_accuracy": 0.7983763587474823,
+      "num_tokens": 2379965.0,
+      "step": 900
+    },
+    {
+      "entropy": 0.6557431703805924,
+      "epoch": 2.3288343558282207,
+      "grad_norm": 0.8323768377304077,
+      "learning_rate": 0.0003823513575054986,
+      "loss": 0.6214478683471679,
+      "mean_token_accuracy": 0.8012632429599762,
+      "num_tokens": 2515593.0,
+      "step": 950
+    },
+    {
+      "entropy": 0.6633540654182434,
+      "epoch": 2.4515337423312884,
+      "grad_norm": 0.8328827023506165,
+      "learning_rate": 0.0003783330473832399,
+      "loss": 0.6310849761962891,
+      "mean_token_accuracy": 0.7995503985881806,
+      "num_tokens": 2642790.0,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6542247200012207,
+      "epoch": 2.574233128834356,
+      "grad_norm": 0.8204748630523682,
+      "learning_rate": 0.000373991728415085,
+      "loss": 0.6195787048339844,
+      "mean_token_accuracy": 0.8006457507610321,
+      "num_tokens": 2774396.0,
+      "step": 1050
+    },
+    {
+      "entropy": 0.6526653742790223,
+      "epoch": 2.6969325153374233,
+      "grad_norm": 1.065635323524475,
+      "learning_rate": 0.0003693353436982218,
+      "loss": 0.6206568145751953,
+      "mean_token_accuracy": 0.8002620726823807,
+      "num_tokens": 2912102.0,
+      "step": 1100
+    },
+    {
+      "entropy": 0.654406590461731,
+      "epoch": 2.819631901840491,
+      "grad_norm": 0.8653210401535034,
+      "learning_rate": 0.00036437241279009834,
+      "loss": 0.6259954452514649,
+      "mean_token_accuracy": 0.8008036535978317,
+      "num_tokens": 3042987.0,
+      "step": 1150
+    },
+    {
+      "entropy": 0.6310120761394501,
+      "epoch": 2.942331288343558,
+      "grad_norm": 0.7713245749473572,
+      "learning_rate": 0.0003591120161206093,
+      "loss": 0.6115898132324219,
+      "mean_token_accuracy": 0.8043569129705429,
+      "num_tokens": 3178044.0,
+      "step": 1200
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6591688336644854,
+      "eval_mean_token_accuracy": 0.7992405581474304,
+      "eval_not_syn_loss": 0.6962469220161438,
+      "eval_not_syn_runtime": 56.161,
+      "eval_not_syn_samples_per_second": 24.857,
+      "eval_not_syn_steps_per_second": 3.116,
+      "eval_num_tokens": 3240588.0,
+      "step": 1224
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6191406740461077,
+      "eval_mean_token_accuracy": 0.7810143651281084,
+      "eval_num_tokens": 3240588.0,
+      "eval_syn_loss": 0.6663697361946106,
+      "eval_syn_runtime": 58.4102,
+      "eval_syn_samples_per_second": 23.9,
+      "eval_syn_steps_per_second": 2.996,
+      "step": 1224
+    },
+    {
+      "entropy": 0.5853322166385073,
+      "epoch": 3.063803680981595,
+      "grad_norm": 0.7313649654388428,
+      "learning_rate": 0.00035356377837808084,
+      "loss": 0.5616408538818359,
+      "mean_token_accuracy": 0.8144418639366073,
+      "num_tokens": 3312797.0,
+      "step": 1250
+    },
+    {
+      "entropy": 0.5654504173994064,
+      "epoch": 3.1865030674846624,
+      "grad_norm": 0.713893711566925,
+      "learning_rate": 0.0003477378508994529,
+      "loss": 0.5305458068847656,
+      "mean_token_accuracy": 0.8218590825796127,
+      "num_tokens": 3441629.0,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5721602493524551,
+      "epoch": 3.30920245398773,
+      "grad_norm": 0.8487221002578735,
+      "learning_rate": 0.00034164489309687927,
+      "loss": 0.5437272262573242,
+      "mean_token_accuracy": 0.8169907212257386,
+      "num_tokens": 3576093.0,
+      "step": 1350
+    },
+    {
+      "entropy": 0.5797226822376251,
+      "epoch": 3.4319018404907977,
+      "grad_norm": 0.9307994842529297,
+      "learning_rate": 0.0003352960529547267,
+      "loss": 0.5405152130126953,
+      "mean_token_accuracy": 0.818563597202301,
+      "num_tokens": 3705912.0,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5822264245152473,
+      "epoch": 3.554601226993865,
+      "grad_norm": 0.71127849817276,
+      "learning_rate": 0.00032870294663265757,
+      "loss": 0.5559784698486329,
+      "mean_token_accuracy": 0.8159428876638413,
+      "num_tokens": 3832376.0,
+      "step": 1450
+    },
+    {
+      "entropy": 0.5702577340602875,
+      "epoch": 3.6773006134969326,
+      "grad_norm": 0.8090758919715881,
+      "learning_rate": 0.0003218776372121157,
+      "loss": 0.5466982650756836,
+      "mean_token_accuracy": 0.8186432421207428,
+      "num_tokens": 3967273.0,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5708746567368508,
+      "epoch": 3.8,
+      "grad_norm": 0.7777291536331177,
+      "learning_rate": 0.000314832612625101,
+      "loss": 0.5412663650512696,
+      "mean_token_accuracy": 0.8195344746112824,
+      "num_tokens": 4102434.0,
+      "step": 1550
+    },
+    {
+      "entropy": 0.578739655315876,
+      "epoch": 3.9226993865030675,
+      "grad_norm": 0.8103565573692322,
+      "learning_rate": 0.0003075807628056167,
+      "loss": 0.5605202865600586,
+      "mean_token_accuracy": 0.815134603381157,
+      "num_tokens": 4234037.0,
+      "step": 1600
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.6183825404303415,
+      "eval_mean_token_accuracy": 0.7838981601170131,
+      "eval_not_syn_loss": 0.6814553737640381,
+      "eval_not_syn_runtime": 56.4858,
+      "eval_not_syn_samples_per_second": 24.714,
+      "eval_not_syn_steps_per_second": 3.098,
+      "eval_num_tokens": 4320784.0,
+      "step": 1632
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5816301565510886,
+      "eval_mean_token_accuracy": 0.8187636685371399,
+      "eval_num_tokens": 4320784.0,
+      "eval_syn_loss": 0.6366032361984253,
+      "eval_syn_runtime": 58.1317,
+      "eval_syn_samples_per_second": 24.014,
+      "eval_syn_steps_per_second": 3.01,
+      "step": 1632
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 4080,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.819362572820186e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2040/README.md b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2040/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2040/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2040/adapter_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2040/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..100cd582d71fb39c74fbba0e56e6d3415622454e
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2040/adapter_config.json
@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.05635041589438875,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "gate_proj",
+    "up_proj",
+    "o_proj",
+    "down_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2040/chat_template.jinja b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2040/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2040/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is true %}
+        {{- '<think>\n' }}
+    {%- else %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2040/tokenizer_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2040/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2040/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2040/trainer_state.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2040/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ecdef5c4a11bc3deb327cd7941d115ac5fa346a2
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2040/trainer_state.json
@@ -0,0 +1,544 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.0,
+  "eval_steps": 500,
+  "global_step": 2040,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8153233402967452,
+      "epoch": 0.12269938650306748,
+      "grad_norm": 1.8510568141937256,
+      "learning_rate": 4.8469594667113335e-05,
+      "loss": 1.7217323303222656,
+      "mean_token_accuracy": 0.6236635231971741,
+      "num_tokens": 134329.0,
+      "step": 50
+    },
+    {
+      "entropy": 0.97831438601017,
+      "epoch": 0.24539877300613497,
+      "grad_norm": 1.889868974685669,
+      "learning_rate": 9.792836473559631e-05,
+      "loss": 0.9050540924072266,
+      "mean_token_accuracy": 0.7377463465929032,
+      "num_tokens": 262407.0,
+      "step": 100
+    },
+    {
+      "entropy": 0.8743131357431412,
+      "epoch": 0.36809815950920244,
+      "grad_norm": 1.1766732931137085,
+      "learning_rate": 0.0001473871348040793,
+      "loss": 0.8129763031005859,
+      "mean_token_accuracy": 0.7583417356014251,
+      "num_tokens": 401136.0,
+      "step": 150
+    },
+    {
+      "entropy": 0.8273966419696808,
+      "epoch": 0.49079754601226994,
+      "grad_norm": 1.1311014890670776,
+      "learning_rate": 0.00019684590487256229,
+      "loss": 0.7711280822753906,
+      "mean_token_accuracy": 0.7668228060007095,
+      "num_tokens": 531647.0,
+      "step": 200
+    },
+    {
+      "entropy": 0.8101438617706299,
+      "epoch": 0.6134969325153374,
+      "grad_norm": 0.8905950784683228,
+      "learning_rate": 0.0002463046749410453,
+      "loss": 0.7609222412109375,
+      "mean_token_accuracy": 0.7717174577713013,
+      "num_tokens": 666127.0,
+      "step": 250
+    },
+    {
+      "entropy": 0.7885055243968964,
+      "epoch": 0.7361963190184049,
+      "grad_norm": 1.1944245100021362,
+      "learning_rate": 0.00029576344500952824,
+      "loss": 0.7477743530273437,
+      "mean_token_accuracy": 0.773262197971344,
+      "num_tokens": 799292.0,
+      "step": 300
+    },
+    {
+      "entropy": 0.7858527797460556,
+      "epoch": 0.8588957055214724,
+      "grad_norm": 1.1311068534851074,
+      "learning_rate": 0.00034522221507801124,
+      "loss": 0.7411377716064453,
+      "mean_token_accuracy": 0.7745680212974548,
+      "num_tokens": 930614.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.7879797518253326,
+      "epoch": 0.9815950920245399,
+      "grad_norm": 0.9370436072349548,
+      "learning_rate": 0.00039468098514649423,
+      "loss": 0.7507785034179687,
+      "mean_token_accuracy": 0.7747444450855255,
+      "num_tokens": 1061654.0,
+      "step": 400
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.8527693561145238,
+      "eval_mean_token_accuracy": 0.7730878329277039,
+      "eval_not_syn_loss": 0.781921923160553,
+      "eval_not_syn_runtime": 56.9157,
+      "eval_not_syn_samples_per_second": 24.527,
+      "eval_not_syn_steps_per_second": 3.075,
+      "eval_num_tokens": 1080196.0,
+      "step": 408
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.7952005583899362,
+      "eval_mean_token_accuracy": 0.7705784467288426,
+      "eval_num_tokens": 1080196.0,
+      "eval_syn_loss": 0.7461121082305908,
+      "eval_syn_runtime": 59.5758,
+      "eval_syn_samples_per_second": 23.432,
+      "eval_syn_steps_per_second": 2.937,
+      "step": 408
+    },
+    {
+      "entropy": 0.7564875923021875,
+      "epoch": 1.1030674846625768,
+      "grad_norm": 1.1334599256515503,
+      "learning_rate": 0.00040345942966973637,
+      "loss": 0.7208955383300781,
+      "mean_token_accuracy": 0.7788144470465304,
+      "num_tokens": 1199170.0,
+      "step": 450
+    },
+    {
+      "entropy": 0.7427570706605912,
+      "epoch": 1.2257668711656442,
+      "grad_norm": 1.0777021646499634,
+      "learning_rate": 0.00040297229629224153,
+      "loss": 0.7049168395996094,
+      "mean_token_accuracy": 0.7830300116539002,
+      "num_tokens": 1334298.0,
+      "step": 500
+    },
+    {
+      "entropy": 0.7533353108167649,
+      "epoch": 1.3484662576687116,
+      "grad_norm": 1.3439561128616333,
+      "learning_rate": 0.00040211707285041045,
+      "loss": 0.7102288818359375,
+      "mean_token_accuracy": 0.7818449640274048,
+      "num_tokens": 1463189.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.7666342407464981,
+      "epoch": 1.471165644171779,
+      "grad_norm": 1.1329729557037354,
+      "learning_rate": 0.00040089532410439227,
+      "loss": 0.7270996856689453,
+      "mean_token_accuracy": 0.778404277563095,
+      "num_tokens": 1590310.0,
+      "step": 600
+    },
+    {
+      "entropy": 0.7359934949874878,
+      "epoch": 1.5938650306748468,
+      "grad_norm": 1.2227331399917603,
+      "learning_rate": 0.0003993092854276068,
+      "loss": 0.6996555328369141,
+      "mean_token_accuracy": 0.7834958267211914,
+      "num_tokens": 1721836.0,
+      "step": 650
+    },
+    {
+      "entropy": 0.7278932738304138,
+      "epoch": 1.716564417177914,
+      "grad_norm": 0.8753179311752319,
+      "learning_rate": 0.0003973618587167924,
+      "loss": 0.694002456665039,
+      "mean_token_accuracy": 0.7849677371978759,
+      "num_tokens": 1853507.0,
+      "step": 700
+    },
+    {
+      "entropy": 0.7190143716335297,
+      "epoch": 1.8392638036809816,
+      "grad_norm": 0.8524742126464844,
+      "learning_rate": 0.0003950566070825483,
+      "loss": 0.6890143585205079,
+      "mean_token_accuracy": 0.7865350896120071,
+      "num_tokens": 1988113.0,
+      "step": 750
+    },
+    {
+      "entropy": 0.7072037667036056,
+      "epoch": 1.961963190184049,
+      "grad_norm": 0.877348005771637,
+      "learning_rate": 0.00039239774833008714,
+      "loss": 0.6795162200927735,
+      "mean_token_accuracy": 0.7885233855247498,
+      "num_tokens": 2121314.0,
+      "step": 800
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.818787784235818,
+      "eval_mean_token_accuracy": 0.7573730857031686,
+      "eval_not_syn_loss": 0.7486392855644226,
+      "eval_not_syn_runtime": 56.7426,
+      "eval_not_syn_samples_per_second": 24.602,
+      "eval_not_syn_steps_per_second": 3.084,
+      "eval_num_tokens": 2160392.0,
+      "step": 816
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.7632522354807173,
+      "eval_mean_token_accuracy": 0.8101163625717163,
+      "eval_num_tokens": 2160392.0,
+      "eval_syn_loss": 0.6829464435577393,
+      "eval_syn_runtime": 58.5255,
+      "eval_syn_samples_per_second": 23.853,
+      "eval_syn_steps_per_second": 2.99,
+      "step": 816
+    },
+    {
+      "entropy": 0.6619202801675508,
+      "epoch": 2.083435582822086,
+      "grad_norm": 0.8124546408653259,
+      "learning_rate": 0.00038939014724212456,
+      "loss": 0.6311377716064454,
+      "mean_token_accuracy": 0.7991988044796567,
+      "num_tokens": 2253179.0,
+      "step": 850
+    },
+    {
+      "entropy": 0.6668815296888352,
+      "epoch": 2.2061349693251535,
+      "grad_norm": 0.8693709969520569,
+      "learning_rate": 0.0003860393066780265,
+      "loss": 0.6279015731811524,
+      "mean_token_accuracy": 0.7983763587474823,
+      "num_tokens": 2379965.0,
+      "step": 900
+    },
+    {
+      "entropy": 0.6557431703805924,
+      "epoch": 2.3288343558282207,
+      "grad_norm": 0.8323768377304077,
+      "learning_rate": 0.0003823513575054986,
+      "loss": 0.6214478683471679,
+      "mean_token_accuracy": 0.8012632429599762,
+      "num_tokens": 2515593.0,
+      "step": 950
+    },
+    {
+      "entropy": 0.6633540654182434,
+      "epoch": 2.4515337423312884,
+      "grad_norm": 0.8328827023506165,
+      "learning_rate": 0.0003783330473832399,
+      "loss": 0.6310849761962891,
+      "mean_token_accuracy": 0.7995503985881806,
+      "num_tokens": 2642790.0,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6542247200012207,
+      "epoch": 2.574233128834356,
+      "grad_norm": 0.8204748630523682,
+      "learning_rate": 0.000373991728415085,
+      "loss": 0.6195787048339844,
+      "mean_token_accuracy": 0.8006457507610321,
+      "num_tokens": 2774396.0,
+      "step": 1050
+    },
+    {
+      "entropy": 0.6526653742790223,
+      "epoch": 2.6969325153374233,
+      "grad_norm": 1.065635323524475,
+      "learning_rate": 0.0003693353436982218,
+      "loss": 0.6206568145751953,
+      "mean_token_accuracy": 0.8002620726823807,
+      "num_tokens": 2912102.0,
+      "step": 1100
+    },
+    {
+      "entropy": 0.654406590461731,
+      "epoch": 2.819631901840491,
+      "grad_norm": 0.8653210401535034,
+      "learning_rate": 0.00036437241279009834,
+      "loss": 0.6259954452514649,
+      "mean_token_accuracy": 0.8008036535978317,
+      "num_tokens": 3042987.0,
+      "step": 1150
+    },
+    {
+      "entropy": 0.6310120761394501,
+      "epoch": 2.942331288343558,
+      "grad_norm": 0.7713245749473572,
+      "learning_rate": 0.0003591120161206093,
+      "loss": 0.6115898132324219,
+      "mean_token_accuracy": 0.8043569129705429,
+      "num_tokens": 3178044.0,
+      "step": 1200
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6591688336644854,
+      "eval_mean_token_accuracy": 0.7992405581474304,
+      "eval_not_syn_loss": 0.6962469220161438,
+      "eval_not_syn_runtime": 56.161,
+      "eval_not_syn_samples_per_second": 24.857,
+      "eval_not_syn_steps_per_second": 3.116,
+      "eval_num_tokens": 3240588.0,
+      "step": 1224
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6191406740461077,
+      "eval_mean_token_accuracy": 0.7810143651281084,
+      "eval_num_tokens": 3240588.0,
+      "eval_syn_loss": 0.6663697361946106,
+      "eval_syn_runtime": 58.4102,
+      "eval_syn_samples_per_second": 23.9,
+      "eval_syn_steps_per_second": 2.996,
+      "step": 1224
+    },
+    {
+      "entropy": 0.5853322166385073,
+      "epoch": 3.063803680981595,
+      "grad_norm": 0.7313649654388428,
+      "learning_rate": 0.00035356377837808084,
+      "loss": 0.5616408538818359,
+      "mean_token_accuracy": 0.8144418639366073,
+      "num_tokens": 3312797.0,
+      "step": 1250
+    },
+    {
+      "entropy": 0.5654504173994064,
+      "epoch": 3.1865030674846624,
+      "grad_norm": 0.713893711566925,
+      "learning_rate": 0.0003477378508994529,
+      "loss": 0.5305458068847656,
+      "mean_token_accuracy": 0.8218590825796127,
+      "num_tokens": 3441629.0,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5721602493524551,
+      "epoch": 3.30920245398773,
+      "grad_norm": 0.8487221002578735,
+      "learning_rate": 0.00034164489309687927,
+      "loss": 0.5437272262573242,
+      "mean_token_accuracy": 0.8169907212257386,
+      "num_tokens": 3576093.0,
+      "step": 1350
+    },
+    {
+      "entropy": 0.5797226822376251,
+      "epoch": 3.4319018404907977,
+      "grad_norm": 0.9307994842529297,
+      "learning_rate": 0.0003352960529547267,
+      "loss": 0.5405152130126953,
+      "mean_token_accuracy": 0.818563597202301,
+      "num_tokens": 3705912.0,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5822264245152473,
+      "epoch": 3.554601226993865,
+      "grad_norm": 0.71127849817276,
+      "learning_rate": 0.00032870294663265757,
+      "loss": 0.5559784698486329,
+      "mean_token_accuracy": 0.8159428876638413,
+      "num_tokens": 3832376.0,
+      "step": 1450
+    },
+    {
+      "entropy": 0.5702577340602875,
+      "epoch": 3.6773006134969326,
+      "grad_norm": 0.8090758919715881,
+      "learning_rate": 0.0003218776372121157,
+      "loss": 0.5466982650756836,
+      "mean_token_accuracy": 0.8186432421207428,
+      "num_tokens": 3967273.0,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5708746567368508,
+      "epoch": 3.8,
+      "grad_norm": 0.7777291536331177,
+      "learning_rate": 0.000314832612625101,
+      "loss": 0.5412663650512696,
+      "mean_token_accuracy": 0.8195344746112824,
+      "num_tokens": 4102434.0,
+      "step": 1550
+    },
+    {
+      "entropy": 0.578739655315876,
+      "epoch": 3.9226993865030675,
+      "grad_norm": 0.8103565573692322,
+      "learning_rate": 0.0003075807628056167,
+      "loss": 0.5605202865600586,
+      "mean_token_accuracy": 0.815134603381157,
+      "num_tokens": 4234037.0,
+      "step": 1600
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.6183825404303415,
+      "eval_mean_token_accuracy": 0.7838981601170131,
+      "eval_not_syn_loss": 0.6814553737640381,
+      "eval_not_syn_runtime": 56.4858,
+      "eval_not_syn_samples_per_second": 24.714,
+      "eval_not_syn_steps_per_second": 3.098,
+      "eval_num_tokens": 4320784.0,
+      "step": 1632
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5816301565510886,
+      "eval_mean_token_accuracy": 0.8187636685371399,
+      "eval_num_tokens": 4320784.0,
+      "eval_syn_loss": 0.6366032361984253,
+      "eval_syn_runtime": 58.1317,
+      "eval_syn_samples_per_second": 24.014,
+      "eval_syn_steps_per_second": 3.01,
+      "step": 1632
+    },
+    {
+      "entropy": 0.5312421954039371,
+      "epoch": 4.044171779141104,
+      "grad_norm": 0.6530773043632507,
+      "learning_rate": 0.00030013535610559226,
+      "loss": 0.5023544311523438,
+      "mean_token_accuracy": 0.8297024351177793,
+      "num_tokens": 4368268.0,
+      "step": 1650
+    },
+    {
+      "entropy": 0.4853435277938843,
+      "epoch": 4.166871165644172,
+      "grad_norm": 1.0383695363998413,
+      "learning_rate": 0.00029251001501843485,
+      "loss": 0.45016471862792967,
+      "mean_token_accuracy": 0.8418685424327851,
+      "num_tokens": 4494525.0,
+      "step": 1700
+    },
+    {
+      "entropy": 0.48153585255146025,
+      "epoch": 4.289570552147239,
+      "grad_norm": 0.7215332388877869,
+      "learning_rate": 0.00028471869125462477,
+      "loss": 0.4460752487182617,
+      "mean_token_accuracy": 0.8421206694841384,
+      "num_tokens": 4631985.0,
+      "step": 1750
+    },
+    {
+      "entropy": 0.49278041124343874,
+      "epoch": 4.412269938650307,
+      "grad_norm": 0.7771472334861755,
+      "learning_rate": 0.0002767756402149588,
+      "loss": 0.45794551849365234,
+      "mean_token_accuracy": 0.8404830145835877,
+      "num_tokens": 4766323.0,
+      "step": 1800
+    },
+    {
+      "entropy": 0.5082326257228851,
+      "epoch": 4.534969325153375,
+      "grad_norm": 0.9560676217079163,
+      "learning_rate": 0.00026869539490814704,
+      "loss": 0.4632451629638672,
+      "mean_token_accuracy": 0.8386451864242553,
+      "num_tokens": 4895606.0,
+      "step": 1850
+    },
+    {
+      "entropy": 0.5081156292557716,
+      "epoch": 4.6576687116564415,
+      "grad_norm": 0.7719491720199585,
+      "learning_rate": 0.0002604927393604828,
+      "loss": 0.4630893325805664,
+      "mean_token_accuracy": 0.839306333065033,
+      "num_tokens": 5025912.0,
+      "step": 1900
+    },
+    {
+      "entropy": 0.5121548187732696,
+      "epoch": 4.780368098159509,
+      "grad_norm": 0.8171904683113098,
+      "learning_rate": 0.00025218268156623985,
+      "loss": 0.46541053771972657,
+      "mean_token_accuracy": 0.8369353520870209,
+      "num_tokens": 5162972.0,
+      "step": 1950
+    },
+    {
+      "entropy": 0.5085560208559037,
+      "epoch": 4.903067484662577,
+      "grad_norm": 0.5919133424758911,
+      "learning_rate": 0.00024378042602828492,
+      "loss": 0.46405506134033203,
+      "mean_token_accuracy": 0.8379521882534027,
+      "num_tokens": 5297935.0,
+      "step": 2000
+    },
+    {
+      "epoch": 5.0,
+      "eval_entropy": 0.5431888212476458,
+      "eval_mean_token_accuracy": 0.7749751152311053,
+      "eval_not_syn_loss": 0.7035888433456421,
+      "eval_not_syn_runtime": 56.5207,
+      "eval_not_syn_samples_per_second": 24.699,
+      "eval_not_syn_steps_per_second": 3.096,
+      "eval_num_tokens": 5400980.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.0,
+      "eval_entropy": 0.5098190077713558,
+      "eval_mean_token_accuracy": 0.832349077633449,
+      "eval_num_tokens": 5400980.0,
+      "eval_syn_loss": 0.6227898597717285,
+      "eval_syn_runtime": 57.6978,
+      "eval_syn_samples_per_second": 24.195,
+      "eval_syn_steps_per_second": 3.033,
+      "step": 2040
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 4080,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.1014210793408448e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2448/README.md b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2448/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2448/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2448/adapter_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2448/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..100cd582d71fb39c74fbba0e56e6d3415622454e
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2448/adapter_config.json
@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.05635041589438875,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "gate_proj",
+    "up_proj",
+    "o_proj",
+    "down_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2448/chat_template.jinja b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2448/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2448/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is true %}
+        {{- '<think>\n' }}
+    {%- else %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2448/tokenizer_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2448/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2448/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2448/trainer_state.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2448/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..bd633ec5467130b9f20492716e1001f23958ba49
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2448/trainer_state.json
@@ -0,0 +1,646 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.0,
+  "eval_steps": 500,
+  "global_step": 2448,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8153233402967452,
+      "epoch": 0.12269938650306748,
+      "grad_norm": 1.8510568141937256,
+      "learning_rate": 4.8469594667113335e-05,
+      "loss": 1.7217323303222656,
+      "mean_token_accuracy": 0.6236635231971741,
+      "num_tokens": 134329.0,
+      "step": 50
+    },
+    {
+      "entropy": 0.97831438601017,
+      "epoch": 0.24539877300613497,
+      "grad_norm": 1.889868974685669,
+      "learning_rate": 9.792836473559631e-05,
+      "loss": 0.9050540924072266,
+      "mean_token_accuracy": 0.7377463465929032,
+      "num_tokens": 262407.0,
+      "step": 100
+    },
+    {
+      "entropy": 0.8743131357431412,
+      "epoch": 0.36809815950920244,
+      "grad_norm": 1.1766732931137085,
+      "learning_rate": 0.0001473871348040793,
+      "loss": 0.8129763031005859,
+      "mean_token_accuracy": 0.7583417356014251,
+      "num_tokens": 401136.0,
+      "step": 150
+    },
+    {
+      "entropy": 0.8273966419696808,
+      "epoch": 0.49079754601226994,
+      "grad_norm": 1.1311014890670776,
+      "learning_rate": 0.00019684590487256229,
+      "loss": 0.7711280822753906,
+      "mean_token_accuracy": 0.7668228060007095,
+      "num_tokens": 531647.0,
+      "step": 200
+    },
+    {
+      "entropy": 0.8101438617706299,
+      "epoch": 0.6134969325153374,
+      "grad_norm": 0.8905950784683228,
+      "learning_rate": 0.0002463046749410453,
+      "loss": 0.7609222412109375,
+      "mean_token_accuracy": 0.7717174577713013,
+      "num_tokens": 666127.0,
+      "step": 250
+    },
+    {
+      "entropy": 0.7885055243968964,
+      "epoch": 0.7361963190184049,
+      "grad_norm": 1.1944245100021362,
+      "learning_rate": 0.00029576344500952824,
+      "loss": 0.7477743530273437,
+      "mean_token_accuracy": 0.773262197971344,
+      "num_tokens": 799292.0,
+      "step": 300
+    },
+    {
+      "entropy": 0.7858527797460556,
+      "epoch": 0.8588957055214724,
+      "grad_norm": 1.1311068534851074,
+      "learning_rate": 0.00034522221507801124,
+      "loss": 0.7411377716064453,
+      "mean_token_accuracy": 0.7745680212974548,
+      "num_tokens": 930614.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.7879797518253326,
+      "epoch": 0.9815950920245399,
+      "grad_norm": 0.9370436072349548,
+      "learning_rate": 0.00039468098514649423,
+      "loss": 0.7507785034179687,
+      "mean_token_accuracy": 0.7747444450855255,
+      "num_tokens": 1061654.0,
+      "step": 400
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.8527693561145238,
+      "eval_mean_token_accuracy": 0.7730878329277039,
+      "eval_not_syn_loss": 0.781921923160553,
+      "eval_not_syn_runtime": 56.9157,
+      "eval_not_syn_samples_per_second": 24.527,
+      "eval_not_syn_steps_per_second": 3.075,
+      "eval_num_tokens": 1080196.0,
+      "step": 408
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.7952005583899362,
+      "eval_mean_token_accuracy": 0.7705784467288426,
+      "eval_num_tokens": 1080196.0,
+      "eval_syn_loss": 0.7461121082305908,
+      "eval_syn_runtime": 59.5758,
+      "eval_syn_samples_per_second": 23.432,
+      "eval_syn_steps_per_second": 2.937,
+      "step": 408
+    },
+    {
+      "entropy": 0.7564875923021875,
+      "epoch": 1.1030674846625768,
+      "grad_norm": 1.1334599256515503,
+      "learning_rate": 0.00040345942966973637,
+      "loss": 0.7208955383300781,
+      "mean_token_accuracy": 0.7788144470465304,
+      "num_tokens": 1199170.0,
+      "step": 450
+    },
+    {
+      "entropy": 0.7427570706605912,
+      "epoch": 1.2257668711656442,
+      "grad_norm": 1.0777021646499634,
+      "learning_rate": 0.00040297229629224153,
+      "loss": 0.7049168395996094,
+      "mean_token_accuracy": 0.7830300116539002,
+      "num_tokens": 1334298.0,
+      "step": 500
+    },
+    {
+      "entropy": 0.7533353108167649,
+      "epoch": 1.3484662576687116,
+      "grad_norm": 1.3439561128616333,
+      "learning_rate": 0.00040211707285041045,
+      "loss": 0.7102288818359375,
+      "mean_token_accuracy": 0.7818449640274048,
+      "num_tokens": 1463189.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.7666342407464981,
+      "epoch": 1.471165644171779,
+      "grad_norm": 1.1329729557037354,
+      "learning_rate": 0.00040089532410439227,
+      "loss": 0.7270996856689453,
+      "mean_token_accuracy": 0.778404277563095,
+      "num_tokens": 1590310.0,
+      "step": 600
+    },
+    {
+      "entropy": 0.7359934949874878,
+      "epoch": 1.5938650306748468,
+      "grad_norm": 1.2227331399917603,
+      "learning_rate": 0.0003993092854276068,
+      "loss": 0.6996555328369141,
+      "mean_token_accuracy": 0.7834958267211914,
+      "num_tokens": 1721836.0,
+      "step": 650
+    },
+    {
+      "entropy": 0.7278932738304138,
+      "epoch": 1.716564417177914,
+      "grad_norm": 0.8753179311752319,
+      "learning_rate": 0.0003973618587167924,
+      "loss": 0.694002456665039,
+      "mean_token_accuracy": 0.7849677371978759,
+      "num_tokens": 1853507.0,
+      "step": 700
+    },
+    {
+      "entropy": 0.7190143716335297,
+      "epoch": 1.8392638036809816,
+      "grad_norm": 0.8524742126464844,
+      "learning_rate": 0.0003950566070825483,
+      "loss": 0.6890143585205079,
+      "mean_token_accuracy": 0.7865350896120071,
+      "num_tokens": 1988113.0,
+      "step": 750
+    },
+    {
+      "entropy": 0.7072037667036056,
+      "epoch": 1.961963190184049,
+      "grad_norm": 0.877348005771637,
+      "learning_rate": 0.00039239774833008714,
+      "loss": 0.6795162200927735,
+      "mean_token_accuracy": 0.7885233855247498,
+      "num_tokens": 2121314.0,
+      "step": 800
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.818787784235818,
+      "eval_mean_token_accuracy": 0.7573730857031686,
+      "eval_not_syn_loss": 0.7486392855644226,
+      "eval_not_syn_runtime": 56.7426,
+      "eval_not_syn_samples_per_second": 24.602,
+      "eval_not_syn_steps_per_second": 3.084,
+      "eval_num_tokens": 2160392.0,
+      "step": 816
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.7632522354807173,
+      "eval_mean_token_accuracy": 0.8101163625717163,
+      "eval_num_tokens": 2160392.0,
+      "eval_syn_loss": 0.6829464435577393,
+      "eval_syn_runtime": 58.5255,
+      "eval_syn_samples_per_second": 23.853,
+      "eval_syn_steps_per_second": 2.99,
+      "step": 816
+    },
+    {
+      "entropy": 0.6619202801675508,
+      "epoch": 2.083435582822086,
+      "grad_norm": 0.8124546408653259,
+      "learning_rate": 0.00038939014724212456,
+      "loss": 0.6311377716064454,
+      "mean_token_accuracy": 0.7991988044796567,
+      "num_tokens": 2253179.0,
+      "step": 850
+    },
+    {
+      "entropy": 0.6668815296888352,
+      "epoch": 2.2061349693251535,
+      "grad_norm": 0.8693709969520569,
+      "learning_rate": 0.0003860393066780265,
+      "loss": 0.6279015731811524,
+      "mean_token_accuracy": 0.7983763587474823,
+      "num_tokens": 2379965.0,
+      "step": 900
+    },
+    {
+      "entropy": 0.6557431703805924,
+      "epoch": 2.3288343558282207,
+      "grad_norm": 0.8323768377304077,
+      "learning_rate": 0.0003823513575054986,
+      "loss": 0.6214478683471679,
+      "mean_token_accuracy": 0.8012632429599762,
+      "num_tokens": 2515593.0,
+      "step": 950
+    },
+    {
+      "entropy": 0.6633540654182434,
+      "epoch": 2.4515337423312884,
+      "grad_norm": 0.8328827023506165,
+      "learning_rate": 0.0003783330473832399,
+      "loss": 0.6310849761962891,
+      "mean_token_accuracy": 0.7995503985881806,
+      "num_tokens": 2642790.0,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6542247200012207,
+      "epoch": 2.574233128834356,
+      "grad_norm": 0.8204748630523682,
+      "learning_rate": 0.000373991728415085,
+      "loss": 0.6195787048339844,
+      "mean_token_accuracy": 0.8006457507610321,
+      "num_tokens": 2774396.0,
+      "step": 1050
+    },
+    {
+      "entropy": 0.6526653742790223,
+      "epoch": 2.6969325153374233,
+      "grad_norm": 1.065635323524475,
+      "learning_rate": 0.0003693353436982218,
+      "loss": 0.6206568145751953,
+      "mean_token_accuracy": 0.8002620726823807,
+      "num_tokens": 2912102.0,
+      "step": 1100
+    },
+    {
+      "entropy": 0.654406590461731,
+      "epoch": 2.819631901840491,
+      "grad_norm": 0.8653210401535034,
+      "learning_rate": 0.00036437241279009834,
+      "loss": 0.6259954452514649,
+      "mean_token_accuracy": 0.8008036535978317,
+      "num_tokens": 3042987.0,
+      "step": 1150
+    },
+    {
+      "entropy": 0.6310120761394501,
+      "epoch": 2.942331288343558,
+      "grad_norm": 0.7713245749473572,
+      "learning_rate": 0.0003591120161206093,
+      "loss": 0.6115898132324219,
+      "mean_token_accuracy": 0.8043569129705429,
+      "num_tokens": 3178044.0,
+      "step": 1200
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6591688336644854,
+      "eval_mean_token_accuracy": 0.7992405581474304,
+      "eval_not_syn_loss": 0.6962469220161438,
+      "eval_not_syn_runtime": 56.161,
+      "eval_not_syn_samples_per_second": 24.857,
+      "eval_not_syn_steps_per_second": 3.116,
+      "eval_num_tokens": 3240588.0,
+      "step": 1224
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6191406740461077,
+      "eval_mean_token_accuracy": 0.7810143651281084,
+      "eval_num_tokens": 3240588.0,
+      "eval_syn_loss": 0.6663697361946106,
+      "eval_syn_runtime": 58.4102,
+      "eval_syn_samples_per_second": 23.9,
+      "eval_syn_steps_per_second": 2.996,
+      "step": 1224
+    },
+    {
+      "entropy": 0.5853322166385073,
+      "epoch": 3.063803680981595,
+      "grad_norm": 0.7313649654388428,
+      "learning_rate": 0.00035356377837808084,
+      "loss": 0.5616408538818359,
+      "mean_token_accuracy": 0.8144418639366073,
+      "num_tokens": 3312797.0,
+      "step": 1250
+    },
+    {
+      "entropy": 0.5654504173994064,
+      "epoch": 3.1865030674846624,
+      "grad_norm": 0.713893711566925,
+      "learning_rate": 0.0003477378508994529,
+      "loss": 0.5305458068847656,
+      "mean_token_accuracy": 0.8218590825796127,
+      "num_tokens": 3441629.0,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5721602493524551,
+      "epoch": 3.30920245398773,
+      "grad_norm": 0.8487221002578735,
+      "learning_rate": 0.00034164489309687927,
+      "loss": 0.5437272262573242,
+      "mean_token_accuracy": 0.8169907212257386,
+      "num_tokens": 3576093.0,
+      "step": 1350
+    },
+    {
+      "entropy": 0.5797226822376251,
+      "epoch": 3.4319018404907977,
+      "grad_norm": 0.9307994842529297,
+      "learning_rate": 0.0003352960529547267,
+      "loss": 0.5405152130126953,
+      "mean_token_accuracy": 0.818563597202301,
+      "num_tokens": 3705912.0,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5822264245152473,
+      "epoch": 3.554601226993865,
+      "grad_norm": 0.71127849817276,
+      "learning_rate": 0.00032870294663265757,
+      "loss": 0.5559784698486329,
+      "mean_token_accuracy": 0.8159428876638413,
+      "num_tokens": 3832376.0,
+      "step": 1450
+    },
+    {
+      "entropy": 0.5702577340602875,
+      "epoch": 3.6773006134969326,
+      "grad_norm": 0.8090758919715881,
+      "learning_rate": 0.0003218776372121157,
+      "loss": 0.5466982650756836,
+      "mean_token_accuracy": 0.8186432421207428,
+      "num_tokens": 3967273.0,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5708746567368508,
+      "epoch": 3.8,
+      "grad_norm": 0.7777291536331177,
+      "learning_rate": 0.000314832612625101,
+      "loss": 0.5412663650512696,
+      "mean_token_accuracy": 0.8195344746112824,
+      "num_tokens": 4102434.0,
+      "step": 1550
+    },
+    {
+      "entropy": 0.578739655315876,
+      "epoch": 3.9226993865030675,
+      "grad_norm": 0.8103565573692322,
+      "learning_rate": 0.0003075807628056167,
+      "loss": 0.5605202865600586,
+      "mean_token_accuracy": 0.815134603381157,
+      "num_tokens": 4234037.0,
+      "step": 1600
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.6183825404303415,
+      "eval_mean_token_accuracy": 0.7838981601170131,
+      "eval_not_syn_loss": 0.6814553737640381,
+      "eval_not_syn_runtime": 56.4858,
+      "eval_not_syn_samples_per_second": 24.714,
+      "eval_not_syn_steps_per_second": 3.098,
+      "eval_num_tokens": 4320784.0,
+      "step": 1632
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5816301565510886,
+      "eval_mean_token_accuracy": 0.8187636685371399,
+      "eval_num_tokens": 4320784.0,
+      "eval_syn_loss": 0.6366032361984253,
+      "eval_syn_runtime": 58.1317,
+      "eval_syn_samples_per_second": 24.014,
+      "eval_syn_steps_per_second": 3.01,
+      "step": 1632
+    },
+    {
+      "entropy": 0.5312421954039371,
+      "epoch": 4.044171779141104,
+      "grad_norm": 0.6530773043632507,
+      "learning_rate": 0.00030013535610559226,
+      "loss": 0.5023544311523438,
+      "mean_token_accuracy": 0.8297024351177793,
+      "num_tokens": 4368268.0,
+      "step": 1650
+    },
+    {
+      "entropy": 0.4853435277938843,
+      "epoch": 4.166871165644172,
+      "grad_norm": 1.0383695363998413,
+      "learning_rate": 0.00029251001501843485,
+      "loss": 0.45016471862792967,
+      "mean_token_accuracy": 0.8418685424327851,
+      "num_tokens": 4494525.0,
+      "step": 1700
+    },
+    {
+      "entropy": 0.48153585255146025,
+      "epoch": 4.289570552147239,
+      "grad_norm": 0.7215332388877869,
+      "learning_rate": 0.00028471869125462477,
+      "loss": 0.4460752487182617,
+      "mean_token_accuracy": 0.8421206694841384,
+      "num_tokens": 4631985.0,
+      "step": 1750
+    },
+    {
+      "entropy": 0.49278041124343874,
+      "epoch": 4.412269938650307,
+      "grad_norm": 0.7771472334861755,
+      "learning_rate": 0.0002767756402149588,
+      "loss": 0.45794551849365234,
+      "mean_token_accuracy": 0.8404830145835877,
+      "num_tokens": 4766323.0,
+      "step": 1800
+    },
+    {
+      "entropy": 0.5082326257228851,
+      "epoch": 4.534969325153375,
+      "grad_norm": 0.9560676217079163,
+      "learning_rate": 0.00026869539490814704,
+      "loss": 0.4632451629638672,
+      "mean_token_accuracy": 0.8386451864242553,
+      "num_tokens": 4895606.0,
+      "step": 1850
+    },
+    {
+      "entropy": 0.5081156292557716,
+      "epoch": 4.6576687116564415,
+      "grad_norm": 0.7719491720199585,
+      "learning_rate": 0.0002604927393604828,
+      "loss": 0.4630893325805664,
+      "mean_token_accuracy": 0.839306333065033,
+      "num_tokens": 5025912.0,
+      "step": 1900
+    },
+    {
+      "entropy": 0.5121548187732696,
+      "epoch": 4.780368098159509,
+      "grad_norm": 0.8171904683113098,
+      "learning_rate": 0.00025218268156623985,
+      "loss": 0.46541053771972657,
+      "mean_token_accuracy": 0.8369353520870209,
+      "num_tokens": 5162972.0,
+      "step": 1950
+    },
+    {
+      "entropy": 0.5085560208559037,
+      "epoch": 4.903067484662577,
+      "grad_norm": 0.5919133424758911,
+      "learning_rate": 0.00024378042602828492,
+      "loss": 0.46405506134033203,
+      "mean_token_accuracy": 0.8379521882534027,
+      "num_tokens": 5297935.0,
+      "step": 2000
+    },
+    {
+      "epoch": 5.0,
+      "eval_entropy": 0.5431888212476458,
+      "eval_mean_token_accuracy": 0.7749751152311053,
+      "eval_not_syn_loss": 0.7035888433456421,
+      "eval_not_syn_runtime": 56.5207,
+      "eval_not_syn_samples_per_second": 24.699,
+      "eval_not_syn_steps_per_second": 3.096,
+      "eval_num_tokens": 5400980.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.0,
+      "eval_entropy": 0.5098190077713558,
+      "eval_mean_token_accuracy": 0.832349077633449,
+      "eval_num_tokens": 5400980.0,
+      "eval_syn_loss": 0.6227898597717285,
+      "eval_syn_runtime": 57.6978,
+      "eval_syn_samples_per_second": 24.195,
+      "eval_syn_steps_per_second": 3.033,
+      "step": 2040
+    },
+    {
+      "entropy": 0.4960306563762703,
+      "epoch": 5.024539877300613,
+      "grad_norm": 0.9411171078681946,
+      "learning_rate": 0.0002353013459391488,
+      "loss": 0.4484004211425781,
+      "mean_token_accuracy": 0.8438825571175778,
+      "num_tokens": 5427362.0,
+      "step": 2050
+    },
+    {
+      "entropy": 0.3916636416316032,
+      "epoch": 5.147239263803681,
+      "grad_norm": 0.8929405808448792,
+      "learning_rate": 0.00022676095505345462,
+      "loss": 0.3446478271484375,
+      "mean_token_accuracy": 0.8732858788967133,
+      "num_tokens": 5564186.0,
+      "step": 2100
+    },
+    {
+      "entropy": 0.39636621803045274,
+      "epoch": 5.269938650306749,
+      "grad_norm": 0.69569331407547,
+      "learning_rate": 0.0002181748793031656,
+      "loss": 0.3548049163818359,
+      "mean_token_accuracy": 0.8699262255430221,
+      "num_tokens": 5694728.0,
+      "step": 2150
+    },
+    {
+      "entropy": 0.4028259950876236,
+      "epoch": 5.392638036809816,
+      "grad_norm": 0.7605823874473572,
+      "learning_rate": 0.00020955882820758868,
+      "loss": 0.35218334197998047,
+      "mean_token_accuracy": 0.8704028391838073,
+      "num_tokens": 5832581.0,
+      "step": 2200
+    },
+    {
+      "entropy": 0.4091260200738907,
+      "epoch": 5.515337423312883,
+      "grad_norm": 0.8990920782089233,
+      "learning_rate": 0.00020092856613044126,
+      "loss": 0.36372703552246094,
+      "mean_token_accuracy": 0.8663889598846436,
+      "num_tokens": 5962417.0,
+      "step": 2250
+    },
+    {
+      "entropy": 0.4057882487773895,
+      "epoch": 5.638036809815951,
+      "grad_norm": 0.9872731566429138,
+      "learning_rate": 0.0001922998834365732,
+      "loss": 0.3588364791870117,
+      "mean_token_accuracy": 0.8686526268720627,
+      "num_tokens": 6092338.0,
+      "step": 2300
+    },
+    {
+      "entropy": 0.41128933161497117,
+      "epoch": 5.7607361963190185,
+      "grad_norm": 0.8626382350921631,
+      "learning_rate": 0.0001836885676011143,
+      "loss": 0.36508998870849607,
+      "mean_token_accuracy": 0.8664334756135941,
+      "num_tokens": 6226327.0,
+      "step": 2350
+    },
+    {
+      "entropy": 0.40721403241157533,
+      "epoch": 5.883435582822086,
+      "grad_norm": 0.775476336479187,
+      "learning_rate": 0.00017511037432391027,
+      "loss": 0.36363380432128906,
+      "mean_token_accuracy": 0.8656364333629608,
+      "num_tokens": 6359083.0,
+      "step": 2400
+    },
+    {
+      "epoch": 6.0,
+      "eval_entropy": 0.4985327124595642,
+      "eval_mean_token_accuracy": 0.8019610796655927,
+      "eval_not_syn_loss": 0.717173159122467,
+      "eval_not_syn_runtime": 56.4169,
+      "eval_not_syn_samples_per_second": 24.744,
+      "eval_not_syn_steps_per_second": 3.102,
+      "eval_num_tokens": 6481176.0,
+      "step": 2448
+    },
+    {
+      "epoch": 6.0,
+      "eval_entropy": 0.4651233986445836,
+      "eval_mean_token_accuracy": 0.8059938202585493,
+      "eval_num_tokens": 6481176.0,
+      "eval_syn_loss": 0.6778863668441772,
+      "eval_syn_runtime": 58.5069,
+      "eval_syn_samples_per_second": 23.86,
+      "eval_syn_steps_per_second": 2.991,
+      "step": 2448
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 4080,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.321195971768434e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2856/README.md b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2856/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2856/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2856/adapter_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2856/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..100cd582d71fb39c74fbba0e56e6d3415622454e
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2856/adapter_config.json
@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.05635041589438875,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "gate_proj",
+    "up_proj",
+    "o_proj",
+    "down_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2856/chat_template.jinja b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2856/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2856/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is true %}
+        {{- '<think>\n' }}
+    {%- else %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2856/tokenizer_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2856/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2856/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2856/trainer_state.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2856/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..83c225fb8c90fe3f001ba06eed7a04abb105ff31
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-2856/trainer_state.json
@@ -0,0 +1,758 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.0,
+  "eval_steps": 500,
+  "global_step": 2856,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8153233402967452,
+      "epoch": 0.12269938650306748,
+      "grad_norm": 1.8510568141937256,
+      "learning_rate": 4.8469594667113335e-05,
+      "loss": 1.7217323303222656,
+      "mean_token_accuracy": 0.6236635231971741,
+      "num_tokens": 134329.0,
+      "step": 50
+    },
+    {
+      "entropy": 0.97831438601017,
+      "epoch": 0.24539877300613497,
+      "grad_norm": 1.889868974685669,
+      "learning_rate": 9.792836473559631e-05,
+      "loss": 0.9050540924072266,
+      "mean_token_accuracy": 0.7377463465929032,
+      "num_tokens": 262407.0,
+      "step": 100
+    },
+    {
+      "entropy": 0.8743131357431412,
+      "epoch": 0.36809815950920244,
+      "grad_norm": 1.1766732931137085,
+      "learning_rate": 0.0001473871348040793,
+      "loss": 0.8129763031005859,
+      "mean_token_accuracy": 0.7583417356014251,
+      "num_tokens": 401136.0,
+      "step": 150
+    },
+    {
+      "entropy": 0.8273966419696808,
+      "epoch": 0.49079754601226994,
+      "grad_norm": 1.1311014890670776,
+      "learning_rate": 0.00019684590487256229,
+      "loss": 0.7711280822753906,
+      "mean_token_accuracy": 0.7668228060007095,
+      "num_tokens": 531647.0,
+      "step": 200
+    },
+    {
+      "entropy": 0.8101438617706299,
+      "epoch": 0.6134969325153374,
+      "grad_norm": 0.8905950784683228,
+      "learning_rate": 0.0002463046749410453,
+      "loss": 0.7609222412109375,
+      "mean_token_accuracy": 0.7717174577713013,
+      "num_tokens": 666127.0,
+      "step": 250
+    },
+    {
+      "entropy": 0.7885055243968964,
+      "epoch": 0.7361963190184049,
+      "grad_norm": 1.1944245100021362,
+      "learning_rate": 0.00029576344500952824,
+      "loss": 0.7477743530273437,
+      "mean_token_accuracy": 0.773262197971344,
+      "num_tokens": 799292.0,
+      "step": 300
+    },
+    {
+      "entropy": 0.7858527797460556,
+      "epoch": 0.8588957055214724,
+      "grad_norm": 1.1311068534851074,
+      "learning_rate": 0.00034522221507801124,
+      "loss": 0.7411377716064453,
+      "mean_token_accuracy": 0.7745680212974548,
+      "num_tokens": 930614.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.7879797518253326,
+      "epoch": 0.9815950920245399,
+      "grad_norm": 0.9370436072349548,
+      "learning_rate": 0.00039468098514649423,
+      "loss": 0.7507785034179687,
+      "mean_token_accuracy": 0.7747444450855255,
+      "num_tokens": 1061654.0,
+      "step": 400
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.8527693561145238,
+      "eval_mean_token_accuracy": 0.7730878329277039,
+      "eval_not_syn_loss": 0.781921923160553,
+      "eval_not_syn_runtime": 56.9157,
+      "eval_not_syn_samples_per_second": 24.527,
+      "eval_not_syn_steps_per_second": 3.075,
+      "eval_num_tokens": 1080196.0,
+      "step": 408
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.7952005583899362,
+      "eval_mean_token_accuracy": 0.7705784467288426,
+      "eval_num_tokens": 1080196.0,
+      "eval_syn_loss": 0.7461121082305908,
+      "eval_syn_runtime": 59.5758,
+      "eval_syn_samples_per_second": 23.432,
+      "eval_syn_steps_per_second": 2.937,
+      "step": 408
+    },
+    {
+      "entropy": 0.7564875923021875,
+      "epoch": 1.1030674846625768,
+      "grad_norm": 1.1334599256515503,
+      "learning_rate": 0.00040345942966973637,
+      "loss": 0.7208955383300781,
+      "mean_token_accuracy": 0.7788144470465304,
+      "num_tokens": 1199170.0,
+      "step": 450
+    },
+    {
+      "entropy": 0.7427570706605912,
+      "epoch": 1.2257668711656442,
+      "grad_norm": 1.0777021646499634,
+      "learning_rate": 0.00040297229629224153,
+      "loss": 0.7049168395996094,
+      "mean_token_accuracy": 0.7830300116539002,
+      "num_tokens": 1334298.0,
+      "step": 500
+    },
+    {
+      "entropy": 0.7533353108167649,
+      "epoch": 1.3484662576687116,
+      "grad_norm": 1.3439561128616333,
+      "learning_rate": 0.00040211707285041045,
+      "loss": 0.7102288818359375,
+      "mean_token_accuracy": 0.7818449640274048,
+      "num_tokens": 1463189.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.7666342407464981,
+      "epoch": 1.471165644171779,
+      "grad_norm": 1.1329729557037354,
+      "learning_rate": 0.00040089532410439227,
+      "loss": 0.7270996856689453,
+      "mean_token_accuracy": 0.778404277563095,
+      "num_tokens": 1590310.0,
+      "step": 600
+    },
+    {
+      "entropy": 0.7359934949874878,
+      "epoch": 1.5938650306748468,
+      "grad_norm": 1.2227331399917603,
+      "learning_rate": 0.0003993092854276068,
+      "loss": 0.6996555328369141,
+      "mean_token_accuracy": 0.7834958267211914,
+      "num_tokens": 1721836.0,
+      "step": 650
+    },
+    {
+      "entropy": 0.7278932738304138,
+      "epoch": 1.716564417177914,
+      "grad_norm": 0.8753179311752319,
+      "learning_rate": 0.0003973618587167924,
+      "loss": 0.694002456665039,
+      "mean_token_accuracy": 0.7849677371978759,
+      "num_tokens": 1853507.0,
+      "step": 700
+    },
+    {
+      "entropy": 0.7190143716335297,
+      "epoch": 1.8392638036809816,
+      "grad_norm": 0.8524742126464844,
+      "learning_rate": 0.0003950566070825483,
+      "loss": 0.6890143585205079,
+      "mean_token_accuracy": 0.7865350896120071,
+      "num_tokens": 1988113.0,
+      "step": 750
+    },
+    {
+      "entropy": 0.7072037667036056,
+      "epoch": 1.961963190184049,
+      "grad_norm": 0.877348005771637,
+      "learning_rate": 0.00039239774833008714,
+      "loss": 0.6795162200927735,
+      "mean_token_accuracy": 0.7885233855247498,
+      "num_tokens": 2121314.0,
+      "step": 800
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.818787784235818,
+      "eval_mean_token_accuracy": 0.7573730857031686,
+      "eval_not_syn_loss": 0.7486392855644226,
+      "eval_not_syn_runtime": 56.7426,
+      "eval_not_syn_samples_per_second": 24.602,
+      "eval_not_syn_steps_per_second": 3.084,
+      "eval_num_tokens": 2160392.0,
+      "step": 816
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.7632522354807173,
+      "eval_mean_token_accuracy": 0.8101163625717163,
+      "eval_num_tokens": 2160392.0,
+      "eval_syn_loss": 0.6829464435577393,
+      "eval_syn_runtime": 58.5255,
+      "eval_syn_samples_per_second": 23.853,
+      "eval_syn_steps_per_second": 2.99,
+      "step": 816
+    },
+    {
+      "entropy": 0.6619202801675508,
+      "epoch": 2.083435582822086,
+      "grad_norm": 0.8124546408653259,
+      "learning_rate": 0.00038939014724212456,
+      "loss": 0.6311377716064454,
+      "mean_token_accuracy": 0.7991988044796567,
+      "num_tokens": 2253179.0,
+      "step": 850
+    },
+    {
+      "entropy": 0.6668815296888352,
+      "epoch": 2.2061349693251535,
+      "grad_norm": 0.8693709969520569,
+      "learning_rate": 0.0003860393066780265,
+      "loss": 0.6279015731811524,
+      "mean_token_accuracy": 0.7983763587474823,
+      "num_tokens": 2379965.0,
+      "step": 900
+    },
+    {
+      "entropy": 0.6557431703805924,
+      "epoch": 2.3288343558282207,
+      "grad_norm": 0.8323768377304077,
+      "learning_rate": 0.0003823513575054986,
+      "loss": 0.6214478683471679,
+      "mean_token_accuracy": 0.8012632429599762,
+      "num_tokens": 2515593.0,
+      "step": 950
+    },
+    {
+      "entropy": 0.6633540654182434,
+      "epoch": 2.4515337423312884,
+      "grad_norm": 0.8328827023506165,
+      "learning_rate": 0.0003783330473832399,
+      "loss": 0.6310849761962891,
+      "mean_token_accuracy": 0.7995503985881806,
+      "num_tokens": 2642790.0,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6542247200012207,
+      "epoch": 2.574233128834356,
+      "grad_norm": 0.8204748630523682,
+      "learning_rate": 0.000373991728415085,
+      "loss": 0.6195787048339844,
+      "mean_token_accuracy": 0.8006457507610321,
+      "num_tokens": 2774396.0,
+      "step": 1050
+    },
+    {
+      "entropy": 0.6526653742790223,
+      "epoch": 2.6969325153374233,
+      "grad_norm": 1.065635323524475,
+      "learning_rate": 0.0003693353436982218,
+      "loss": 0.6206568145751953,
+      "mean_token_accuracy": 0.8002620726823807,
+      "num_tokens": 2912102.0,
+      "step": 1100
+    },
+    {
+      "entropy": 0.654406590461731,
+      "epoch": 2.819631901840491,
+      "grad_norm": 0.8653210401535034,
+      "learning_rate": 0.00036437241279009834,
+      "loss": 0.6259954452514649,
+      "mean_token_accuracy": 0.8008036535978317,
+      "num_tokens": 3042987.0,
+      "step": 1150
+    },
+    {
+      "entropy": 0.6310120761394501,
+      "epoch": 2.942331288343558,
+      "grad_norm": 0.7713245749473572,
+      "learning_rate": 0.0003591120161206093,
+      "loss": 0.6115898132324219,
+      "mean_token_accuracy": 0.8043569129705429,
+      "num_tokens": 3178044.0,
+      "step": 1200
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6591688336644854,
+      "eval_mean_token_accuracy": 0.7992405581474304,
+      "eval_not_syn_loss": 0.6962469220161438,
+      "eval_not_syn_runtime": 56.161,
+      "eval_not_syn_samples_per_second": 24.857,
+      "eval_not_syn_steps_per_second": 3.116,
+      "eval_num_tokens": 3240588.0,
+      "step": 1224
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6191406740461077,
+      "eval_mean_token_accuracy": 0.7810143651281084,
+      "eval_num_tokens": 3240588.0,
+      "eval_syn_loss": 0.6663697361946106,
+      "eval_syn_runtime": 58.4102,
+      "eval_syn_samples_per_second": 23.9,
+      "eval_syn_steps_per_second": 2.996,
+      "step": 1224
+    },
+    {
+      "entropy": 0.5853322166385073,
+      "epoch": 3.063803680981595,
+      "grad_norm": 0.7313649654388428,
+      "learning_rate": 0.00035356377837808084,
+      "loss": 0.5616408538818359,
+      "mean_token_accuracy": 0.8144418639366073,
+      "num_tokens": 3312797.0,
+      "step": 1250
+    },
+    {
+      "entropy": 0.5654504173994064,
+      "epoch": 3.1865030674846624,
+      "grad_norm": 0.713893711566925,
+      "learning_rate": 0.0003477378508994529,
+      "loss": 0.5305458068847656,
+      "mean_token_accuracy": 0.8218590825796127,
+      "num_tokens": 3441629.0,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5721602493524551,
+      "epoch": 3.30920245398773,
+      "grad_norm": 0.8487221002578735,
+      "learning_rate": 0.00034164489309687927,
+      "loss": 0.5437272262573242,
+      "mean_token_accuracy": 0.8169907212257386,
+      "num_tokens": 3576093.0,
+      "step": 1350
+    },
+    {
+      "entropy": 0.5797226822376251,
+      "epoch": 3.4319018404907977,
+      "grad_norm": 0.9307994842529297,
+      "learning_rate": 0.0003352960529547267,
+      "loss": 0.5405152130126953,
+      "mean_token_accuracy": 0.818563597202301,
+      "num_tokens": 3705912.0,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5822264245152473,
+      "epoch": 3.554601226993865,
+      "grad_norm": 0.71127849817276,
+      "learning_rate": 0.00032870294663265757,
+      "loss": 0.5559784698486329,
+      "mean_token_accuracy": 0.8159428876638413,
+      "num_tokens": 3832376.0,
+      "step": 1450
+    },
+    {
+      "entropy": 0.5702577340602875,
+      "epoch": 3.6773006134969326,
+      "grad_norm": 0.8090758919715881,
+      "learning_rate": 0.0003218776372121157,
+      "loss": 0.5466982650756836,
+      "mean_token_accuracy": 0.8186432421207428,
+      "num_tokens": 3967273.0,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5708746567368508,
+      "epoch": 3.8,
+      "grad_norm": 0.7777291536331177,
+      "learning_rate": 0.000314832612625101,
+      "loss": 0.5412663650512696,
+      "mean_token_accuracy": 0.8195344746112824,
+      "num_tokens": 4102434.0,
+      "step": 1550
+    },
+    {
+      "entropy": 0.578739655315876,
+      "epoch": 3.9226993865030675,
+      "grad_norm": 0.8103565573692322,
+      "learning_rate": 0.0003075807628056167,
+      "loss": 0.5605202865600586,
+      "mean_token_accuracy": 0.815134603381157,
+      "num_tokens": 4234037.0,
+      "step": 1600
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.6183825404303415,
+      "eval_mean_token_accuracy": 0.7838981601170131,
+      "eval_not_syn_loss": 0.6814553737640381,
+      "eval_not_syn_runtime": 56.4858,
+      "eval_not_syn_samples_per_second": 24.714,
+      "eval_not_syn_steps_per_second": 3.098,
+      "eval_num_tokens": 4320784.0,
+      "step": 1632
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5816301565510886,
+      "eval_mean_token_accuracy": 0.8187636685371399,
+      "eval_num_tokens": 4320784.0,
+      "eval_syn_loss": 0.6366032361984253,
+      "eval_syn_runtime": 58.1317,
+      "eval_syn_samples_per_second": 24.014,
+      "eval_syn_steps_per_second": 3.01,
+      "step": 1632
+    },
+    {
+      "entropy": 0.5312421954039371,
+      "epoch": 4.044171779141104,
+      "grad_norm": 0.6530773043632507,
+      "learning_rate": 0.00030013535610559226,
+      "loss": 0.5023544311523438,
+      "mean_token_accuracy": 0.8297024351177793,
+      "num_tokens": 4368268.0,
+      "step": 1650
+    },
+    {
+      "entropy": 0.4853435277938843,
+      "epoch": 4.166871165644172,
+      "grad_norm": 1.0383695363998413,
+      "learning_rate": 0.00029251001501843485,
+      "loss": 0.45016471862792967,
+      "mean_token_accuracy": 0.8418685424327851,
+      "num_tokens": 4494525.0,
+      "step": 1700
+    },
+    {
+      "entropy": 0.48153585255146025,
+      "epoch": 4.289570552147239,
+      "grad_norm": 0.7215332388877869,
+      "learning_rate": 0.00028471869125462477,
+      "loss": 0.4460752487182617,
+      "mean_token_accuracy": 0.8421206694841384,
+      "num_tokens": 4631985.0,
+      "step": 1750
+    },
+    {
+      "entropy": 0.49278041124343874,
+      "epoch": 4.412269938650307,
+      "grad_norm": 0.7771472334861755,
+      "learning_rate": 0.0002767756402149588,
+      "loss": 0.45794551849365234,
+      "mean_token_accuracy": 0.8404830145835877,
+      "num_tokens": 4766323.0,
+      "step": 1800
+    },
+    {
+      "entropy": 0.5082326257228851,
+      "epoch": 4.534969325153375,
+      "grad_norm": 0.9560676217079163,
+      "learning_rate": 0.00026869539490814704,
+      "loss": 0.4632451629638672,
+      "mean_token_accuracy": 0.8386451864242553,
+      "num_tokens": 4895606.0,
+      "step": 1850
+    },
+    {
+      "entropy": 0.5081156292557716,
+      "epoch": 4.6576687116564415,
+      "grad_norm": 0.7719491720199585,
+      "learning_rate": 0.0002604927393604828,
+      "loss": 0.4630893325805664,
+      "mean_token_accuracy": 0.839306333065033,
+      "num_tokens": 5025912.0,
+      "step": 1900
+    },
+    {
+      "entropy": 0.5121548187732696,
+      "epoch": 4.780368098159509,
+      "grad_norm": 0.8171904683113098,
+      "learning_rate": 0.00025218268156623985,
+      "loss": 0.46541053771972657,
+      "mean_token_accuracy": 0.8369353520870209,
+      "num_tokens": 5162972.0,
+      "step": 1950
+    },
+    {
+      "entropy": 0.5085560208559037,
+      "epoch": 4.903067484662577,
+      "grad_norm": 0.5919133424758911,
+      "learning_rate": 0.00024378042602828492,
+      "loss": 0.46405506134033203,
+      "mean_token_accuracy": 0.8379521882534027,
+      "num_tokens": 5297935.0,
+      "step": 2000
+    },
+    {
+      "epoch": 5.0,
+      "eval_entropy": 0.5431888212476458,
+      "eval_mean_token_accuracy": 0.7749751152311053,
+      "eval_not_syn_loss": 0.7035888433456421,
+      "eval_not_syn_runtime": 56.5207,
+      "eval_not_syn_samples_per_second": 24.699,
+      "eval_not_syn_steps_per_second": 3.096,
+      "eval_num_tokens": 5400980.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.0,
+      "eval_entropy": 0.5098190077713558,
+      "eval_mean_token_accuracy": 0.832349077633449,
+      "eval_num_tokens": 5400980.0,
+      "eval_syn_loss": 0.6227898597717285,
+      "eval_syn_runtime": 57.6978,
+      "eval_syn_samples_per_second": 24.195,
+      "eval_syn_steps_per_second": 3.033,
+      "step": 2040
+    },
+    {
+      "entropy": 0.4960306563762703,
+      "epoch": 5.024539877300613,
+      "grad_norm": 0.9411171078681946,
+      "learning_rate": 0.0002353013459391488,
+      "loss": 0.4484004211425781,
+      "mean_token_accuracy": 0.8438825571175778,
+      "num_tokens": 5427362.0,
+      "step": 2050
+    },
+    {
+      "entropy": 0.3916636416316032,
+      "epoch": 5.147239263803681,
+      "grad_norm": 0.8929405808448792,
+      "learning_rate": 0.00022676095505345462,
+      "loss": 0.3446478271484375,
+      "mean_token_accuracy": 0.8732858788967133,
+      "num_tokens": 5564186.0,
+      "step": 2100
+    },
+    {
+      "entropy": 0.39636621803045274,
+      "epoch": 5.269938650306749,
+      "grad_norm": 0.69569331407547,
+      "learning_rate": 0.0002181748793031656,
+      "loss": 0.3548049163818359,
+      "mean_token_accuracy": 0.8699262255430221,
+      "num_tokens": 5694728.0,
+      "step": 2150
+    },
+    {
+      "entropy": 0.4028259950876236,
+      "epoch": 5.392638036809816,
+      "grad_norm": 0.7605823874473572,
+      "learning_rate": 0.00020955882820758868,
+      "loss": 0.35218334197998047,
+      "mean_token_accuracy": 0.8704028391838073,
+      "num_tokens": 5832581.0,
+      "step": 2200
+    },
+    {
+      "entropy": 0.4091260200738907,
+      "epoch": 5.515337423312883,
+      "grad_norm": 0.8990920782089233,
+      "learning_rate": 0.00020092856613044126,
+      "loss": 0.36372703552246094,
+      "mean_token_accuracy": 0.8663889598846436,
+      "num_tokens": 5962417.0,
+      "step": 2250
+    },
+    {
+      "entropy": 0.4057882487773895,
+      "epoch": 5.638036809815951,
+      "grad_norm": 0.9872731566429138,
+      "learning_rate": 0.0001922998834365732,
+      "loss": 0.3588364791870117,
+      "mean_token_accuracy": 0.8686526268720627,
+      "num_tokens": 6092338.0,
+      "step": 2300
+    },
+    {
+      "entropy": 0.41128933161497117,
+      "epoch": 5.7607361963190185,
+      "grad_norm": 0.8626382350921631,
+      "learning_rate": 0.0001836885676011143,
+      "loss": 0.36508998870849607,
+      "mean_token_accuracy": 0.8664334756135941,
+      "num_tokens": 6226327.0,
+      "step": 2350
+    },
+    {
+      "entropy": 0.40721403241157533,
+      "epoch": 5.883435582822086,
+      "grad_norm": 0.775476336479187,
+      "learning_rate": 0.00017511037432391027,
+      "loss": 0.36363380432128906,
+      "mean_token_accuracy": 0.8656364333629608,
+      "num_tokens": 6359083.0,
+      "step": 2400
+    },
+    {
+      "epoch": 6.0,
+      "eval_entropy": 0.4985327124595642,
+      "eval_mean_token_accuracy": 0.8019610796655927,
+      "eval_not_syn_loss": 0.717173159122467,
+      "eval_not_syn_runtime": 56.4169,
+      "eval_not_syn_samples_per_second": 24.744,
+      "eval_not_syn_steps_per_second": 3.102,
+      "eval_num_tokens": 6481176.0,
+      "step": 2448
+    },
+    {
+      "epoch": 6.0,
+      "eval_entropy": 0.4651233986445836,
+      "eval_mean_token_accuracy": 0.8059938202585493,
+      "eval_num_tokens": 6481176.0,
+      "eval_syn_loss": 0.6778863668441772,
+      "eval_syn_runtime": 58.5069,
+      "eval_syn_samples_per_second": 23.86,
+      "eval_syn_steps_per_second": 2.991,
+      "step": 2448
+    },
+    {
+      "entropy": 0.40582828509687174,
+      "epoch": 6.004907975460123,
+      "grad_norm": 0.6788471341133118,
+      "learning_rate": 0.0001665809987020958,
+      "loss": 0.3566559600830078,
+      "mean_token_accuracy": 0.868294438930473,
+      "num_tokens": 6486549.0,
+      "step": 2450
+    },
+    {
+      "entropy": 0.2933592677116394,
+      "epoch": 6.12760736196319,
+      "grad_norm": 0.6598446369171143,
+      "learning_rate": 0.00015811604651354912,
+      "loss": 0.24192693710327148,
+      "mean_token_accuracy": 0.9082661873102188,
+      "num_tokens": 6622384.0,
+      "step": 2500
+    },
+    {
+      "entropy": 0.2975004023313522,
+      "epoch": 6.250306748466258,
+      "grad_norm": 0.7413120865821838,
+      "learning_rate": 0.00014973100566377046,
+      "loss": 0.2502699661254883,
+      "mean_token_accuracy": 0.9056042343378067,
+      "num_tokens": 6758637.0,
+      "step": 2550
+    },
+    {
+      "entropy": 0.3052474121749401,
+      "epoch": 6.373006134969325,
+      "grad_norm": 0.8629281520843506,
+      "learning_rate": 0.00014144121784842467,
+      "loss": 0.2543815040588379,
+      "mean_token_accuracy": 0.9046169209480286,
+      "num_tokens": 6888533.0,
+      "step": 2600
+    },
+    {
+      "entropy": 0.303673956990242,
+      "epoch": 6.495705521472392,
+      "grad_norm": 0.8680911660194397,
+      "learning_rate": 0.000133261850483398,
+      "loss": 0.2529561424255371,
+      "mean_token_accuracy": 0.9040863698720932,
+      "num_tokens": 7023729.0,
+      "step": 2650
+    },
+    {
+      "entropy": 0.2993183335661888,
+      "epoch": 6.61840490797546,
+      "grad_norm": 0.7648728489875793,
+      "learning_rate": 0.00012520786895372493,
+      "loss": 0.2517524528503418,
+      "mean_token_accuracy": 0.9038773095607757,
+      "num_tokens": 7153821.0,
+      "step": 2700
+    },
+    {
+      "entropy": 0.3006903246045113,
+      "epoch": 6.741104294478528,
+      "grad_norm": 0.7433948516845703,
+      "learning_rate": 0.00011729400923216141,
+      "loss": 0.25459712982177735,
+      "mean_token_accuracy": 0.9037652736902237,
+      "num_tokens": 7283513.0,
+      "step": 2750
+    },
+    {
+      "entropy": 0.30416361182928087,
+      "epoch": 6.863803680981595,
+      "grad_norm": 0.8744818568229675,
+      "learning_rate": 0.00010953475091750244,
+      "loss": 0.25621614456176756,
+      "mean_token_accuracy": 0.9019701254367828,
+      "num_tokens": 7411578.0,
+      "step": 2800
+    },
+    {
+      "entropy": 0.2956543755531311,
+      "epoch": 6.986503067484662,
+      "grad_norm": 0.8242411017417908,
+      "learning_rate": 0.00010194429074197415,
+      "loss": 0.24829059600830078,
+      "mean_token_accuracy": 0.9054388856887817,
+      "num_tokens": 7548356.0,
+      "step": 2850
+    },
+    {
+      "epoch": 7.0,
+      "eval_entropy": 0.4070556444781167,
+      "eval_mean_token_accuracy": 0.7848251972879682,
+      "eval_not_syn_loss": 0.8054981827735901,
+      "eval_not_syn_runtime": 56.6822,
+      "eval_not_syn_samples_per_second": 24.629,
+      "eval_not_syn_steps_per_second": 3.087,
+      "eval_num_tokens": 7561372.0,
+      "step": 2856
+    },
+    {
+      "epoch": 7.0,
+      "eval_entropy": 0.3784746325016022,
+      "eval_mean_token_accuracy": 0.8244705755370004,
+      "eval_num_tokens": 7561372.0,
+      "eval_syn_loss": 0.7377892732620239,
+      "eval_syn_runtime": 58.6534,
+      "eval_syn_samples_per_second": 23.801,
+      "eval_syn_steps_per_second": 2.984,
+      "step": 2856
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 4080,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.5400107074712845e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3264/README.md b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3264/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3264/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3264/adapter_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3264/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..100cd582d71fb39c74fbba0e56e6d3415622454e
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3264/adapter_config.json
@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.05635041589438875,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "gate_proj",
+    "up_proj",
+    "o_proj",
+    "down_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3264/chat_template.jinja b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3264/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3264/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is true %}
+        {{- '<think>\n' }}
+    {%- else %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3264/tokenizer_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3264/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3264/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3264/trainer_state.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3264/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e3d680e4a4134f7178937f809d8ae398d9d4293d
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3264/trainer_state.json
@@ -0,0 +1,860 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.0,
+  "eval_steps": 500,
+  "global_step": 3264,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8153233402967452,
+      "epoch": 0.12269938650306748,
+      "grad_norm": 1.8510568141937256,
+      "learning_rate": 4.8469594667113335e-05,
+      "loss": 1.7217323303222656,
+      "mean_token_accuracy": 0.6236635231971741,
+      "num_tokens": 134329.0,
+      "step": 50
+    },
+    {
+      "entropy": 0.97831438601017,
+      "epoch": 0.24539877300613497,
+      "grad_norm": 1.889868974685669,
+      "learning_rate": 9.792836473559631e-05,
+      "loss": 0.9050540924072266,
+      "mean_token_accuracy": 0.7377463465929032,
+      "num_tokens": 262407.0,
+      "step": 100
+    },
+    {
+      "entropy": 0.8743131357431412,
+      "epoch": 0.36809815950920244,
+      "grad_norm": 1.1766732931137085,
+      "learning_rate": 0.0001473871348040793,
+      "loss": 0.8129763031005859,
+      "mean_token_accuracy": 0.7583417356014251,
+      "num_tokens": 401136.0,
+      "step": 150
+    },
+    {
+      "entropy": 0.8273966419696808,
+      "epoch": 0.49079754601226994,
+      "grad_norm": 1.1311014890670776,
+      "learning_rate": 0.00019684590487256229,
+      "loss": 0.7711280822753906,
+      "mean_token_accuracy": 0.7668228060007095,
+      "num_tokens": 531647.0,
+      "step": 200
+    },
+    {
+      "entropy": 0.8101438617706299,
+      "epoch": 0.6134969325153374,
+      "grad_norm": 0.8905950784683228,
+      "learning_rate": 0.0002463046749410453,
+      "loss": 0.7609222412109375,
+      "mean_token_accuracy": 0.7717174577713013,
+      "num_tokens": 666127.0,
+      "step": 250
+    },
+    {
+      "entropy": 0.7885055243968964,
+      "epoch": 0.7361963190184049,
+      "grad_norm": 1.1944245100021362,
+      "learning_rate": 0.00029576344500952824,
+      "loss": 0.7477743530273437,
+      "mean_token_accuracy": 0.773262197971344,
+      "num_tokens": 799292.0,
+      "step": 300
+    },
+    {
+      "entropy": 0.7858527797460556,
+      "epoch": 0.8588957055214724,
+      "grad_norm": 1.1311068534851074,
+      "learning_rate": 0.00034522221507801124,
+      "loss": 0.7411377716064453,
+      "mean_token_accuracy": 0.7745680212974548,
+      "num_tokens": 930614.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.7879797518253326,
+      "epoch": 0.9815950920245399,
+      "grad_norm": 0.9370436072349548,
+      "learning_rate": 0.00039468098514649423,
+      "loss": 0.7507785034179687,
+      "mean_token_accuracy": 0.7747444450855255,
+      "num_tokens": 1061654.0,
+      "step": 400
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.8527693561145238,
+      "eval_mean_token_accuracy": 0.7730878329277039,
+      "eval_not_syn_loss": 0.781921923160553,
+      "eval_not_syn_runtime": 56.9157,
+      "eval_not_syn_samples_per_second": 24.527,
+      "eval_not_syn_steps_per_second": 3.075,
+      "eval_num_tokens": 1080196.0,
+      "step": 408
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.7952005583899362,
+      "eval_mean_token_accuracy": 0.7705784467288426,
+      "eval_num_tokens": 1080196.0,
+      "eval_syn_loss": 0.7461121082305908,
+      "eval_syn_runtime": 59.5758,
+      "eval_syn_samples_per_second": 23.432,
+      "eval_syn_steps_per_second": 2.937,
+      "step": 408
+    },
+    {
+      "entropy": 0.7564875923021875,
+      "epoch": 1.1030674846625768,
+      "grad_norm": 1.1334599256515503,
+      "learning_rate": 0.00040345942966973637,
+      "loss": 0.7208955383300781,
+      "mean_token_accuracy": 0.7788144470465304,
+      "num_tokens": 1199170.0,
+      "step": 450
+    },
+    {
+      "entropy": 0.7427570706605912,
+      "epoch": 1.2257668711656442,
+      "grad_norm": 1.0777021646499634,
+      "learning_rate": 0.00040297229629224153,
+      "loss": 0.7049168395996094,
+      "mean_token_accuracy": 0.7830300116539002,
+      "num_tokens": 1334298.0,
+      "step": 500
+    },
+    {
+      "entropy": 0.7533353108167649,
+      "epoch": 1.3484662576687116,
+      "grad_norm": 1.3439561128616333,
+      "learning_rate": 0.00040211707285041045,
+      "loss": 0.7102288818359375,
+      "mean_token_accuracy": 0.7818449640274048,
+      "num_tokens": 1463189.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.7666342407464981,
+      "epoch": 1.471165644171779,
+      "grad_norm": 1.1329729557037354,
+      "learning_rate": 0.00040089532410439227,
+      "loss": 0.7270996856689453,
+      "mean_token_accuracy": 0.778404277563095,
+      "num_tokens": 1590310.0,
+      "step": 600
+    },
+    {
+      "entropy": 0.7359934949874878,
+      "epoch": 1.5938650306748468,
+      "grad_norm": 1.2227331399917603,
+      "learning_rate": 0.0003993092854276068,
+      "loss": 0.6996555328369141,
+      "mean_token_accuracy": 0.7834958267211914,
+      "num_tokens": 1721836.0,
+      "step": 650
+    },
+    {
+      "entropy": 0.7278932738304138,
+      "epoch": 1.716564417177914,
+      "grad_norm": 0.8753179311752319,
+      "learning_rate": 0.0003973618587167924,
+      "loss": 0.694002456665039,
+      "mean_token_accuracy": 0.7849677371978759,
+      "num_tokens": 1853507.0,
+      "step": 700
+    },
+    {
+      "entropy": 0.7190143716335297,
+      "epoch": 1.8392638036809816,
+      "grad_norm": 0.8524742126464844,
+      "learning_rate": 0.0003950566070825483,
+      "loss": 0.6890143585205079,
+      "mean_token_accuracy": 0.7865350896120071,
+      "num_tokens": 1988113.0,
+      "step": 750
+    },
+    {
+      "entropy": 0.7072037667036056,
+      "epoch": 1.961963190184049,
+      "grad_norm": 0.877348005771637,
+      "learning_rate": 0.00039239774833008714,
+      "loss": 0.6795162200927735,
+      "mean_token_accuracy": 0.7885233855247498,
+      "num_tokens": 2121314.0,
+      "step": 800
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.818787784235818,
+      "eval_mean_token_accuracy": 0.7573730857031686,
+      "eval_not_syn_loss": 0.7486392855644226,
+      "eval_not_syn_runtime": 56.7426,
+      "eval_not_syn_samples_per_second": 24.602,
+      "eval_not_syn_steps_per_second": 3.084,
+      "eval_num_tokens": 2160392.0,
+      "step": 816
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.7632522354807173,
+      "eval_mean_token_accuracy": 0.8101163625717163,
+      "eval_num_tokens": 2160392.0,
+      "eval_syn_loss": 0.6829464435577393,
+      "eval_syn_runtime": 58.5255,
+      "eval_syn_samples_per_second": 23.853,
+      "eval_syn_steps_per_second": 2.99,
+      "step": 816
+    },
+    {
+      "entropy": 0.6619202801675508,
+      "epoch": 2.083435582822086,
+      "grad_norm": 0.8124546408653259,
+      "learning_rate": 0.00038939014724212456,
+      "loss": 0.6311377716064454,
+      "mean_token_accuracy": 0.7991988044796567,
+      "num_tokens": 2253179.0,
+      "step": 850
+    },
+    {
+      "entropy": 0.6668815296888352,
+      "epoch": 2.2061349693251535,
+      "grad_norm": 0.8693709969520569,
+      "learning_rate": 0.0003860393066780265,
+      "loss": 0.6279015731811524,
+      "mean_token_accuracy": 0.7983763587474823,
+      "num_tokens": 2379965.0,
+      "step": 900
+    },
+    {
+      "entropy": 0.6557431703805924,
+      "epoch": 2.3288343558282207,
+      "grad_norm": 0.8323768377304077,
+      "learning_rate": 0.0003823513575054986,
+      "loss": 0.6214478683471679,
+      "mean_token_accuracy": 0.8012632429599762,
+      "num_tokens": 2515593.0,
+      "step": 950
+    },
+    {
+      "entropy": 0.6633540654182434,
+      "epoch": 2.4515337423312884,
+      "grad_norm": 0.8328827023506165,
+      "learning_rate": 0.0003783330473832399,
+      "loss": 0.6310849761962891,
+      "mean_token_accuracy": 0.7995503985881806,
+      "num_tokens": 2642790.0,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6542247200012207,
+      "epoch": 2.574233128834356,
+      "grad_norm": 0.8204748630523682,
+      "learning_rate": 0.000373991728415085,
+      "loss": 0.6195787048339844,
+      "mean_token_accuracy": 0.8006457507610321,
+      "num_tokens": 2774396.0,
+      "step": 1050
+    },
+    {
+      "entropy": 0.6526653742790223,
+      "epoch": 2.6969325153374233,
+      "grad_norm": 1.065635323524475,
+      "learning_rate": 0.0003693353436982218,
+      "loss": 0.6206568145751953,
+      "mean_token_accuracy": 0.8002620726823807,
+      "num_tokens": 2912102.0,
+      "step": 1100
+    },
+    {
+      "entropy": 0.654406590461731,
+      "epoch": 2.819631901840491,
+      "grad_norm": 0.8653210401535034,
+      "learning_rate": 0.00036437241279009834,
+      "loss": 0.6259954452514649,
+      "mean_token_accuracy": 0.8008036535978317,
+      "num_tokens": 3042987.0,
+      "step": 1150
+    },
+    {
+      "entropy": 0.6310120761394501,
+      "epoch": 2.942331288343558,
+      "grad_norm": 0.7713245749473572,
+      "learning_rate": 0.0003591120161206093,
+      "loss": 0.6115898132324219,
+      "mean_token_accuracy": 0.8043569129705429,
+      "num_tokens": 3178044.0,
+      "step": 1200
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6591688336644854,
+      "eval_mean_token_accuracy": 0.7992405581474304,
+      "eval_not_syn_loss": 0.6962469220161438,
+      "eval_not_syn_runtime": 56.161,
+      "eval_not_syn_samples_per_second": 24.857,
+      "eval_not_syn_steps_per_second": 3.116,
+      "eval_num_tokens": 3240588.0,
+      "step": 1224
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6191406740461077,
+      "eval_mean_token_accuracy": 0.7810143651281084,
+      "eval_num_tokens": 3240588.0,
+      "eval_syn_loss": 0.6663697361946106,
+      "eval_syn_runtime": 58.4102,
+      "eval_syn_samples_per_second": 23.9,
+      "eval_syn_steps_per_second": 2.996,
+      "step": 1224
+    },
+    {
+      "entropy": 0.5853322166385073,
+      "epoch": 3.063803680981595,
+      "grad_norm": 0.7313649654388428,
+      "learning_rate": 0.00035356377837808084,
+      "loss": 0.5616408538818359,
+      "mean_token_accuracy": 0.8144418639366073,
+      "num_tokens": 3312797.0,
+      "step": 1250
+    },
+    {
+      "entropy": 0.5654504173994064,
+      "epoch": 3.1865030674846624,
+      "grad_norm": 0.713893711566925,
+      "learning_rate": 0.0003477378508994529,
+      "loss": 0.5305458068847656,
+      "mean_token_accuracy": 0.8218590825796127,
+      "num_tokens": 3441629.0,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5721602493524551,
+      "epoch": 3.30920245398773,
+      "grad_norm": 0.8487221002578735,
+      "learning_rate": 0.00034164489309687927,
+      "loss": 0.5437272262573242,
+      "mean_token_accuracy": 0.8169907212257386,
+      "num_tokens": 3576093.0,
+      "step": 1350
+    },
+    {
+      "entropy": 0.5797226822376251,
+      "epoch": 3.4319018404907977,
+      "grad_norm": 0.9307994842529297,
+      "learning_rate": 0.0003352960529547267,
+      "loss": 0.5405152130126953,
+      "mean_token_accuracy": 0.818563597202301,
+      "num_tokens": 3705912.0,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5822264245152473,
+      "epoch": 3.554601226993865,
+      "grad_norm": 0.71127849817276,
+      "learning_rate": 0.00032870294663265757,
+      "loss": 0.5559784698486329,
+      "mean_token_accuracy": 0.8159428876638413,
+      "num_tokens": 3832376.0,
+      "step": 1450
+    },
+    {
+      "entropy": 0.5702577340602875,
+      "epoch": 3.6773006134969326,
+      "grad_norm": 0.8090758919715881,
+      "learning_rate": 0.0003218776372121157,
+      "loss": 0.5466982650756836,
+      "mean_token_accuracy": 0.8186432421207428,
+      "num_tokens": 3967273.0,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5708746567368508,
+      "epoch": 3.8,
+      "grad_norm": 0.7777291536331177,
+      "learning_rate": 0.000314832612625101,
+      "loss": 0.5412663650512696,
+      "mean_token_accuracy": 0.8195344746112824,
+      "num_tokens": 4102434.0,
+      "step": 1550
+    },
+    {
+      "entropy": 0.578739655315876,
+      "epoch": 3.9226993865030675,
+      "grad_norm": 0.8103565573692322,
+      "learning_rate": 0.0003075807628056167,
+      "loss": 0.5605202865600586,
+      "mean_token_accuracy": 0.815134603381157,
+      "num_tokens": 4234037.0,
+      "step": 1600
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.6183825404303415,
+      "eval_mean_token_accuracy": 0.7838981601170131,
+      "eval_not_syn_loss": 0.6814553737640381,
+      "eval_not_syn_runtime": 56.4858,
+      "eval_not_syn_samples_per_second": 24.714,
+      "eval_not_syn_steps_per_second": 3.098,
+      "eval_num_tokens": 4320784.0,
+      "step": 1632
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5816301565510886,
+      "eval_mean_token_accuracy": 0.8187636685371399,
+      "eval_num_tokens": 4320784.0,
+      "eval_syn_loss": 0.6366032361984253,
+      "eval_syn_runtime": 58.1317,
+      "eval_syn_samples_per_second": 24.014,
+      "eval_syn_steps_per_second": 3.01,
+      "step": 1632
+    },
+    {
+      "entropy": 0.5312421954039371,
+      "epoch": 4.044171779141104,
+      "grad_norm": 0.6530773043632507,
+      "learning_rate": 0.00030013535610559226,
+      "loss": 0.5023544311523438,
+      "mean_token_accuracy": 0.8297024351177793,
+      "num_tokens": 4368268.0,
+      "step": 1650
+    },
+    {
+      "entropy": 0.4853435277938843,
+      "epoch": 4.166871165644172,
+      "grad_norm": 1.0383695363998413,
+      "learning_rate": 0.00029251001501843485,
+      "loss": 0.45016471862792967,
+      "mean_token_accuracy": 0.8418685424327851,
+      "num_tokens": 4494525.0,
+      "step": 1700
+    },
+    {
+      "entropy": 0.48153585255146025,
+      "epoch": 4.289570552147239,
+      "grad_norm": 0.7215332388877869,
+      "learning_rate": 0.00028471869125462477,
+      "loss": 0.4460752487182617,
+      "mean_token_accuracy": 0.8421206694841384,
+      "num_tokens": 4631985.0,
+      "step": 1750
+    },
+    {
+      "entropy": 0.49278041124343874,
+      "epoch": 4.412269938650307,
+      "grad_norm": 0.7771472334861755,
+      "learning_rate": 0.0002767756402149588,
+      "loss": 0.45794551849365234,
+      "mean_token_accuracy": 0.8404830145835877,
+      "num_tokens": 4766323.0,
+      "step": 1800
+    },
+    {
+      "entropy": 0.5082326257228851,
+      "epoch": 4.534969325153375,
+      "grad_norm": 0.9560676217079163,
+      "learning_rate": 0.00026869539490814704,
+      "loss": 0.4632451629638672,
+      "mean_token_accuracy": 0.8386451864242553,
+      "num_tokens": 4895606.0,
+      "step": 1850
+    },
+    {
+      "entropy": 0.5081156292557716,
+      "epoch": 4.6576687116564415,
+      "grad_norm": 0.7719491720199585,
+      "learning_rate": 0.0002604927393604828,
+      "loss": 0.4630893325805664,
+      "mean_token_accuracy": 0.839306333065033,
+      "num_tokens": 5025912.0,
+      "step": 1900
+    },
+    {
+      "entropy": 0.5121548187732696,
+      "epoch": 4.780368098159509,
+      "grad_norm": 0.8171904683113098,
+      "learning_rate": 0.00025218268156623985,
+      "loss": 0.46541053771972657,
+      "mean_token_accuracy": 0.8369353520870209,
+      "num_tokens": 5162972.0,
+      "step": 1950
+    },
+    {
+      "entropy": 0.5085560208559037,
+      "epoch": 4.903067484662577,
+      "grad_norm": 0.5919133424758911,
+      "learning_rate": 0.00024378042602828492,
+      "loss": 0.46405506134033203,
+      "mean_token_accuracy": 0.8379521882534027,
+      "num_tokens": 5297935.0,
+      "step": 2000
+    },
+    {
+      "epoch": 5.0,
+      "eval_entropy": 0.5431888212476458,
+      "eval_mean_token_accuracy": 0.7749751152311053,
+      "eval_not_syn_loss": 0.7035888433456421,
+      "eval_not_syn_runtime": 56.5207,
+      "eval_not_syn_samples_per_second": 24.699,
+      "eval_not_syn_steps_per_second": 3.096,
+      "eval_num_tokens": 5400980.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.0,
+      "eval_entropy": 0.5098190077713558,
+      "eval_mean_token_accuracy": 0.832349077633449,
+      "eval_num_tokens": 5400980.0,
+      "eval_syn_loss": 0.6227898597717285,
+      "eval_syn_runtime": 57.6978,
+      "eval_syn_samples_per_second": 24.195,
+      "eval_syn_steps_per_second": 3.033,
+      "step": 2040
+    },
+    {
+      "entropy": 0.4960306563762703,
+      "epoch": 5.024539877300613,
+      "grad_norm": 0.9411171078681946,
+      "learning_rate": 0.0002353013459391488,
+      "loss": 0.4484004211425781,
+      "mean_token_accuracy": 0.8438825571175778,
+      "num_tokens": 5427362.0,
+      "step": 2050
+    },
+    {
+      "entropy": 0.3916636416316032,
+      "epoch": 5.147239263803681,
+      "grad_norm": 0.8929405808448792,
+      "learning_rate": 0.00022676095505345462,
+      "loss": 0.3446478271484375,
+      "mean_token_accuracy": 0.8732858788967133,
+      "num_tokens": 5564186.0,
+      "step": 2100
+    },
+    {
+      "entropy": 0.39636621803045274,
+      "epoch": 5.269938650306749,
+      "grad_norm": 0.69569331407547,
+      "learning_rate": 0.0002181748793031656,
+      "loss": 0.3548049163818359,
+      "mean_token_accuracy": 0.8699262255430221,
+      "num_tokens": 5694728.0,
+      "step": 2150
+    },
+    {
+      "entropy": 0.4028259950876236,
+      "epoch": 5.392638036809816,
+      "grad_norm": 0.7605823874473572,
+      "learning_rate": 0.00020955882820758868,
+      "loss": 0.35218334197998047,
+      "mean_token_accuracy": 0.8704028391838073,
+      "num_tokens": 5832581.0,
+      "step": 2200
+    },
+    {
+      "entropy": 0.4091260200738907,
+      "epoch": 5.515337423312883,
+      "grad_norm": 0.8990920782089233,
+      "learning_rate": 0.00020092856613044126,
+      "loss": 0.36372703552246094,
+      "mean_token_accuracy": 0.8663889598846436,
+      "num_tokens": 5962417.0,
+      "step": 2250
+    },
+    {
+      "entropy": 0.4057882487773895,
+      "epoch": 5.638036809815951,
+      "grad_norm": 0.9872731566429138,
+      "learning_rate": 0.0001922998834365732,
+      "loss": 0.3588364791870117,
+      "mean_token_accuracy": 0.8686526268720627,
+      "num_tokens": 6092338.0,
+      "step": 2300
+    },
+    {
+      "entropy": 0.41128933161497117,
+      "epoch": 5.7607361963190185,
+      "grad_norm": 0.8626382350921631,
+      "learning_rate": 0.0001836885676011143,
+      "loss": 0.36508998870849607,
+      "mean_token_accuracy": 0.8664334756135941,
+      "num_tokens": 6226327.0,
+      "step": 2350
+    },
+    {
+      "entropy": 0.40721403241157533,
+      "epoch": 5.883435582822086,
+      "grad_norm": 0.775476336479187,
+      "learning_rate": 0.00017511037432391027,
+      "loss": 0.36363380432128906,
+      "mean_token_accuracy": 0.8656364333629608,
+      "num_tokens": 6359083.0,
+      "step": 2400
+    },
+    {
+      "epoch": 6.0,
+      "eval_entropy": 0.4985327124595642,
+      "eval_mean_token_accuracy": 0.8019610796655927,
+      "eval_not_syn_loss": 0.717173159122467,
+      "eval_not_syn_runtime": 56.4169,
+      "eval_not_syn_samples_per_second": 24.744,
+      "eval_not_syn_steps_per_second": 3.102,
+      "eval_num_tokens": 6481176.0,
+      "step": 2448
+    },
+    {
+      "epoch": 6.0,
+      "eval_entropy": 0.4651233986445836,
+      "eval_mean_token_accuracy": 0.8059938202585493,
+      "eval_num_tokens": 6481176.0,
+      "eval_syn_loss": 0.6778863668441772,
+      "eval_syn_runtime": 58.5069,
+      "eval_syn_samples_per_second": 23.86,
+      "eval_syn_steps_per_second": 2.991,
+      "step": 2448
+    },
+    {
+      "entropy": 0.40582828509687174,
+      "epoch": 6.004907975460123,
+      "grad_norm": 0.6788471341133118,
+      "learning_rate": 0.0001665809987020958,
+      "loss": 0.3566559600830078,
+      "mean_token_accuracy": 0.868294438930473,
+      "num_tokens": 6486549.0,
+      "step": 2450
+    },
+    {
+      "entropy": 0.2933592677116394,
+      "epoch": 6.12760736196319,
+      "grad_norm": 0.6598446369171143,
+      "learning_rate": 0.00015811604651354912,
+      "loss": 0.24192693710327148,
+      "mean_token_accuracy": 0.9082661873102188,
+      "num_tokens": 6622384.0,
+      "step": 2500
+    },
+    {
+      "entropy": 0.2975004023313522,
+      "epoch": 6.250306748466258,
+      "grad_norm": 0.7413120865821838,
+      "learning_rate": 0.00014973100566377046,
+      "loss": 0.2502699661254883,
+      "mean_token_accuracy": 0.9056042343378067,
+      "num_tokens": 6758637.0,
+      "step": 2550
+    },
+    {
+      "entropy": 0.3052474121749401,
+      "epoch": 6.373006134969325,
+      "grad_norm": 0.8629281520843506,
+      "learning_rate": 0.00014144121784842467,
+      "loss": 0.2543815040588379,
+      "mean_token_accuracy": 0.9046169209480286,
+      "num_tokens": 6888533.0,
+      "step": 2600
+    },
+    {
+      "entropy": 0.303673956990242,
+      "epoch": 6.495705521472392,
+      "grad_norm": 0.8680911660194397,
+      "learning_rate": 0.000133261850483398,
+      "loss": 0.2529561424255371,
+      "mean_token_accuracy": 0.9040863698720932,
+      "num_tokens": 7023729.0,
+      "step": 2650
+    },
+    {
+      "entropy": 0.2993183335661888,
+      "epoch": 6.61840490797546,
+      "grad_norm": 0.7648728489875793,
+      "learning_rate": 0.00012520786895372493,
+      "loss": 0.2517524528503418,
+      "mean_token_accuracy": 0.9038773095607757,
+      "num_tokens": 7153821.0,
+      "step": 2700
+    },
+    {
+      "entropy": 0.3006903246045113,
+      "epoch": 6.741104294478528,
+      "grad_norm": 0.7433948516845703,
+      "learning_rate": 0.00011729400923216141,
+      "loss": 0.25459712982177735,
+      "mean_token_accuracy": 0.9037652736902237,
+      "num_tokens": 7283513.0,
+      "step": 2750
+    },
+    {
+      "entropy": 0.30416361182928087,
+      "epoch": 6.863803680981595,
+      "grad_norm": 0.8744818568229675,
+      "learning_rate": 0.00010953475091750244,
+      "loss": 0.25621614456176756,
+      "mean_token_accuracy": 0.9019701254367828,
+      "num_tokens": 7411578.0,
+      "step": 2800
+    },
+    {
+      "entropy": 0.2956543755531311,
+      "epoch": 6.986503067484662,
+      "grad_norm": 0.8242411017417908,
+      "learning_rate": 0.00010194429074197415,
+      "loss": 0.24829059600830078,
+      "mean_token_accuracy": 0.9054388856887817,
+      "num_tokens": 7548356.0,
+      "step": 2850
+    },
+    {
+      "epoch": 7.0,
+      "eval_entropy": 0.4070556444781167,
+      "eval_mean_token_accuracy": 0.7848251972879682,
+      "eval_not_syn_loss": 0.8054981827735901,
+      "eval_not_syn_runtime": 56.6822,
+      "eval_not_syn_samples_per_second": 24.629,
+      "eval_not_syn_steps_per_second": 3.087,
+      "eval_num_tokens": 7561372.0,
+      "step": 2856
+    },
+    {
+      "epoch": 7.0,
+      "eval_entropy": 0.3784746325016022,
+      "eval_mean_token_accuracy": 0.8244705755370004,
+      "eval_num_tokens": 7561372.0,
+      "eval_syn_loss": 0.7377892732620239,
+      "eval_syn_runtime": 58.6534,
+      "eval_syn_samples_per_second": 23.801,
+      "eval_syn_steps_per_second": 2.984,
+      "step": 2856
+    },
+    {
+      "entropy": 0.22089176542229122,
+      "epoch": 7.1079754601227,
+      "grad_norm": 0.7850068211555481,
+      "learning_rate": 9.453651659617315e-05,
+      "loss": 0.17431402206420898,
+      "mean_token_accuracy": 0.9356574006754943,
+      "num_tokens": 7671657.0,
+      "step": 2900
+    },
+    {
+      "entropy": 0.2014119729399681,
+      "epoch": 7.230674846625767,
+      "grad_norm": 0.6396872401237488,
+      "learning_rate": 8.732498211907838e-05,
+      "loss": 0.15425819396972656,
+      "mean_token_accuracy": 0.9421506607532502,
+      "num_tokens": 7806989.0,
+      "step": 2950
+    },
+    {
+      "entropy": 0.19853905692696572,
+      "epoch": 7.353374233128834,
+      "grad_norm": 0.8545557260513306,
+      "learning_rate": 8.032288189962588e-05,
+      "loss": 0.1555456066131592,
+      "mean_token_accuracy": 0.9425770407915115,
+      "num_tokens": 7941760.0,
+      "step": 3000
+    },
+    {
+      "entropy": 0.19454577103257178,
+      "epoch": 7.476073619631902,
+      "grad_norm": 0.9338676929473877,
+      "learning_rate": 7.354302733522059e-05,
+      "loss": 0.15240780830383302,
+      "mean_token_accuracy": 0.9427168095111846,
+      "num_tokens": 8080328.0,
+      "step": 3050
+    },
+    {
+      "entropy": 0.19621057718992232,
+      "epoch": 7.598773006134969,
+      "grad_norm": 0.920280933380127,
+      "learning_rate": 6.699782319135416e-05,
+      "loss": 0.1560393238067627,
+      "mean_token_accuracy": 0.9418092548847199,
+      "num_tokens": 8210831.0,
+      "step": 3100
+    },
+    {
+      "entropy": 0.19532680556178092,
+      "epoch": 7.721472392638037,
+      "grad_norm": 0.8807936906814575,
+      "learning_rate": 6.069924490521774e-05,
+      "loss": 0.15486297607421876,
+      "mean_token_accuracy": 0.9418806570768357,
+      "num_tokens": 8344985.0,
+      "step": 3150
+    },
+    {
+      "entropy": 0.19740462571382522,
+      "epoch": 7.844171779141105,
+      "grad_norm": 1.0521959066390991,
+      "learning_rate": 5.4658816674834886e-05,
+      "loss": 0.15762215614318847,
+      "mean_token_accuracy": 0.9403509825468064,
+      "num_tokens": 8475716.0,
+      "step": 3200
+    },
+    {
+      "entropy": 0.1947207669913769,
+      "epoch": 7.9668711656441715,
+      "grad_norm": 0.9047127366065979,
+      "learning_rate": 4.888759037380488e-05,
+      "loss": 0.15496024131774902,
+      "mean_token_accuracy": 0.9418564343452454,
+      "num_tokens": 8606709.0,
+      "step": 3250
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.34109559910637993,
+      "eval_mean_token_accuracy": 0.7936825769288199,
+      "eval_not_syn_loss": 0.9095119833946228,
+      "eval_not_syn_runtime": 56.9258,
+      "eval_not_syn_samples_per_second": 24.523,
+      "eval_not_syn_steps_per_second": 3.074,
+      "eval_num_tokens": 8641568.0,
+      "step": 3264
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.31804646364280154,
+      "eval_mean_token_accuracy": 0.8126268771716526,
+      "eval_num_tokens": 8641568.0,
+      "eval_syn_loss": 0.8467251658439636,
+      "eval_syn_runtime": 58.7085,
+      "eval_syn_samples_per_second": 23.779,
+      "eval_syn_steps_per_second": 2.981,
+      "step": 3264
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 4080,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.7583979889495923e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3672/README.md b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3672/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3672/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3672/adapter_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3672/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..100cd582d71fb39c74fbba0e56e6d3415622454e
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3672/adapter_config.json
@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.05635041589438875,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "gate_proj",
+    "up_proj",
+    "o_proj",
+    "down_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3672/chat_template.jinja b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3672/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3672/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is true %}
+        {{- '<think>\n' }}
+    {%- else %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3672/tokenizer_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3672/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3672/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3672/trainer_state.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3672/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..2d33c455fc458afa0f96ad0ead1e339ef647195a
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-3672/trainer_state.json
@@ -0,0 +1,962 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.0,
+  "eval_steps": 500,
+  "global_step": 3672,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8153233402967452,
+      "epoch": 0.12269938650306748,
+      "grad_norm": 1.8510568141937256,
+      "learning_rate": 4.8469594667113335e-05,
+      "loss": 1.7217323303222656,
+      "mean_token_accuracy": 0.6236635231971741,
+      "num_tokens": 134329.0,
+      "step": 50
+    },
+    {
+      "entropy": 0.97831438601017,
+      "epoch": 0.24539877300613497,
+      "grad_norm": 1.889868974685669,
+      "learning_rate": 9.792836473559631e-05,
+      "loss": 0.9050540924072266,
+      "mean_token_accuracy": 0.7377463465929032,
+      "num_tokens": 262407.0,
+      "step": 100
+    },
+    {
+      "entropy": 0.8743131357431412,
+      "epoch": 0.36809815950920244,
+      "grad_norm": 1.1766732931137085,
+      "learning_rate": 0.0001473871348040793,
+      "loss": 0.8129763031005859,
+      "mean_token_accuracy": 0.7583417356014251,
+      "num_tokens": 401136.0,
+      "step": 150
+    },
+    {
+      "entropy": 0.8273966419696808,
+      "epoch": 0.49079754601226994,
+      "grad_norm": 1.1311014890670776,
+      "learning_rate": 0.00019684590487256229,
+      "loss": 0.7711280822753906,
+      "mean_token_accuracy": 0.7668228060007095,
+      "num_tokens": 531647.0,
+      "step": 200
+    },
+    {
+      "entropy": 0.8101438617706299,
+      "epoch": 0.6134969325153374,
+      "grad_norm": 0.8905950784683228,
+      "learning_rate": 0.0002463046749410453,
+      "loss": 0.7609222412109375,
+      "mean_token_accuracy": 0.7717174577713013,
+      "num_tokens": 666127.0,
+      "step": 250
+    },
+    {
+      "entropy": 0.7885055243968964,
+      "epoch": 0.7361963190184049,
+      "grad_norm": 1.1944245100021362,
+      "learning_rate": 0.00029576344500952824,
+      "loss": 0.7477743530273437,
+      "mean_token_accuracy": 0.773262197971344,
+      "num_tokens": 799292.0,
+      "step": 300
+    },
+    {
+      "entropy": 0.7858527797460556,
+      "epoch": 0.8588957055214724,
+      "grad_norm": 1.1311068534851074,
+      "learning_rate": 0.00034522221507801124,
+      "loss": 0.7411377716064453,
+      "mean_token_accuracy": 0.7745680212974548,
+      "num_tokens": 930614.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.7879797518253326,
+      "epoch": 0.9815950920245399,
+      "grad_norm": 0.9370436072349548,
+      "learning_rate": 0.00039468098514649423,
+      "loss": 0.7507785034179687,
+      "mean_token_accuracy": 0.7747444450855255,
+      "num_tokens": 1061654.0,
+      "step": 400
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.8527693561145238,
+      "eval_mean_token_accuracy": 0.7730878329277039,
+      "eval_not_syn_loss": 0.781921923160553,
+      "eval_not_syn_runtime": 56.9157,
+      "eval_not_syn_samples_per_second": 24.527,
+      "eval_not_syn_steps_per_second": 3.075,
+      "eval_num_tokens": 1080196.0,
+      "step": 408
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.7952005583899362,
+      "eval_mean_token_accuracy": 0.7705784467288426,
+      "eval_num_tokens": 1080196.0,
+      "eval_syn_loss": 0.7461121082305908,
+      "eval_syn_runtime": 59.5758,
+      "eval_syn_samples_per_second": 23.432,
+      "eval_syn_steps_per_second": 2.937,
+      "step": 408
+    },
+    {
+      "entropy": 0.7564875923021875,
+      "epoch": 1.1030674846625768,
+      "grad_norm": 1.1334599256515503,
+      "learning_rate": 0.00040345942966973637,
+      "loss": 0.7208955383300781,
+      "mean_token_accuracy": 0.7788144470465304,
+      "num_tokens": 1199170.0,
+      "step": 450
+    },
+    {
+      "entropy": 0.7427570706605912,
+      "epoch": 1.2257668711656442,
+      "grad_norm": 1.0777021646499634,
+      "learning_rate": 0.00040297229629224153,
+      "loss": 0.7049168395996094,
+      "mean_token_accuracy": 0.7830300116539002,
+      "num_tokens": 1334298.0,
+      "step": 500
+    },
+    {
+      "entropy": 0.7533353108167649,
+      "epoch": 1.3484662576687116,
+      "grad_norm": 1.3439561128616333,
+      "learning_rate": 0.00040211707285041045,
+      "loss": 0.7102288818359375,
+      "mean_token_accuracy": 0.7818449640274048,
+      "num_tokens": 1463189.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.7666342407464981,
+      "epoch": 1.471165644171779,
+      "grad_norm": 1.1329729557037354,
+      "learning_rate": 0.00040089532410439227,
+      "loss": 0.7270996856689453,
+      "mean_token_accuracy": 0.778404277563095,
+      "num_tokens": 1590310.0,
+      "step": 600
+    },
+    {
+      "entropy": 0.7359934949874878,
+      "epoch": 1.5938650306748468,
+      "grad_norm": 1.2227331399917603,
+      "learning_rate": 0.0003993092854276068,
+      "loss": 0.6996555328369141,
+      "mean_token_accuracy": 0.7834958267211914,
+      "num_tokens": 1721836.0,
+      "step": 650
+    },
+    {
+      "entropy": 0.7278932738304138,
+      "epoch": 1.716564417177914,
+      "grad_norm": 0.8753179311752319,
+      "learning_rate": 0.0003973618587167924,
+      "loss": 0.694002456665039,
+      "mean_token_accuracy": 0.7849677371978759,
+      "num_tokens": 1853507.0,
+      "step": 700
+    },
+    {
+      "entropy": 0.7190143716335297,
+      "epoch": 1.8392638036809816,
+      "grad_norm": 0.8524742126464844,
+      "learning_rate": 0.0003950566070825483,
+      "loss": 0.6890143585205079,
+      "mean_token_accuracy": 0.7865350896120071,
+      "num_tokens": 1988113.0,
+      "step": 750
+    },
+    {
+      "entropy": 0.7072037667036056,
+      "epoch": 1.961963190184049,
+      "grad_norm": 0.877348005771637,
+      "learning_rate": 0.00039239774833008714,
+      "loss": 0.6795162200927735,
+      "mean_token_accuracy": 0.7885233855247498,
+      "num_tokens": 2121314.0,
+      "step": 800
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.818787784235818,
+      "eval_mean_token_accuracy": 0.7573730857031686,
+      "eval_not_syn_loss": 0.7486392855644226,
+      "eval_not_syn_runtime": 56.7426,
+      "eval_not_syn_samples_per_second": 24.602,
+      "eval_not_syn_steps_per_second": 3.084,
+      "eval_num_tokens": 2160392.0,
+      "step": 816
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.7632522354807173,
+      "eval_mean_token_accuracy": 0.8101163625717163,
+      "eval_num_tokens": 2160392.0,
+      "eval_syn_loss": 0.6829464435577393,
+      "eval_syn_runtime": 58.5255,
+      "eval_syn_samples_per_second": 23.853,
+      "eval_syn_steps_per_second": 2.99,
+      "step": 816
+    },
+    {
+      "entropy": 0.6619202801675508,
+      "epoch": 2.083435582822086,
+      "grad_norm": 0.8124546408653259,
+      "learning_rate": 0.00038939014724212456,
+      "loss": 0.6311377716064454,
+      "mean_token_accuracy": 0.7991988044796567,
+      "num_tokens": 2253179.0,
+      "step": 850
+    },
+    {
+      "entropy": 0.6668815296888352,
+      "epoch": 2.2061349693251535,
+      "grad_norm": 0.8693709969520569,
+      "learning_rate": 0.0003860393066780265,
+      "loss": 0.6279015731811524,
+      "mean_token_accuracy": 0.7983763587474823,
+      "num_tokens": 2379965.0,
+      "step": 900
+    },
+    {
+      "entropy": 0.6557431703805924,
+      "epoch": 2.3288343558282207,
+      "grad_norm": 0.8323768377304077,
+      "learning_rate": 0.0003823513575054986,
+      "loss": 0.6214478683471679,
+      "mean_token_accuracy": 0.8012632429599762,
+      "num_tokens": 2515593.0,
+      "step": 950
+    },
+    {
+      "entropy": 0.6633540654182434,
+      "epoch": 2.4515337423312884,
+      "grad_norm": 0.8328827023506165,
+      "learning_rate": 0.0003783330473832399,
+      "loss": 0.6310849761962891,
+      "mean_token_accuracy": 0.7995503985881806,
+      "num_tokens": 2642790.0,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6542247200012207,
+      "epoch": 2.574233128834356,
+      "grad_norm": 0.8204748630523682,
+      "learning_rate": 0.000373991728415085,
+      "loss": 0.6195787048339844,
+      "mean_token_accuracy": 0.8006457507610321,
+      "num_tokens": 2774396.0,
+      "step": 1050
+    },
+    {
+      "entropy": 0.6526653742790223,
+      "epoch": 2.6969325153374233,
+      "grad_norm": 1.065635323524475,
+      "learning_rate": 0.0003693353436982218,
+      "loss": 0.6206568145751953,
+      "mean_token_accuracy": 0.8002620726823807,
+      "num_tokens": 2912102.0,
+      "step": 1100
+    },
+    {
+      "entropy": 0.654406590461731,
+      "epoch": 2.819631901840491,
+      "grad_norm": 0.8653210401535034,
+      "learning_rate": 0.00036437241279009834,
+      "loss": 0.6259954452514649,
+      "mean_token_accuracy": 0.8008036535978317,
+      "num_tokens": 3042987.0,
+      "step": 1150
+    },
+    {
+      "entropy": 0.6310120761394501,
+      "epoch": 2.942331288343558,
+      "grad_norm": 0.7713245749473572,
+      "learning_rate": 0.0003591120161206093,
+      "loss": 0.6115898132324219,
+      "mean_token_accuracy": 0.8043569129705429,
+      "num_tokens": 3178044.0,
+      "step": 1200
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6591688336644854,
+      "eval_mean_token_accuracy": 0.7992405581474304,
+      "eval_not_syn_loss": 0.6962469220161438,
+      "eval_not_syn_runtime": 56.161,
+      "eval_not_syn_samples_per_second": 24.857,
+      "eval_not_syn_steps_per_second": 3.116,
+      "eval_num_tokens": 3240588.0,
+      "step": 1224
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6191406740461077,
+      "eval_mean_token_accuracy": 0.7810143651281084,
+      "eval_num_tokens": 3240588.0,
+      "eval_syn_loss": 0.6663697361946106,
+      "eval_syn_runtime": 58.4102,
+      "eval_syn_samples_per_second": 23.9,
+      "eval_syn_steps_per_second": 2.996,
+      "step": 1224
+    },
+    {
+      "entropy": 0.5853322166385073,
+      "epoch": 3.063803680981595,
+      "grad_norm": 0.7313649654388428,
+      "learning_rate": 0.00035356377837808084,
+      "loss": 0.5616408538818359,
+      "mean_token_accuracy": 0.8144418639366073,
+      "num_tokens": 3312797.0,
+      "step": 1250
+    },
+    {
+      "entropy": 0.5654504173994064,
+      "epoch": 3.1865030674846624,
+      "grad_norm": 0.713893711566925,
+      "learning_rate": 0.0003477378508994529,
+      "loss": 0.5305458068847656,
+      "mean_token_accuracy": 0.8218590825796127,
+      "num_tokens": 3441629.0,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5721602493524551,
+      "epoch": 3.30920245398773,
+      "grad_norm": 0.8487221002578735,
+      "learning_rate": 0.00034164489309687927,
+      "loss": 0.5437272262573242,
+      "mean_token_accuracy": 0.8169907212257386,
+      "num_tokens": 3576093.0,
+      "step": 1350
+    },
+    {
+      "entropy": 0.5797226822376251,
+      "epoch": 3.4319018404907977,
+      "grad_norm": 0.9307994842529297,
+      "learning_rate": 0.0003352960529547267,
+      "loss": 0.5405152130126953,
+      "mean_token_accuracy": 0.818563597202301,
+      "num_tokens": 3705912.0,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5822264245152473,
+      "epoch": 3.554601226993865,
+      "grad_norm": 0.71127849817276,
+      "learning_rate": 0.00032870294663265757,
+      "loss": 0.5559784698486329,
+      "mean_token_accuracy": 0.8159428876638413,
+      "num_tokens": 3832376.0,
+      "step": 1450
+    },
+    {
+      "entropy": 0.5702577340602875,
+      "epoch": 3.6773006134969326,
+      "grad_norm": 0.8090758919715881,
+      "learning_rate": 0.0003218776372121157,
+      "loss": 0.5466982650756836,
+      "mean_token_accuracy": 0.8186432421207428,
+      "num_tokens": 3967273.0,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5708746567368508,
+      "epoch": 3.8,
+      "grad_norm": 0.7777291536331177,
+      "learning_rate": 0.000314832612625101,
+      "loss": 0.5412663650512696,
+      "mean_token_accuracy": 0.8195344746112824,
+      "num_tokens": 4102434.0,
+      "step": 1550
+    },
+    {
+      "entropy": 0.578739655315876,
+      "epoch": 3.9226993865030675,
+      "grad_norm": 0.8103565573692322,
+      "learning_rate": 0.0003075807628056167,
+      "loss": 0.5605202865600586,
+      "mean_token_accuracy": 0.815134603381157,
+      "num_tokens": 4234037.0,
+      "step": 1600
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.6183825404303415,
+      "eval_mean_token_accuracy": 0.7838981601170131,
+      "eval_not_syn_loss": 0.6814553737640381,
+      "eval_not_syn_runtime": 56.4858,
+      "eval_not_syn_samples_per_second": 24.714,
+      "eval_not_syn_steps_per_second": 3.098,
+      "eval_num_tokens": 4320784.0,
+      "step": 1632
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5816301565510886,
+      "eval_mean_token_accuracy": 0.8187636685371399,
+      "eval_num_tokens": 4320784.0,
+      "eval_syn_loss": 0.6366032361984253,
+      "eval_syn_runtime": 58.1317,
+      "eval_syn_samples_per_second": 24.014,
+      "eval_syn_steps_per_second": 3.01,
+      "step": 1632
+    },
+    {
+      "entropy": 0.5312421954039371,
+      "epoch": 4.044171779141104,
+      "grad_norm": 0.6530773043632507,
+      "learning_rate": 0.00030013535610559226,
+      "loss": 0.5023544311523438,
+      "mean_token_accuracy": 0.8297024351177793,
+      "num_tokens": 4368268.0,
+      "step": 1650
+    },
+    {
+      "entropy": 0.4853435277938843,
+      "epoch": 4.166871165644172,
+      "grad_norm": 1.0383695363998413,
+      "learning_rate": 0.00029251001501843485,
+      "loss": 0.45016471862792967,
+      "mean_token_accuracy": 0.8418685424327851,
+      "num_tokens": 4494525.0,
+      "step": 1700
+    },
+    {
+      "entropy": 0.48153585255146025,
+      "epoch": 4.289570552147239,
+      "grad_norm": 0.7215332388877869,
+      "learning_rate": 0.00028471869125462477,
+      "loss": 0.4460752487182617,
+      "mean_token_accuracy": 0.8421206694841384,
+      "num_tokens": 4631985.0,
+      "step": 1750
+    },
+    {
+      "entropy": 0.49278041124343874,
+      "epoch": 4.412269938650307,
+      "grad_norm": 0.7771472334861755,
+      "learning_rate": 0.0002767756402149588,
+      "loss": 0.45794551849365234,
+      "mean_token_accuracy": 0.8404830145835877,
+      "num_tokens": 4766323.0,
+      "step": 1800
+    },
+    {
+      "entropy": 0.5082326257228851,
+      "epoch": 4.534969325153375,
+      "grad_norm": 0.9560676217079163,
+      "learning_rate": 0.00026869539490814704,
+      "loss": 0.4632451629638672,
+      "mean_token_accuracy": 0.8386451864242553,
+      "num_tokens": 4895606.0,
+      "step": 1850
+    },
+    {
+      "entropy": 0.5081156292557716,
+      "epoch": 4.6576687116564415,
+      "grad_norm": 0.7719491720199585,
+      "learning_rate": 0.0002604927393604828,
+      "loss": 0.4630893325805664,
+      "mean_token_accuracy": 0.839306333065033,
+      "num_tokens": 5025912.0,
+      "step": 1900
+    },
+    {
+      "entropy": 0.5121548187732696,
+      "epoch": 4.780368098159509,
+      "grad_norm": 0.8171904683113098,
+      "learning_rate": 0.00025218268156623985,
+      "loss": 0.46541053771972657,
+      "mean_token_accuracy": 0.8369353520870209,
+      "num_tokens": 5162972.0,
+      "step": 1950
+    },
+    {
+      "entropy": 0.5085560208559037,
+      "epoch": 4.903067484662577,
+      "grad_norm": 0.5919133424758911,
+      "learning_rate": 0.00024378042602828492,
+      "loss": 0.46405506134033203,
+      "mean_token_accuracy": 0.8379521882534027,
+      "num_tokens": 5297935.0,
+      "step": 2000
+    },
+    {
+      "epoch": 5.0,
+      "eval_entropy": 0.5431888212476458,
+      "eval_mean_token_accuracy": 0.7749751152311053,
+      "eval_not_syn_loss": 0.7035888433456421,
+      "eval_not_syn_runtime": 56.5207,
+      "eval_not_syn_samples_per_second": 24.699,
+      "eval_not_syn_steps_per_second": 3.096,
+      "eval_num_tokens": 5400980.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.0,
+      "eval_entropy": 0.5098190077713558,
+      "eval_mean_token_accuracy": 0.832349077633449,
+      "eval_num_tokens": 5400980.0,
+      "eval_syn_loss": 0.6227898597717285,
+      "eval_syn_runtime": 57.6978,
+      "eval_syn_samples_per_second": 24.195,
+      "eval_syn_steps_per_second": 3.033,
+      "step": 2040
+    },
+    {
+      "entropy": 0.4960306563762703,
+      "epoch": 5.024539877300613,
+      "grad_norm": 0.9411171078681946,
+      "learning_rate": 0.0002353013459391488,
+      "loss": 0.4484004211425781,
+      "mean_token_accuracy": 0.8438825571175778,
+      "num_tokens": 5427362.0,
+      "step": 2050
+    },
+    {
+      "entropy": 0.3916636416316032,
+      "epoch": 5.147239263803681,
+      "grad_norm": 0.8929405808448792,
+      "learning_rate": 0.00022676095505345462,
+      "loss": 0.3446478271484375,
+      "mean_token_accuracy": 0.8732858788967133,
+      "num_tokens": 5564186.0,
+      "step": 2100
+    },
+    {
+      "entropy": 0.39636621803045274,
+      "epoch": 5.269938650306749,
+      "grad_norm": 0.69569331407547,
+      "learning_rate": 0.0002181748793031656,
+      "loss": 0.3548049163818359,
+      "mean_token_accuracy": 0.8699262255430221,
+      "num_tokens": 5694728.0,
+      "step": 2150
+    },
+    {
+      "entropy": 0.4028259950876236,
+      "epoch": 5.392638036809816,
+      "grad_norm": 0.7605823874473572,
+      "learning_rate": 0.00020955882820758868,
+      "loss": 0.35218334197998047,
+      "mean_token_accuracy": 0.8704028391838073,
+      "num_tokens": 5832581.0,
+      "step": 2200
+    },
+    {
+      "entropy": 0.4091260200738907,
+      "epoch": 5.515337423312883,
+      "grad_norm": 0.8990920782089233,
+      "learning_rate": 0.00020092856613044126,
+      "loss": 0.36372703552246094,
+      "mean_token_accuracy": 0.8663889598846436,
+      "num_tokens": 5962417.0,
+      "step": 2250
+    },
+    {
+      "entropy": 0.4057882487773895,
+      "epoch": 5.638036809815951,
+      "grad_norm": 0.9872731566429138,
+      "learning_rate": 0.0001922998834365732,
+      "loss": 0.3588364791870117,
+      "mean_token_accuracy": 0.8686526268720627,
+      "num_tokens": 6092338.0,
+      "step": 2300
+    },
+    {
+      "entropy": 0.41128933161497117,
+      "epoch": 5.7607361963190185,
+      "grad_norm": 0.8626382350921631,
+      "learning_rate": 0.0001836885676011143,
+      "loss": 0.36508998870849607,
+      "mean_token_accuracy": 0.8664334756135941,
+      "num_tokens": 6226327.0,
+      "step": 2350
+    },
+    {
+      "entropy": 0.40721403241157533,
+      "epoch": 5.883435582822086,
+      "grad_norm": 0.775476336479187,
+      "learning_rate": 0.00017511037432391027,
+      "loss": 0.36363380432128906,
+      "mean_token_accuracy": 0.8656364333629608,
+      "num_tokens": 6359083.0,
+      "step": 2400
+    },
+    {
+      "epoch": 6.0,
+      "eval_entropy": 0.4985327124595642,
+      "eval_mean_token_accuracy": 0.8019610796655927,
+      "eval_not_syn_loss": 0.717173159122467,
+      "eval_not_syn_runtime": 56.4169,
+      "eval_not_syn_samples_per_second": 24.744,
+      "eval_not_syn_steps_per_second": 3.102,
+      "eval_num_tokens": 6481176.0,
+      "step": 2448
+    },
+    {
+      "epoch": 6.0,
+      "eval_entropy": 0.4651233986445836,
+      "eval_mean_token_accuracy": 0.8059938202585493,
+      "eval_num_tokens": 6481176.0,
+      "eval_syn_loss": 0.6778863668441772,
+      "eval_syn_runtime": 58.5069,
+      "eval_syn_samples_per_second": 23.86,
+      "eval_syn_steps_per_second": 2.991,
+      "step": 2448
+    },
+    {
+      "entropy": 0.40582828509687174,
+      "epoch": 6.004907975460123,
+      "grad_norm": 0.6788471341133118,
+      "learning_rate": 0.0001665809987020958,
+      "loss": 0.3566559600830078,
+      "mean_token_accuracy": 0.868294438930473,
+      "num_tokens": 6486549.0,
+      "step": 2450
+    },
+    {
+      "entropy": 0.2933592677116394,
+      "epoch": 6.12760736196319,
+      "grad_norm": 0.6598446369171143,
+      "learning_rate": 0.00015811604651354912,
+      "loss": 0.24192693710327148,
+      "mean_token_accuracy": 0.9082661873102188,
+      "num_tokens": 6622384.0,
+      "step": 2500
+    },
+    {
+      "entropy": 0.2975004023313522,
+      "epoch": 6.250306748466258,
+      "grad_norm": 0.7413120865821838,
+      "learning_rate": 0.00014973100566377046,
+      "loss": 0.2502699661254883,
+      "mean_token_accuracy": 0.9056042343378067,
+      "num_tokens": 6758637.0,
+      "step": 2550
+    },
+    {
+      "entropy": 0.3052474121749401,
+      "epoch": 6.373006134969325,
+      "grad_norm": 0.8629281520843506,
+      "learning_rate": 0.00014144121784842467,
+      "loss": 0.2543815040588379,
+      "mean_token_accuracy": 0.9046169209480286,
+      "num_tokens": 6888533.0,
+      "step": 2600
+    },
+    {
+      "entropy": 0.303673956990242,
+      "epoch": 6.495705521472392,
+      "grad_norm": 0.8680911660194397,
+      "learning_rate": 0.000133261850483398,
+      "loss": 0.2529561424255371,
+      "mean_token_accuracy": 0.9040863698720932,
+      "num_tokens": 7023729.0,
+      "step": 2650
+    },
+    {
+      "entropy": 0.2993183335661888,
+      "epoch": 6.61840490797546,
+      "grad_norm": 0.7648728489875793,
+      "learning_rate": 0.00012520786895372493,
+      "loss": 0.2517524528503418,
+      "mean_token_accuracy": 0.9038773095607757,
+      "num_tokens": 7153821.0,
+      "step": 2700
+    },
+    {
+      "entropy": 0.3006903246045113,
+      "epoch": 6.741104294478528,
+      "grad_norm": 0.7433948516845703,
+      "learning_rate": 0.00011729400923216141,
+      "loss": 0.25459712982177735,
+      "mean_token_accuracy": 0.9037652736902237,
+      "num_tokens": 7283513.0,
+      "step": 2750
+    },
+    {
+      "entropy": 0.30416361182928087,
+      "epoch": 6.863803680981595,
+      "grad_norm": 0.8744818568229675,
+      "learning_rate": 0.00010953475091750244,
+      "loss": 0.25621614456176756,
+      "mean_token_accuracy": 0.9019701254367828,
+      "num_tokens": 7411578.0,
+      "step": 2800
+    },
+    {
+      "entropy": 0.2956543755531311,
+      "epoch": 6.986503067484662,
+      "grad_norm": 0.8242411017417908,
+      "learning_rate": 0.00010194429074197415,
+      "loss": 0.24829059600830078,
+      "mean_token_accuracy": 0.9054388856887817,
+      "num_tokens": 7548356.0,
+      "step": 2850
+    },
+    {
+      "epoch": 7.0,
+      "eval_entropy": 0.4070556444781167,
+      "eval_mean_token_accuracy": 0.7848251972879682,
+      "eval_not_syn_loss": 0.8054981827735901,
+      "eval_not_syn_runtime": 56.6822,
+      "eval_not_syn_samples_per_second": 24.629,
+      "eval_not_syn_steps_per_second": 3.087,
+      "eval_num_tokens": 7561372.0,
+      "step": 2856
+    },
+    {
+      "epoch": 7.0,
+      "eval_entropy": 0.3784746325016022,
+      "eval_mean_token_accuracy": 0.8244705755370004,
+      "eval_num_tokens": 7561372.0,
+      "eval_syn_loss": 0.7377892732620239,
+      "eval_syn_runtime": 58.6534,
+      "eval_syn_samples_per_second": 23.801,
+      "eval_syn_steps_per_second": 2.984,
+      "step": 2856
+    },
+    {
+      "entropy": 0.22089176542229122,
+      "epoch": 7.1079754601227,
+      "grad_norm": 0.7850068211555481,
+      "learning_rate": 9.453651659617315e-05,
+      "loss": 0.17431402206420898,
+      "mean_token_accuracy": 0.9356574006754943,
+      "num_tokens": 7671657.0,
+      "step": 2900
+    },
+    {
+      "entropy": 0.2014119729399681,
+      "epoch": 7.230674846625767,
+      "grad_norm": 0.6396872401237488,
+      "learning_rate": 8.732498211907838e-05,
+      "loss": 0.15425819396972656,
+      "mean_token_accuracy": 0.9421506607532502,
+      "num_tokens": 7806989.0,
+      "step": 2950
+    },
+    {
+      "entropy": 0.19853905692696572,
+      "epoch": 7.353374233128834,
+      "grad_norm": 0.8545557260513306,
+      "learning_rate": 8.032288189962588e-05,
+      "loss": 0.1555456066131592,
+      "mean_token_accuracy": 0.9425770407915115,
+      "num_tokens": 7941760.0,
+      "step": 3000
+    },
+    {
+      "entropy": 0.19454577103257178,
+      "epoch": 7.476073619631902,
+      "grad_norm": 0.9338676929473877,
+      "learning_rate": 7.354302733522059e-05,
+      "loss": 0.15240780830383302,
+      "mean_token_accuracy": 0.9427168095111846,
+      "num_tokens": 8080328.0,
+      "step": 3050
+    },
+    {
+      "entropy": 0.19621057718992232,
+      "epoch": 7.598773006134969,
+      "grad_norm": 0.920280933380127,
+      "learning_rate": 6.699782319135416e-05,
+      "loss": 0.1560393238067627,
+      "mean_token_accuracy": 0.9418092548847199,
+      "num_tokens": 8210831.0,
+      "step": 3100
+    },
+    {
+      "entropy": 0.19532680556178092,
+      "epoch": 7.721472392638037,
+      "grad_norm": 0.8807936906814575,
+      "learning_rate": 6.069924490521774e-05,
+      "loss": 0.15486297607421876,
+      "mean_token_accuracy": 0.9418806570768357,
+      "num_tokens": 8344985.0,
+      "step": 3150
+    },
+    {
+      "entropy": 0.19740462571382522,
+      "epoch": 7.844171779141105,
+      "grad_norm": 1.0521959066390991,
+      "learning_rate": 5.4658816674834886e-05,
+      "loss": 0.15762215614318847,
+      "mean_token_accuracy": 0.9403509825468064,
+      "num_tokens": 8475716.0,
+      "step": 3200
+    },
+    {
+      "entropy": 0.1947207669913769,
+      "epoch": 7.9668711656441715,
+      "grad_norm": 0.9047127366065979,
+      "learning_rate": 4.888759037380488e-05,
+      "loss": 0.15496024131774902,
+      "mean_token_accuracy": 0.9418564343452454,
+      "num_tokens": 8606709.0,
+      "step": 3250
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.34109559910637993,
+      "eval_mean_token_accuracy": 0.7936825769288199,
+      "eval_not_syn_loss": 0.9095119833946228,
+      "eval_not_syn_runtime": 56.9258,
+      "eval_not_syn_samples_per_second": 24.523,
+      "eval_not_syn_steps_per_second": 3.074,
+      "eval_num_tokens": 8641568.0,
+      "step": 3264
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.31804646364280154,
+      "eval_mean_token_accuracy": 0.8126268771716526,
+      "eval_num_tokens": 8641568.0,
+      "eval_syn_loss": 0.8467251658439636,
+      "eval_syn_runtime": 58.7085,
+      "eval_syn_samples_per_second": 23.779,
+      "eval_syn_steps_per_second": 2.981,
+      "step": 3264
+    },
+    {
+      "entropy": 0.16006369128672762,
+      "epoch": 8.088343558282208,
+      "grad_norm": 0.6774137616157532,
+      "learning_rate": 4.339612533023478e-05,
+      "loss": 0.11336767196655273,
+      "mean_token_accuracy": 0.9597395754823781,
+      "num_tokens": 8733256.0,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13934748992323875,
+      "epoch": 8.211042944785277,
+      "grad_norm": 0.7012852430343628,
+      "learning_rate": 3.8194469006857826e-05,
+      "loss": 0.09432272911071778,
+      "mean_token_accuracy": 0.9670496737957001,
+      "num_tokens": 8864507.0,
+      "step": 3350
+    },
+    {
+      "entropy": 0.13847702488303185,
+      "epoch": 8.333742331288343,
+      "grad_norm": 0.6204716563224792,
+      "learning_rate": 3.329213861768602e-05,
+      "loss": 0.09572757720947266,
+      "mean_token_accuracy": 0.9653251791000366,
+      "num_tokens": 8993933.0,
+      "step": 3400
+    },
+    {
+      "entropy": 0.1341943299770355,
+      "epoch": 8.45644171779141,
+      "grad_norm": 0.6529184579849243,
+      "learning_rate": 2.8698103714833297e-05,
+      "loss": 0.09508653640747071,
+      "mean_token_accuracy": 0.9664993113279343,
+      "num_tokens": 9124819.0,
+      "step": 3450
+    },
+    {
+      "entropy": 0.13126649804413318,
+      "epoch": 8.579141104294479,
+      "grad_norm": 0.6677636504173279,
+      "learning_rate": 2.4420769777368618e-05,
+      "loss": 0.09260921478271485,
+      "mean_token_accuracy": 0.967041677236557,
+      "num_tokens": 9266865.0,
+      "step": 3500
+    },
+    {
+      "entropy": 0.13248525604605674,
+      "epoch": 8.701840490797546,
+      "grad_norm": 0.7775622010231018,
+      "learning_rate": 2.0467962832225135e-05,
+      "loss": 0.09405729293823242,
+      "mean_token_accuracy": 0.9667006832361221,
+      "num_tokens": 9402291.0,
+      "step": 3550
+    },
+    {
+      "entropy": 0.13282355941832066,
+      "epoch": 8.824539877300614,
+      "grad_norm": 0.6643354296684265,
+      "learning_rate": 1.6846915135304847e-05,
+      "loss": 0.09236414909362793,
+      "mean_token_accuracy": 0.967280547618866,
+      "num_tokens": 9536676.0,
+      "step": 3600
+    },
+    {
+      "entropy": 0.13367731645703315,
+      "epoch": 8.94723926380368,
+      "grad_norm": 0.7113834023475647,
+      "learning_rate": 1.3564251938976921e-05,
+      "loss": 0.09302605628967285,
+      "mean_token_accuracy": 0.9672682428359985,
+      "num_tokens": 9668941.0,
+      "step": 3650
+    },
+    {
+      "epoch": 9.0,
+      "eval_entropy": 0.30514796699796404,
+      "eval_mean_token_accuracy": 0.7953711404119219,
+      "eval_not_syn_loss": 1.0386866331100464,
+      "eval_not_syn_runtime": 56.9221,
+      "eval_not_syn_samples_per_second": 24.525,
+      "eval_not_syn_steps_per_second": 3.074,
+      "eval_num_tokens": 9721764.0,
+      "step": 3672
+    },
+    {
+      "epoch": 9.0,
+      "eval_entropy": 0.28337277259145466,
+      "eval_mean_token_accuracy": 0.8083208533695766,
+      "eval_num_tokens": 9721764.0,
+      "eval_syn_loss": 0.9751896262168884,
+      "eval_syn_runtime": 58.4388,
+      "eval_syn_samples_per_second": 23.888,
+      "eval_syn_steps_per_second": 2.995,
+      "step": 3672
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 4080,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.9782483744860506e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-408/README.md b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-408/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-408/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-408/adapter_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-408/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..100cd582d71fb39c74fbba0e56e6d3415622454e
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-408/adapter_config.json
@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.05635041589438875,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "gate_proj",
+    "up_proj",
+    "o_proj",
+    "down_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-408/chat_template.jinja b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-408/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-408/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is true %}
+        {{- '<think>\n' }}
+    {%- else %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-408/tokenizer_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-408/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-408/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-408/trainer_state.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-408/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6ee0790c390bc635393f04c95116bbc5614f6737
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-408/trainer_state.json
@@ -0,0 +1,136 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 408,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8153233402967452,
+      "epoch": 0.12269938650306748,
+      "grad_norm": 1.8510568141937256,
+      "learning_rate": 4.8469594667113335e-05,
+      "loss": 1.7217323303222656,
+      "mean_token_accuracy": 0.6236635231971741,
+      "num_tokens": 134329.0,
+      "step": 50
+    },
+    {
+      "entropy": 0.97831438601017,
+      "epoch": 0.24539877300613497,
+      "grad_norm": 1.889868974685669,
+      "learning_rate": 9.792836473559631e-05,
+      "loss": 0.9050540924072266,
+      "mean_token_accuracy": 0.7377463465929032,
+      "num_tokens": 262407.0,
+      "step": 100
+    },
+    {
+      "entropy": 0.8743131357431412,
+      "epoch": 0.36809815950920244,
+      "grad_norm": 1.1766732931137085,
+      "learning_rate": 0.0001473871348040793,
+      "loss": 0.8129763031005859,
+      "mean_token_accuracy": 0.7583417356014251,
+      "num_tokens": 401136.0,
+      "step": 150
+    },
+    {
+      "entropy": 0.8273966419696808,
+      "epoch": 0.49079754601226994,
+      "grad_norm": 1.1311014890670776,
+      "learning_rate": 0.00019684590487256229,
+      "loss": 0.7711280822753906,
+      "mean_token_accuracy": 0.7668228060007095,
+      "num_tokens": 531647.0,
+      "step": 200
+    },
+    {
+      "entropy": 0.8101438617706299,
+      "epoch": 0.6134969325153374,
+      "grad_norm": 0.8905950784683228,
+      "learning_rate": 0.0002463046749410453,
+      "loss": 0.7609222412109375,
+      "mean_token_accuracy": 0.7717174577713013,
+      "num_tokens": 666127.0,
+      "step": 250
+    },
+    {
+      "entropy": 0.7885055243968964,
+      "epoch": 0.7361963190184049,
+      "grad_norm": 1.1944245100021362,
+      "learning_rate": 0.00029576344500952824,
+      "loss": 0.7477743530273437,
+      "mean_token_accuracy": 0.773262197971344,
+      "num_tokens": 799292.0,
+      "step": 300
+    },
+    {
+      "entropy": 0.7858527797460556,
+      "epoch": 0.8588957055214724,
+      "grad_norm": 1.1311068534851074,
+      "learning_rate": 0.00034522221507801124,
+      "loss": 0.7411377716064453,
+      "mean_token_accuracy": 0.7745680212974548,
+      "num_tokens": 930614.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.7879797518253326,
+      "epoch": 0.9815950920245399,
+      "grad_norm": 0.9370436072349548,
+      "learning_rate": 0.00039468098514649423,
+      "loss": 0.7507785034179687,
+      "mean_token_accuracy": 0.7747444450855255,
+      "num_tokens": 1061654.0,
+      "step": 400
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.8527693561145238,
+      "eval_mean_token_accuracy": 0.7730878329277039,
+      "eval_not_syn_loss": 0.781921923160553,
+      "eval_not_syn_runtime": 56.9157,
+      "eval_not_syn_samples_per_second": 24.527,
+      "eval_not_syn_steps_per_second": 3.075,
+      "eval_num_tokens": 1080196.0,
+      "step": 408
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.7952005583899362,
+      "eval_mean_token_accuracy": 0.7705784467288426,
+      "eval_num_tokens": 1080196.0,
+      "eval_syn_loss": 0.7461121082305908,
+      "eval_syn_runtime": 59.5758,
+      "eval_syn_samples_per_second": 23.432,
+      "eval_syn_steps_per_second": 2.937,
+      "step": 408
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 4080,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.2127480787778944e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-4080/README.md b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-4080/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-4080/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-4080/adapter_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-4080/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..100cd582d71fb39c74fbba0e56e6d3415622454e
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-4080/adapter_config.json
@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.05635041589438875,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "gate_proj",
+    "up_proj",
+    "o_proj",
+    "down_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-4080/chat_template.jinja b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-4080/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-4080/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is true %}
+        {{- '<think>\n' }}
+    {%- else %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-4080/tokenizer_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-4080/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-4080/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-4080/trainer_state.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-4080/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..47e42076eba8e4d7fd1caf467481447eb4663dbe
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-4080/trainer_state.json
@@ -0,0 +1,1064 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 10.0,
+  "eval_steps": 500,
+  "global_step": 4080,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8153233402967452,
+      "epoch": 0.12269938650306748,
+      "grad_norm": 1.8510568141937256,
+      "learning_rate": 4.8469594667113335e-05,
+      "loss": 1.7217323303222656,
+      "mean_token_accuracy": 0.6236635231971741,
+      "num_tokens": 134329.0,
+      "step": 50
+    },
+    {
+      "entropy": 0.97831438601017,
+      "epoch": 0.24539877300613497,
+      "grad_norm": 1.889868974685669,
+      "learning_rate": 9.792836473559631e-05,
+      "loss": 0.9050540924072266,
+      "mean_token_accuracy": 0.7377463465929032,
+      "num_tokens": 262407.0,
+      "step": 100
+    },
+    {
+      "entropy": 0.8743131357431412,
+      "epoch": 0.36809815950920244,
+      "grad_norm": 1.1766732931137085,
+      "learning_rate": 0.0001473871348040793,
+      "loss": 0.8129763031005859,
+      "mean_token_accuracy": 0.7583417356014251,
+      "num_tokens": 401136.0,
+      "step": 150
+    },
+    {
+      "entropy": 0.8273966419696808,
+      "epoch": 0.49079754601226994,
+      "grad_norm": 1.1311014890670776,
+      "learning_rate": 0.00019684590487256229,
+      "loss": 0.7711280822753906,
+      "mean_token_accuracy": 0.7668228060007095,
+      "num_tokens": 531647.0,
+      "step": 200
+    },
+    {
+      "entropy": 0.8101438617706299,
+      "epoch": 0.6134969325153374,
+      "grad_norm": 0.8905950784683228,
+      "learning_rate": 0.0002463046749410453,
+      "loss": 0.7609222412109375,
+      "mean_token_accuracy": 0.7717174577713013,
+      "num_tokens": 666127.0,
+      "step": 250
+    },
+    {
+      "entropy": 0.7885055243968964,
+      "epoch": 0.7361963190184049,
+      "grad_norm": 1.1944245100021362,
+      "learning_rate": 0.00029576344500952824,
+      "loss": 0.7477743530273437,
+      "mean_token_accuracy": 0.773262197971344,
+      "num_tokens": 799292.0,
+      "step": 300
+    },
+    {
+      "entropy": 0.7858527797460556,
+      "epoch": 0.8588957055214724,
+      "grad_norm": 1.1311068534851074,
+      "learning_rate": 0.00034522221507801124,
+      "loss": 0.7411377716064453,
+      "mean_token_accuracy": 0.7745680212974548,
+      "num_tokens": 930614.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.7879797518253326,
+      "epoch": 0.9815950920245399,
+      "grad_norm": 0.9370436072349548,
+      "learning_rate": 0.00039468098514649423,
+      "loss": 0.7507785034179687,
+      "mean_token_accuracy": 0.7747444450855255,
+      "num_tokens": 1061654.0,
+      "step": 400
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.8527693561145238,
+      "eval_mean_token_accuracy": 0.7730878329277039,
+      "eval_not_syn_loss": 0.781921923160553,
+      "eval_not_syn_runtime": 56.9157,
+      "eval_not_syn_samples_per_second": 24.527,
+      "eval_not_syn_steps_per_second": 3.075,
+      "eval_num_tokens": 1080196.0,
+      "step": 408
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.7952005583899362,
+      "eval_mean_token_accuracy": 0.7705784467288426,
+      "eval_num_tokens": 1080196.0,
+      "eval_syn_loss": 0.7461121082305908,
+      "eval_syn_runtime": 59.5758,
+      "eval_syn_samples_per_second": 23.432,
+      "eval_syn_steps_per_second": 2.937,
+      "step": 408
+    },
+    {
+      "entropy": 0.7564875923021875,
+      "epoch": 1.1030674846625768,
+      "grad_norm": 1.1334599256515503,
+      "learning_rate": 0.00040345942966973637,
+      "loss": 0.7208955383300781,
+      "mean_token_accuracy": 0.7788144470465304,
+      "num_tokens": 1199170.0,
+      "step": 450
+    },
+    {
+      "entropy": 0.7427570706605912,
+      "epoch": 1.2257668711656442,
+      "grad_norm": 1.0777021646499634,
+      "learning_rate": 0.00040297229629224153,
+      "loss": 0.7049168395996094,
+      "mean_token_accuracy": 0.7830300116539002,
+      "num_tokens": 1334298.0,
+      "step": 500
+    },
+    {
+      "entropy": 0.7533353108167649,
+      "epoch": 1.3484662576687116,
+      "grad_norm": 1.3439561128616333,
+      "learning_rate": 0.00040211707285041045,
+      "loss": 0.7102288818359375,
+      "mean_token_accuracy": 0.7818449640274048,
+      "num_tokens": 1463189.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.7666342407464981,
+      "epoch": 1.471165644171779,
+      "grad_norm": 1.1329729557037354,
+      "learning_rate": 0.00040089532410439227,
+      "loss": 0.7270996856689453,
+      "mean_token_accuracy": 0.778404277563095,
+      "num_tokens": 1590310.0,
+      "step": 600
+    },
+    {
+      "entropy": 0.7359934949874878,
+      "epoch": 1.5938650306748468,
+      "grad_norm": 1.2227331399917603,
+      "learning_rate": 0.0003993092854276068,
+      "loss": 0.6996555328369141,
+      "mean_token_accuracy": 0.7834958267211914,
+      "num_tokens": 1721836.0,
+      "step": 650
+    },
+    {
+      "entropy": 0.7278932738304138,
+      "epoch": 1.716564417177914,
+      "grad_norm": 0.8753179311752319,
+      "learning_rate": 0.0003973618587167924,
+      "loss": 0.694002456665039,
+      "mean_token_accuracy": 0.7849677371978759,
+      "num_tokens": 1853507.0,
+      "step": 700
+    },
+    {
+      "entropy": 0.7190143716335297,
+      "epoch": 1.8392638036809816,
+      "grad_norm": 0.8524742126464844,
+      "learning_rate": 0.0003950566070825483,
+      "loss": 0.6890143585205079,
+      "mean_token_accuracy": 0.7865350896120071,
+      "num_tokens": 1988113.0,
+      "step": 750
+    },
+    {
+      "entropy": 0.7072037667036056,
+      "epoch": 1.961963190184049,
+      "grad_norm": 0.877348005771637,
+      "learning_rate": 0.00039239774833008714,
+      "loss": 0.6795162200927735,
+      "mean_token_accuracy": 0.7885233855247498,
+      "num_tokens": 2121314.0,
+      "step": 800
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.818787784235818,
+      "eval_mean_token_accuracy": 0.7573730857031686,
+      "eval_not_syn_loss": 0.7486392855644226,
+      "eval_not_syn_runtime": 56.7426,
+      "eval_not_syn_samples_per_second": 24.602,
+      "eval_not_syn_steps_per_second": 3.084,
+      "eval_num_tokens": 2160392.0,
+      "step": 816
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.7632522354807173,
+      "eval_mean_token_accuracy": 0.8101163625717163,
+      "eval_num_tokens": 2160392.0,
+      "eval_syn_loss": 0.6829464435577393,
+      "eval_syn_runtime": 58.5255,
+      "eval_syn_samples_per_second": 23.853,
+      "eval_syn_steps_per_second": 2.99,
+      "step": 816
+    },
+    {
+      "entropy": 0.6619202801675508,
+      "epoch": 2.083435582822086,
+      "grad_norm": 0.8124546408653259,
+      "learning_rate": 0.00038939014724212456,
+      "loss": 0.6311377716064454,
+      "mean_token_accuracy": 0.7991988044796567,
+      "num_tokens": 2253179.0,
+      "step": 850
+    },
+    {
+      "entropy": 0.6668815296888352,
+      "epoch": 2.2061349693251535,
+      "grad_norm": 0.8693709969520569,
+      "learning_rate": 0.0003860393066780265,
+      "loss": 0.6279015731811524,
+      "mean_token_accuracy": 0.7983763587474823,
+      "num_tokens": 2379965.0,
+      "step": 900
+    },
+    {
+      "entropy": 0.6557431703805924,
+      "epoch": 2.3288343558282207,
+      "grad_norm": 0.8323768377304077,
+      "learning_rate": 0.0003823513575054986,
+      "loss": 0.6214478683471679,
+      "mean_token_accuracy": 0.8012632429599762,
+      "num_tokens": 2515593.0,
+      "step": 950
+    },
+    {
+      "entropy": 0.6633540654182434,
+      "epoch": 2.4515337423312884,
+      "grad_norm": 0.8328827023506165,
+      "learning_rate": 0.0003783330473832399,
+      "loss": 0.6310849761962891,
+      "mean_token_accuracy": 0.7995503985881806,
+      "num_tokens": 2642790.0,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6542247200012207,
+      "epoch": 2.574233128834356,
+      "grad_norm": 0.8204748630523682,
+      "learning_rate": 0.000373991728415085,
+      "loss": 0.6195787048339844,
+      "mean_token_accuracy": 0.8006457507610321,
+      "num_tokens": 2774396.0,
+      "step": 1050
+    },
+    {
+      "entropy": 0.6526653742790223,
+      "epoch": 2.6969325153374233,
+      "grad_norm": 1.065635323524475,
+      "learning_rate": 0.0003693353436982218,
+      "loss": 0.6206568145751953,
+      "mean_token_accuracy": 0.8002620726823807,
+      "num_tokens": 2912102.0,
+      "step": 1100
+    },
+    {
+      "entropy": 0.654406590461731,
+      "epoch": 2.819631901840491,
+      "grad_norm": 0.8653210401535034,
+      "learning_rate": 0.00036437241279009834,
+      "loss": 0.6259954452514649,
+      "mean_token_accuracy": 0.8008036535978317,
+      "num_tokens": 3042987.0,
+      "step": 1150
+    },
+    {
+      "entropy": 0.6310120761394501,
+      "epoch": 2.942331288343558,
+      "grad_norm": 0.7713245749473572,
+      "learning_rate": 0.0003591120161206093,
+      "loss": 0.6115898132324219,
+      "mean_token_accuracy": 0.8043569129705429,
+      "num_tokens": 3178044.0,
+      "step": 1200
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6591688336644854,
+      "eval_mean_token_accuracy": 0.7992405581474304,
+      "eval_not_syn_loss": 0.6962469220161438,
+      "eval_not_syn_runtime": 56.161,
+      "eval_not_syn_samples_per_second": 24.857,
+      "eval_not_syn_steps_per_second": 3.116,
+      "eval_num_tokens": 3240588.0,
+      "step": 1224
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.6191406740461077,
+      "eval_mean_token_accuracy": 0.7810143651281084,
+      "eval_num_tokens": 3240588.0,
+      "eval_syn_loss": 0.6663697361946106,
+      "eval_syn_runtime": 58.4102,
+      "eval_syn_samples_per_second": 23.9,
+      "eval_syn_steps_per_second": 2.996,
+      "step": 1224
+    },
+    {
+      "entropy": 0.5853322166385073,
+      "epoch": 3.063803680981595,
+      "grad_norm": 0.7313649654388428,
+      "learning_rate": 0.00035356377837808084,
+      "loss": 0.5616408538818359,
+      "mean_token_accuracy": 0.8144418639366073,
+      "num_tokens": 3312797.0,
+      "step": 1250
+    },
+    {
+      "entropy": 0.5654504173994064,
+      "epoch": 3.1865030674846624,
+      "grad_norm": 0.713893711566925,
+      "learning_rate": 0.0003477378508994529,
+      "loss": 0.5305458068847656,
+      "mean_token_accuracy": 0.8218590825796127,
+      "num_tokens": 3441629.0,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5721602493524551,
+      "epoch": 3.30920245398773,
+      "grad_norm": 0.8487221002578735,
+      "learning_rate": 0.00034164489309687927,
+      "loss": 0.5437272262573242,
+      "mean_token_accuracy": 0.8169907212257386,
+      "num_tokens": 3576093.0,
+      "step": 1350
+    },
+    {
+      "entropy": 0.5797226822376251,
+      "epoch": 3.4319018404907977,
+      "grad_norm": 0.9307994842529297,
+      "learning_rate": 0.0003352960529547267,
+      "loss": 0.5405152130126953,
+      "mean_token_accuracy": 0.818563597202301,
+      "num_tokens": 3705912.0,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5822264245152473,
+      "epoch": 3.554601226993865,
+      "grad_norm": 0.71127849817276,
+      "learning_rate": 0.00032870294663265757,
+      "loss": 0.5559784698486329,
+      "mean_token_accuracy": 0.8159428876638413,
+      "num_tokens": 3832376.0,
+      "step": 1450
+    },
+    {
+      "entropy": 0.5702577340602875,
+      "epoch": 3.6773006134969326,
+      "grad_norm": 0.8090758919715881,
+      "learning_rate": 0.0003218776372121157,
+      "loss": 0.5466982650756836,
+      "mean_token_accuracy": 0.8186432421207428,
+      "num_tokens": 3967273.0,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5708746567368508,
+      "epoch": 3.8,
+      "grad_norm": 0.7777291536331177,
+      "learning_rate": 0.000314832612625101,
+      "loss": 0.5412663650512696,
+      "mean_token_accuracy": 0.8195344746112824,
+      "num_tokens": 4102434.0,
+      "step": 1550
+    },
+    {
+      "entropy": 0.578739655315876,
+      "epoch": 3.9226993865030675,
+      "grad_norm": 0.8103565573692322,
+      "learning_rate": 0.0003075807628056167,
+      "loss": 0.5605202865600586,
+      "mean_token_accuracy": 0.815134603381157,
+      "num_tokens": 4234037.0,
+      "step": 1600
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.6183825404303415,
+      "eval_mean_token_accuracy": 0.7838981601170131,
+      "eval_not_syn_loss": 0.6814553737640381,
+      "eval_not_syn_runtime": 56.4858,
+      "eval_not_syn_samples_per_second": 24.714,
+      "eval_not_syn_steps_per_second": 3.098,
+      "eval_num_tokens": 4320784.0,
+      "step": 1632
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5816301565510886,
+      "eval_mean_token_accuracy": 0.8187636685371399,
+      "eval_num_tokens": 4320784.0,
+      "eval_syn_loss": 0.6366032361984253,
+      "eval_syn_runtime": 58.1317,
+      "eval_syn_samples_per_second": 24.014,
+      "eval_syn_steps_per_second": 3.01,
+      "step": 1632
+    },
+    {
+      "entropy": 0.5312421954039371,
+      "epoch": 4.044171779141104,
+      "grad_norm": 0.6530773043632507,
+      "learning_rate": 0.00030013535610559226,
+      "loss": 0.5023544311523438,
+      "mean_token_accuracy": 0.8297024351177793,
+      "num_tokens": 4368268.0,
+      "step": 1650
+    },
+    {
+      "entropy": 0.4853435277938843,
+      "epoch": 4.166871165644172,
+      "grad_norm": 1.0383695363998413,
+      "learning_rate": 0.00029251001501843485,
+      "loss": 0.45016471862792967,
+      "mean_token_accuracy": 0.8418685424327851,
+      "num_tokens": 4494525.0,
+      "step": 1700
+    },
+    {
+      "entropy": 0.48153585255146025,
+      "epoch": 4.289570552147239,
+      "grad_norm": 0.7215332388877869,
+      "learning_rate": 0.00028471869125462477,
+      "loss": 0.4460752487182617,
+      "mean_token_accuracy": 0.8421206694841384,
+      "num_tokens": 4631985.0,
+      "step": 1750
+    },
+    {
+      "entropy": 0.49278041124343874,
+      "epoch": 4.412269938650307,
+      "grad_norm": 0.7771472334861755,
+      "learning_rate": 0.0002767756402149588,
+      "loss": 0.45794551849365234,
+      "mean_token_accuracy": 0.8404830145835877,
+      "num_tokens": 4766323.0,
+      "step": 1800
+    },
+    {
+      "entropy": 0.5082326257228851,
+      "epoch": 4.534969325153375,
+      "grad_norm": 0.9560676217079163,
+      "learning_rate": 0.00026869539490814704,
+      "loss": 0.4632451629638672,
+      "mean_token_accuracy": 0.8386451864242553,
+      "num_tokens": 4895606.0,
+      "step": 1850
+    },
+    {
+      "entropy": 0.5081156292557716,
+      "epoch": 4.6576687116564415,
+      "grad_norm": 0.7719491720199585,
+      "learning_rate": 0.0002604927393604828,
+      "loss": 0.4630893325805664,
+      "mean_token_accuracy": 0.839306333065033,
+      "num_tokens": 5025912.0,
+      "step": 1900
+    },
+    {
+      "entropy": 0.5121548187732696,
+      "epoch": 4.780368098159509,
+      "grad_norm": 0.8171904683113098,
+      "learning_rate": 0.00025218268156623985,
+      "loss": 0.46541053771972657,
+      "mean_token_accuracy": 0.8369353520870209,
+      "num_tokens": 5162972.0,
+      "step": 1950
+    },
+    {
+      "entropy": 0.5085560208559037,
+      "epoch": 4.903067484662577,
+      "grad_norm": 0.5919133424758911,
+      "learning_rate": 0.00024378042602828492,
+      "loss": 0.46405506134033203,
+      "mean_token_accuracy": 0.8379521882534027,
+      "num_tokens": 5297935.0,
+      "step": 2000
+    },
+    {
+      "epoch": 5.0,
+      "eval_entropy": 0.5431888212476458,
+      "eval_mean_token_accuracy": 0.7749751152311053,
+      "eval_not_syn_loss": 0.7035888433456421,
+      "eval_not_syn_runtime": 56.5207,
+      "eval_not_syn_samples_per_second": 24.699,
+      "eval_not_syn_steps_per_second": 3.096,
+      "eval_num_tokens": 5400980.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.0,
+      "eval_entropy": 0.5098190077713558,
+      "eval_mean_token_accuracy": 0.832349077633449,
+      "eval_num_tokens": 5400980.0,
+      "eval_syn_loss": 0.6227898597717285,
+      "eval_syn_runtime": 57.6978,
+      "eval_syn_samples_per_second": 24.195,
+      "eval_syn_steps_per_second": 3.033,
+      "step": 2040
+    },
+    {
+      "entropy": 0.4960306563762703,
+      "epoch": 5.024539877300613,
+      "grad_norm": 0.9411171078681946,
+      "learning_rate": 0.0002353013459391488,
+      "loss": 0.4484004211425781,
+      "mean_token_accuracy": 0.8438825571175778,
+      "num_tokens": 5427362.0,
+      "step": 2050
+    },
+    {
+      "entropy": 0.3916636416316032,
+      "epoch": 5.147239263803681,
+      "grad_norm": 0.8929405808448792,
+      "learning_rate": 0.00022676095505345462,
+      "loss": 0.3446478271484375,
+      "mean_token_accuracy": 0.8732858788967133,
+      "num_tokens": 5564186.0,
+      "step": 2100
+    },
+    {
+      "entropy": 0.39636621803045274,
+      "epoch": 5.269938650306749,
+      "grad_norm": 0.69569331407547,
+      "learning_rate": 0.0002181748793031656,
+      "loss": 0.3548049163818359,
+      "mean_token_accuracy": 0.8699262255430221,
+      "num_tokens": 5694728.0,
+      "step": 2150
+    },
+    {
+      "entropy": 0.4028259950876236,
+      "epoch": 5.392638036809816,
+      "grad_norm": 0.7605823874473572,
+      "learning_rate": 0.00020955882820758868,
+      "loss": 0.35218334197998047,
+      "mean_token_accuracy": 0.8704028391838073,
+      "num_tokens": 5832581.0,
+      "step": 2200
+    },
+    {
+      "entropy": 0.4091260200738907,
+      "epoch": 5.515337423312883,
+      "grad_norm": 0.8990920782089233,
+      "learning_rate": 0.00020092856613044126,
+      "loss": 0.36372703552246094,
+      "mean_token_accuracy": 0.8663889598846436,
+      "num_tokens": 5962417.0,
+      "step": 2250
+    },
+    {
+      "entropy": 0.4057882487773895,
+      "epoch": 5.638036809815951,
+      "grad_norm": 0.9872731566429138,
+      "learning_rate": 0.0001922998834365732,
+      "loss": 0.3588364791870117,
+      "mean_token_accuracy": 0.8686526268720627,
+      "num_tokens": 6092338.0,
+      "step": 2300
+    },
+    {
+      "entropy": 0.41128933161497117,
+      "epoch": 5.7607361963190185,
+      "grad_norm": 0.8626382350921631,
+      "learning_rate": 0.0001836885676011143,
+      "loss": 0.36508998870849607,
+      "mean_token_accuracy": 0.8664334756135941,
+      "num_tokens": 6226327.0,
+      "step": 2350
+    },
+    {
+      "entropy": 0.40721403241157533,
+      "epoch": 5.883435582822086,
+      "grad_norm": 0.775476336479187,
+      "learning_rate": 0.00017511037432391027,
+      "loss": 0.36363380432128906,
+      "mean_token_accuracy": 0.8656364333629608,
+      "num_tokens": 6359083.0,
+      "step": 2400
+    },
+    {
+      "epoch": 6.0,
+      "eval_entropy": 0.4985327124595642,
+      "eval_mean_token_accuracy": 0.8019610796655927,
+      "eval_not_syn_loss": 0.717173159122467,
+      "eval_not_syn_runtime": 56.4169,
+      "eval_not_syn_samples_per_second": 24.744,
+      "eval_not_syn_steps_per_second": 3.102,
+      "eval_num_tokens": 6481176.0,
+      "step": 2448
+    },
+    {
+      "epoch": 6.0,
+      "eval_entropy": 0.4651233986445836,
+      "eval_mean_token_accuracy": 0.8059938202585493,
+      "eval_num_tokens": 6481176.0,
+      "eval_syn_loss": 0.6778863668441772,
+      "eval_syn_runtime": 58.5069,
+      "eval_syn_samples_per_second": 23.86,
+      "eval_syn_steps_per_second": 2.991,
+      "step": 2448
+    },
+    {
+      "entropy": 0.40582828509687174,
+      "epoch": 6.004907975460123,
+      "grad_norm": 0.6788471341133118,
+      "learning_rate": 0.0001665809987020958,
+      "loss": 0.3566559600830078,
+      "mean_token_accuracy": 0.868294438930473,
+      "num_tokens": 6486549.0,
+      "step": 2450
+    },
+    {
+      "entropy": 0.2933592677116394,
+      "epoch": 6.12760736196319,
+      "grad_norm": 0.6598446369171143,
+      "learning_rate": 0.00015811604651354912,
+      "loss": 0.24192693710327148,
+      "mean_token_accuracy": 0.9082661873102188,
+      "num_tokens": 6622384.0,
+      "step": 2500
+    },
+    {
+      "entropy": 0.2975004023313522,
+      "epoch": 6.250306748466258,
+      "grad_norm": 0.7413120865821838,
+      "learning_rate": 0.00014973100566377046,
+      "loss": 0.2502699661254883,
+      "mean_token_accuracy": 0.9056042343378067,
+      "num_tokens": 6758637.0,
+      "step": 2550
+    },
+    {
+      "entropy": 0.3052474121749401,
+      "epoch": 6.373006134969325,
+      "grad_norm": 0.8629281520843506,
+      "learning_rate": 0.00014144121784842467,
+      "loss": 0.2543815040588379,
+      "mean_token_accuracy": 0.9046169209480286,
+      "num_tokens": 6888533.0,
+      "step": 2600
+    },
+    {
+      "entropy": 0.303673956990242,
+      "epoch": 6.495705521472392,
+      "grad_norm": 0.8680911660194397,
+      "learning_rate": 0.000133261850483398,
+      "loss": 0.2529561424255371,
+      "mean_token_accuracy": 0.9040863698720932,
+      "num_tokens": 7023729.0,
+      "step": 2650
+    },
+    {
+      "entropy": 0.2993183335661888,
+      "epoch": 6.61840490797546,
+      "grad_norm": 0.7648728489875793,
+      "learning_rate": 0.00012520786895372493,
+      "loss": 0.2517524528503418,
+      "mean_token_accuracy": 0.9038773095607757,
+      "num_tokens": 7153821.0,
+      "step": 2700
+    },
+    {
+      "entropy": 0.3006903246045113,
+      "epoch": 6.741104294478528,
+      "grad_norm": 0.7433948516845703,
+      "learning_rate": 0.00011729400923216141,
+      "loss": 0.25459712982177735,
+      "mean_token_accuracy": 0.9037652736902237,
+      "num_tokens": 7283513.0,
+      "step": 2750
+    },
+    {
+      "entropy": 0.30416361182928087,
+      "epoch": 6.863803680981595,
+      "grad_norm": 0.8744818568229675,
+      "learning_rate": 0.00010953475091750244,
+      "loss": 0.25621614456176756,
+      "mean_token_accuracy": 0.9019701254367828,
+      "num_tokens": 7411578.0,
+      "step": 2800
+    },
+    {
+      "entropy": 0.2956543755531311,
+      "epoch": 6.986503067484662,
+      "grad_norm": 0.8242411017417908,
+      "learning_rate": 0.00010194429074197415,
+      "loss": 0.24829059600830078,
+      "mean_token_accuracy": 0.9054388856887817,
+      "num_tokens": 7548356.0,
+      "step": 2850
+    },
+    {
+      "epoch": 7.0,
+      "eval_entropy": 0.4070556444781167,
+      "eval_mean_token_accuracy": 0.7848251972879682,
+      "eval_not_syn_loss": 0.8054981827735901,
+      "eval_not_syn_runtime": 56.6822,
+      "eval_not_syn_samples_per_second": 24.629,
+      "eval_not_syn_steps_per_second": 3.087,
+      "eval_num_tokens": 7561372.0,
+      "step": 2856
+    },
+    {
+      "epoch": 7.0,
+      "eval_entropy": 0.3784746325016022,
+      "eval_mean_token_accuracy": 0.8244705755370004,
+      "eval_num_tokens": 7561372.0,
+      "eval_syn_loss": 0.7377892732620239,
+      "eval_syn_runtime": 58.6534,
+      "eval_syn_samples_per_second": 23.801,
+      "eval_syn_steps_per_second": 2.984,
+      "step": 2856
+    },
+    {
+      "entropy": 0.22089176542229122,
+      "epoch": 7.1079754601227,
+      "grad_norm": 0.7850068211555481,
+      "learning_rate": 9.453651659617315e-05,
+      "loss": 0.17431402206420898,
+      "mean_token_accuracy": 0.9356574006754943,
+      "num_tokens": 7671657.0,
+      "step": 2900
+    },
+    {
+      "entropy": 0.2014119729399681,
+      "epoch": 7.230674846625767,
+      "grad_norm": 0.6396872401237488,
+      "learning_rate": 8.732498211907838e-05,
+      "loss": 0.15425819396972656,
+      "mean_token_accuracy": 0.9421506607532502,
+      "num_tokens": 7806989.0,
+      "step": 2950
+    },
+    {
+      "entropy": 0.19853905692696572,
+      "epoch": 7.353374233128834,
+      "grad_norm": 0.8545557260513306,
+      "learning_rate": 8.032288189962588e-05,
+      "loss": 0.1555456066131592,
+      "mean_token_accuracy": 0.9425770407915115,
+      "num_tokens": 7941760.0,
+      "step": 3000
+    },
+    {
+      "entropy": 0.19454577103257178,
+      "epoch": 7.476073619631902,
+      "grad_norm": 0.9338676929473877,
+      "learning_rate": 7.354302733522059e-05,
+      "loss": 0.15240780830383302,
+      "mean_token_accuracy": 0.9427168095111846,
+      "num_tokens": 8080328.0,
+      "step": 3050
+    },
+    {
+      "entropy": 0.19621057718992232,
+      "epoch": 7.598773006134969,
+      "grad_norm": 0.920280933380127,
+      "learning_rate": 6.699782319135416e-05,
+      "loss": 0.1560393238067627,
+      "mean_token_accuracy": 0.9418092548847199,
+      "num_tokens": 8210831.0,
+      "step": 3100
+    },
+    {
+      "entropy": 0.19532680556178092,
+      "epoch": 7.721472392638037,
+      "grad_norm": 0.8807936906814575,
+      "learning_rate": 6.069924490521774e-05,
+      "loss": 0.15486297607421876,
+      "mean_token_accuracy": 0.9418806570768357,
+      "num_tokens": 8344985.0,
+      "step": 3150
+    },
+    {
+      "entropy": 0.19740462571382522,
+      "epoch": 7.844171779141105,
+      "grad_norm": 1.0521959066390991,
+      "learning_rate": 5.4658816674834886e-05,
+      "loss": 0.15762215614318847,
+      "mean_token_accuracy": 0.9403509825468064,
+      "num_tokens": 8475716.0,
+      "step": 3200
+    },
+    {
+      "entropy": 0.1947207669913769,
+      "epoch": 7.9668711656441715,
+      "grad_norm": 0.9047127366065979,
+      "learning_rate": 4.888759037380488e-05,
+      "loss": 0.15496024131774902,
+      "mean_token_accuracy": 0.9418564343452454,
+      "num_tokens": 8606709.0,
+      "step": 3250
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.34109559910637993,
+      "eval_mean_token_accuracy": 0.7936825769288199,
+      "eval_not_syn_loss": 0.9095119833946228,
+      "eval_not_syn_runtime": 56.9258,
+      "eval_not_syn_samples_per_second": 24.523,
+      "eval_not_syn_steps_per_second": 3.074,
+      "eval_num_tokens": 8641568.0,
+      "step": 3264
+    },
+    {
+      "epoch": 8.0,
+      "eval_entropy": 0.31804646364280154,
+      "eval_mean_token_accuracy": 0.8126268771716526,
+      "eval_num_tokens": 8641568.0,
+      "eval_syn_loss": 0.8467251658439636,
+      "eval_syn_runtime": 58.7085,
+      "eval_syn_samples_per_second": 23.779,
+      "eval_syn_steps_per_second": 2.981,
+      "step": 3264
+    },
+    {
+      "entropy": 0.16006369128672762,
+      "epoch": 8.088343558282208,
+      "grad_norm": 0.6774137616157532,
+      "learning_rate": 4.339612533023478e-05,
+      "loss": 0.11336767196655273,
+      "mean_token_accuracy": 0.9597395754823781,
+      "num_tokens": 8733256.0,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13934748992323875,
+      "epoch": 8.211042944785277,
+      "grad_norm": 0.7012852430343628,
+      "learning_rate": 3.8194469006857826e-05,
+      "loss": 0.09432272911071778,
+      "mean_token_accuracy": 0.9670496737957001,
+      "num_tokens": 8864507.0,
+      "step": 3350
+    },
+    {
+      "entropy": 0.13847702488303185,
+      "epoch": 8.333742331288343,
+      "grad_norm": 0.6204716563224792,
+      "learning_rate": 3.329213861768602e-05,
+      "loss": 0.09572757720947266,
+      "mean_token_accuracy": 0.9653251791000366,
+      "num_tokens": 8993933.0,
+      "step": 3400
+    },
+    {
+      "entropy": 0.1341943299770355,
+      "epoch": 8.45644171779141,
+      "grad_norm": 0.6529184579849243,
+      "learning_rate": 2.8698103714833297e-05,
+      "loss": 0.09508653640747071,
+      "mean_token_accuracy": 0.9664993113279343,
+      "num_tokens": 9124819.0,
+      "step": 3450
+    },
+    {
+      "entropy": 0.13126649804413318,
+      "epoch": 8.579141104294479,
+      "grad_norm": 0.6677636504173279,
+      "learning_rate": 2.4420769777368618e-05,
+      "loss": 0.09260921478271485,
+      "mean_token_accuracy": 0.967041677236557,
+      "num_tokens": 9266865.0,
+      "step": 3500
+    },
+    {
+      "entropy": 0.13248525604605674,
+      "epoch": 8.701840490797546,
+      "grad_norm": 0.7775622010231018,
+      "learning_rate": 2.0467962832225135e-05,
+      "loss": 0.09405729293823242,
+      "mean_token_accuracy": 0.9667006832361221,
+      "num_tokens": 9402291.0,
+      "step": 3550
+    },
+    {
+      "entropy": 0.13282355941832066,
+      "epoch": 8.824539877300614,
+      "grad_norm": 0.6643354296684265,
+      "learning_rate": 1.6846915135304847e-05,
+      "loss": 0.09236414909362793,
+      "mean_token_accuracy": 0.967280547618866,
+      "num_tokens": 9536676.0,
+      "step": 3600
+    },
+    {
+      "entropy": 0.13367731645703315,
+      "epoch": 8.94723926380368,
+      "grad_norm": 0.7113834023475647,
+      "learning_rate": 1.3564251938976921e-05,
+      "loss": 0.09302605628967285,
+      "mean_token_accuracy": 0.9672682428359985,
+      "num_tokens": 9668941.0,
+      "step": 3650
+    },
+    {
+      "epoch": 9.0,
+      "eval_entropy": 0.30514796699796404,
+      "eval_mean_token_accuracy": 0.7953711404119219,
+      "eval_not_syn_loss": 1.0386866331100464,
+      "eval_not_syn_runtime": 56.9221,
+      "eval_not_syn_samples_per_second": 24.525,
+      "eval_not_syn_steps_per_second": 3.074,
+      "eval_num_tokens": 9721764.0,
+      "step": 3672
+    },
+    {
+      "epoch": 9.0,
+      "eval_entropy": 0.28337277259145466,
+      "eval_mean_token_accuracy": 0.8083208533695766,
+      "eval_num_tokens": 9721764.0,
+      "eval_syn_loss": 0.9751896262168884,
+      "eval_syn_runtime": 58.4388,
+      "eval_syn_samples_per_second": 23.888,
+      "eval_syn_steps_per_second": 2.995,
+      "step": 3672
+    },
+    {
+      "entropy": 0.12280672442431402,
+      "epoch": 9.068711656441717,
+      "grad_norm": 0.5349636673927307,
+      "learning_rate": 1.062597937017992e-05,
+      "loss": 0.08112168312072754,
+      "mean_token_accuracy": 0.9721009038915538,
+      "num_tokens": 9799022.0,
+      "step": 3700
+    },
+    {
+      "entropy": 0.11758630983531475,
+      "epoch": 9.191411042944786,
+      "grad_norm": 0.5184180736541748,
+      "learning_rate": 8.03747344130768e-06,
+      "loss": 0.07424304962158203,
+      "mean_token_accuracy": 0.9749062228202819,
+      "num_tokens": 9927323.0,
+      "step": 3750
+    },
+    {
+      "entropy": 0.11504819758236408,
+      "epoch": 9.314110429447853,
+      "grad_norm": 0.4414903223514557,
+      "learning_rate": 5.803470213984592e-06,
+      "loss": 0.07420271396636963,
+      "mean_token_accuracy": 0.9745804452896119,
+      "num_tokens": 10057034.0,
+      "step": 3800
+    },
+    {
+      "entropy": 0.11219210438430309,
+      "epoch": 9.43680981595092,
+      "grad_norm": 0.6505016684532166,
+      "learning_rate": 3.928057133727374e-06,
+      "loss": 0.07242022514343262,
+      "mean_token_accuracy": 0.9752460938692092,
+      "num_tokens": 10190986.0,
+      "step": 3850
+    },
+    {
+      "entropy": 0.1172893676161766,
+      "epoch": 9.559509202453988,
+      "grad_norm": 0.4053712785243988,
+      "learning_rate": 2.4146655513473907e-06,
+      "loss": 0.07574840545654297,
+      "mean_token_accuracy": 0.9739107710123062,
+      "num_tokens": 10318963.0,
+      "step": 3900
+    },
+    {
+      "entropy": 0.11008265137672424,
+      "epoch": 9.682208588957055,
+      "grad_norm": 0.4532715082168579,
+      "learning_rate": 1.2660644447774456e-06,
+      "loss": 0.0717643404006958,
+      "mean_token_accuracy": 0.9753157407045364,
+      "num_tokens": 10453673.0,
+      "step": 3950
+    },
+    {
+      "entropy": 0.11246884763240814,
+      "epoch": 9.804907975460123,
+      "grad_norm": 0.46811601519584656,
+      "learning_rate": 4.843553528094893e-07,
+      "loss": 0.07214241027832032,
+      "mean_token_accuracy": 0.9749426013231277,
+      "num_tokens": 10588217.0,
+      "step": 4000
+    },
+    {
+      "entropy": 0.1111858068406582,
+      "epoch": 9.92760736196319,
+      "grad_norm": 0.5065276026725769,
+      "learning_rate": 7.096853001258902e-08,
+      "loss": 0.07076848030090332,
+      "mean_token_accuracy": 0.9755072641372681,
+      "num_tokens": 10725457.0,
+      "step": 4050
+    },
+    {
+      "epoch": 10.0,
+      "eval_entropy": 0.29344515562057494,
+      "eval_mean_token_accuracy": 0.7949213702338083,
+      "eval_not_syn_loss": 1.1034446954727173,
+      "eval_not_syn_runtime": 55.6095,
+      "eval_not_syn_samples_per_second": 25.104,
+      "eval_not_syn_steps_per_second": 3.147,
+      "eval_num_tokens": 10801960.0,
+      "step": 4080
+    },
+    {
+      "epoch": 10.0,
+      "eval_entropy": 0.271583269068173,
+      "eval_mean_token_accuracy": 0.809326388154711,
+      "eval_num_tokens": 10801960.0,
+      "eval_syn_loss": 1.03242027759552,
+      "eval_syn_runtime": 58.6681,
+      "eval_syn_samples_per_second": 23.795,
+      "eval_syn_steps_per_second": 2.983,
+      "step": 4080
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 4080,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.1980494514604096e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-816/README.md b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-816/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..7ac50e2706e5b4eae8f28e204601afb241c49e55
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-816/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-2B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-2B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-816/adapter_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-816/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..100cd582d71fb39c74fbba0e56e6d3415622454e
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-816/adapter_config.json
@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-2B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.05635041589438875,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "k_proj",
+    "gate_proj",
+    "up_proj",
+    "o_proj",
+    "down_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-816/chat_template.jinja b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-816/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-816/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is true %}
+        {{- '<think>\n' }}
+    {%- else %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-816/tokenizer_config.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-816/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-816/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-816/trainer_state.json b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-816/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..aafb56dfefab7f93810a88683941f29d1cc3d0e6
--- /dev/null
+++ b/substitutivity_original_Estonian/Qwen3.5-2B-Base_substitutivity_splits_original_features_train_substitutivity_splits_original_features_test2/checkpoint-816/trainer_state.json
@@ -0,0 +1,238 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.0,
+  "eval_steps": 500,
+  "global_step": 816,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.8153233402967452,
+      "epoch": 0.12269938650306748,
+      "grad_norm": 1.8510568141937256,
+      "learning_rate": 4.8469594667113335e-05,
+      "loss": 1.7217323303222656,
+      "mean_token_accuracy": 0.6236635231971741,
+      "num_tokens": 134329.0,
+      "step": 50
+    },
+    {
+      "entropy": 0.97831438601017,
+      "epoch": 0.24539877300613497,
+      "grad_norm": 1.889868974685669,
+      "learning_rate": 9.792836473559631e-05,
+      "loss": 0.9050540924072266,
+      "mean_token_accuracy": 0.7377463465929032,
+      "num_tokens": 262407.0,
+      "step": 100
+    },
+    {
+      "entropy": 0.8743131357431412,
+      "epoch": 0.36809815950920244,
+      "grad_norm": 1.1766732931137085,
+      "learning_rate": 0.0001473871348040793,
+      "loss": 0.8129763031005859,
+      "mean_token_accuracy": 0.7583417356014251,
+      "num_tokens": 401136.0,
+      "step": 150
+    },
+    {
+      "entropy": 0.8273966419696808,
+      "epoch": 0.49079754601226994,
+      "grad_norm": 1.1311014890670776,
+      "learning_rate": 0.00019684590487256229,
+      "loss": 0.7711280822753906,
+      "mean_token_accuracy": 0.7668228060007095,
+      "num_tokens": 531647.0,
+      "step": 200
+    },
+    {
+      "entropy": 0.8101438617706299,
+      "epoch": 0.6134969325153374,
+      "grad_norm": 0.8905950784683228,
+      "learning_rate": 0.0002463046749410453,
+      "loss": 0.7609222412109375,
+      "mean_token_accuracy": 0.7717174577713013,
+      "num_tokens": 666127.0,
+      "step": 250
+    },
+    {
+      "entropy": 0.7885055243968964,
+      "epoch": 0.7361963190184049,
+      "grad_norm": 1.1944245100021362,
+      "learning_rate": 0.00029576344500952824,
+      "loss": 0.7477743530273437,
+      "mean_token_accuracy": 0.773262197971344,
+      "num_tokens": 799292.0,
+      "step": 300
+    },
+    {
+      "entropy": 0.7858527797460556,
+      "epoch": 0.8588957055214724,
+      "grad_norm": 1.1311068534851074,
+      "learning_rate": 0.00034522221507801124,
+      "loss": 0.7411377716064453,
+      "mean_token_accuracy": 0.7745680212974548,
+      "num_tokens": 930614.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.7879797518253326,
+      "epoch": 0.9815950920245399,
+      "grad_norm": 0.9370436072349548,
+      "learning_rate": 0.00039468098514649423,
+      "loss": 0.7507785034179687,
+      "mean_token_accuracy": 0.7747444450855255,
+      "num_tokens": 1061654.0,
+      "step": 400
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.8527693561145238,
+      "eval_mean_token_accuracy": 0.7730878329277039,
+      "eval_not_syn_loss": 0.781921923160553,
+      "eval_not_syn_runtime": 56.9157,
+      "eval_not_syn_samples_per_second": 24.527,
+      "eval_not_syn_steps_per_second": 3.075,
+      "eval_num_tokens": 1080196.0,
+      "step": 408
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.7952005583899362,
+      "eval_mean_token_accuracy": 0.7705784467288426,
+      "eval_num_tokens": 1080196.0,
+      "eval_syn_loss": 0.7461121082305908,
+      "eval_syn_runtime": 59.5758,
+      "eval_syn_samples_per_second": 23.432,
+      "eval_syn_steps_per_second": 2.937,
+      "step": 408
+    },
+    {
+      "entropy": 0.7564875923021875,
+      "epoch": 1.1030674846625768,
+      "grad_norm": 1.1334599256515503,
+      "learning_rate": 0.00040345942966973637,
+      "loss": 0.7208955383300781,
+      "mean_token_accuracy": 0.7788144470465304,
+      "num_tokens": 1199170.0,
+      "step": 450
+    },
+    {
+      "entropy": 0.7427570706605912,
+      "epoch": 1.2257668711656442,
+      "grad_norm": 1.0777021646499634,
+      "learning_rate": 0.00040297229629224153,
+      "loss": 0.7049168395996094,
+      "mean_token_accuracy": 0.7830300116539002,
+      "num_tokens": 1334298.0,
+      "step": 500
+    },
+    {
+      "entropy": 0.7533353108167649,
+      "epoch": 1.3484662576687116,
+      "grad_norm": 1.3439561128616333,
+      "learning_rate": 0.00040211707285041045,
+      "loss": 0.7102288818359375,
+      "mean_token_accuracy": 0.7818449640274048,
+      "num_tokens": 1463189.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.7666342407464981,
+      "epoch": 1.471165644171779,
+      "grad_norm": 1.1329729557037354,
+      "learning_rate": 0.00040089532410439227,
+      "loss": 0.7270996856689453,
+      "mean_token_accuracy": 0.778404277563095,
+      "num_tokens": 1590310.0,
+      "step": 600
+    },
+    {
+      "entropy": 0.7359934949874878,
+      "epoch": 1.5938650306748468,
+      "grad_norm": 1.2227331399917603,
+      "learning_rate": 0.0003993092854276068,
+      "loss": 0.6996555328369141,
+      "mean_token_accuracy": 0.7834958267211914,
+      "num_tokens": 1721836.0,
+      "step": 650
+    },
+    {
+      "entropy": 0.7278932738304138,
+      "epoch": 1.716564417177914,
+      "grad_norm": 0.8753179311752319,
+      "learning_rate": 0.0003973618587167924,
+      "loss": 0.694002456665039,
+      "mean_token_accuracy": 0.7849677371978759,
+      "num_tokens": 1853507.0,
+      "step": 700
+    },
+    {
+      "entropy": 0.7190143716335297,
+      "epoch": 1.8392638036809816,
+      "grad_norm": 0.8524742126464844,
+      "learning_rate": 0.0003950566070825483,
+      "loss": 0.6890143585205079,
+      "mean_token_accuracy": 0.7865350896120071,
+      "num_tokens": 1988113.0,
+      "step": 750
+    },
+    {
+      "entropy": 0.7072037667036056,
+      "epoch": 1.961963190184049,
+      "grad_norm": 0.877348005771637,
+      "learning_rate": 0.00039239774833008714,
+      "loss": 0.6795162200927735,
+      "mean_token_accuracy": 0.7885233855247498,
+      "num_tokens": 2121314.0,
+      "step": 800
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.818787784235818,
+      "eval_mean_token_accuracy": 0.7573730857031686,
+      "eval_not_syn_loss": 0.7486392855644226,
+      "eval_not_syn_runtime": 56.7426,
+      "eval_not_syn_samples_per_second": 24.602,
+      "eval_not_syn_steps_per_second": 3.084,
+      "eval_num_tokens": 2160392.0,
+      "step": 816
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.7632522354807173,
+      "eval_mean_token_accuracy": 0.8101163625717163,
+      "eval_num_tokens": 2160392.0,
+      "eval_syn_loss": 0.6829464435577393,
+      "eval_syn_runtime": 58.5255,
+      "eval_syn_samples_per_second": 23.853,
+      "eval_syn_steps_per_second": 2.99,
+      "step": 816
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 4080,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.410561614273088e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/systematicity_original_Estonian/Qwen3.5-2B-Base_systematicity_splits_original_features_train_systematicity_splits_original_features_test1/checkpoint-914/chat_template.jinja b/systematicity_original_Estonian/Qwen3.5-2B-Base_systematicity_splits_original_features_train_systematicity_splits_original_features_test1/checkpoint-914/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..0ef09f214eaa6d9bca297988afc1454b5827b2c7
--- /dev/null
+++ b/systematicity_original_Estonian/Qwen3.5-2B-Base_systematicity_splits_original_features_train_systematicity_splits_original_features_test1/checkpoint-914/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is true %}
+        {{- '<think>\n' }}
+    {%- else %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/systematicity_original_Estonian/Qwen3.5-2B-Base_systematicity_splits_original_features_train_systematicity_splits_original_features_test1/checkpoint-914/tokenizer_config.json b/systematicity_original_Estonian/Qwen3.5-2B-Base_systematicity_splits_original_features_train_systematicity_splits_original_features_test1/checkpoint-914/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/systematicity_original_Estonian/Qwen3.5-2B-Base_systematicity_splits_original_features_train_systematicity_splits_original_features_test1/checkpoint-914/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/systematicity_original_Estonian/Qwen3.5-2B-Base_systematicity_splits_original_features_train_systematicity_splits_original_features_test1/checkpoint-914/trainer_state.json b/systematicity_original_Estonian/Qwen3.5-2B-Base_systematicity_splits_original_features_train_systematicity_splits_original_features_test1/checkpoint-914/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..67a56b17ccacd3d45157c7fb2fe74cbd86f232fd
--- /dev/null
+++ b/systematicity_original_Estonian/Qwen3.5-2B-Base_systematicity_splits_original_features_train_systematicity_splits_original_features_test1/checkpoint-914/trainer_state.json
@@ -0,0 +1,236 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.0,
+  "eval_steps": 500,
+  "global_step": 914,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.074948468208313,
+      "epoch": 0.10952902519167579,
+      "grad_norm": 2.3511903285980225,
+      "learning_rate": 1.2698308676050528e-05,
+      "loss": 1.9153688049316406,
+      "mean_token_accuracy": 0.6077924159169197,
+      "num_tokens": 113087.0,
+      "step": 50
+    },
+    {
+      "entropy": 1.054056368470192,
+      "epoch": 0.21905805038335158,
+      "grad_norm": 2.3081915378570557,
+      "learning_rate": 2.5655766508755147e-05,
+      "loss": 0.9633240509033203,
+      "mean_token_accuracy": 0.741833764910698,
+      "num_tokens": 230392.0,
+      "step": 100
+    },
+    {
+      "entropy": 0.9184817910194397,
+      "epoch": 0.32858707557502737,
+      "grad_norm": 2.0965280532836914,
+      "learning_rate": 3.861322434145977e-05,
+      "loss": 0.8405924987792969,
+      "mean_token_accuracy": 0.7688284432888031,
+      "num_tokens": 341746.0,
+      "step": 150
+    },
+    {
+      "entropy": 0.8618861585855484,
+      "epoch": 0.43811610076670315,
+      "grad_norm": 2.2506542205810547,
+      "learning_rate": 5.157068217416438e-05,
+      "loss": 0.7841030883789063,
+      "mean_token_accuracy": 0.7796466761827469,
+      "num_tokens": 460483.0,
+      "step": 200
+    },
+    {
+      "entropy": 0.819869134426117,
+      "epoch": 0.547645125958379,
+      "grad_norm": 1.4058395624160767,
+      "learning_rate": 6.4528140006869e-05,
+      "loss": 0.7524736785888672,
+      "mean_token_accuracy": 0.7900535291433335,
+      "num_tokens": 570627.0,
+      "step": 250
+    },
+    {
+      "entropy": 0.8075837278366089,
+      "epoch": 0.6571741511500547,
+      "grad_norm": 1.5815014839172363,
+      "learning_rate": 7.748559783957362e-05,
+      "loss": 0.7388697814941406,
+      "mean_token_accuracy": 0.7914055424928665,
+      "num_tokens": 677217.0,
+      "step": 300
+    },
+    {
+      "entropy": 0.7868322026729584,
+      "epoch": 0.7667031763417306,
+      "grad_norm": 1.2377994060516357,
+      "learning_rate": 9.044305567227824e-05,
+      "loss": 0.7226168060302735,
+      "mean_token_accuracy": 0.7953901988267899,
+      "num_tokens": 790185.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.7526325100660324,
+      "epoch": 0.8762322015334063,
+      "grad_norm": 1.203137755393982,
+      "learning_rate": 0.00010340051350498286,
+      "loss": 0.6991680908203125,
+      "mean_token_accuracy": 0.7992895567417144,
+      "num_tokens": 910080.0,
+      "step": 400
+    },
+    {
+      "entropy": 0.7626404595375061,
+      "epoch": 0.9857612267250822,
+      "grad_norm": 1.1625452041625977,
+      "learning_rate": 0.00011635797133768749,
+      "loss": 0.7055020904541016,
+      "mean_token_accuracy": 0.7997505795955658,
+      "num_tokens": 1023473.0,
+      "step": 450
+    },
+    {
+      "epoch": 1.0,
+      "eval_entropy": 0.7441067624659765,
+      "eval_loss": 0.7209385633468628,
+      "eval_mean_token_accuracy": 0.7895652093584575,
+      "eval_num_tokens": 1038250.0,
+      "eval_runtime": 45.5076,
+      "eval_samples_per_second": 22.018,
+      "eval_steps_per_second": 2.769,
+      "step": 457
+    },
+    {
+      "entropy": 0.720724637460227,
+      "epoch": 1.0941949616648412,
+      "grad_norm": 1.2537180185317993,
+      "learning_rate": 0.00011840069618947894,
+      "loss": 0.6586117553710937,
+      "mean_token_accuracy": 0.8081232917429221,
+      "num_tokens": 1128922.0,
+      "step": 500
+    },
+    {
+      "entropy": 0.7084310132265091,
+      "epoch": 1.203723986856517,
+      "grad_norm": 1.2347161769866943,
+      "learning_rate": 0.00011828501915154198,
+      "loss": 0.650611572265625,
+      "mean_token_accuracy": 0.808387525677681,
+      "num_tokens": 1245635.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.706004958152771,
+      "epoch": 1.3132530120481927,
+      "grad_norm": 1.2140169143676758,
+      "learning_rate": 0.00011808319665683977,
+      "loss": 0.6442465209960937,
+      "mean_token_accuracy": 0.8122780376672745,
+      "num_tokens": 1354885.0,
+      "step": 600
+    },
+    {
+      "entropy": 0.6854291236400605,
+      "epoch": 1.4227820372398685,
+      "grad_norm": 1.1137559413909912,
+      "learning_rate": 0.00011779552303848118,
+      "loss": 0.6293138885498046,
+      "mean_token_accuracy": 0.8158071845769882,
+      "num_tokens": 1472232.0,
+      "step": 650
+    },
+    {
+      "entropy": 0.6885070586204529,
+      "epoch": 1.5323110624315444,
+      "grad_norm": 1.1074714660644531,
+      "learning_rate": 0.00011742241783280468,
+      "loss": 0.6259001541137695,
+      "mean_token_accuracy": 0.8164563828706741,
+      "num_tokens": 1587494.0,
+      "step": 700
+    },
+    {
+      "entropy": 0.6932114177942276,
+      "epoch": 1.6418400876232202,
+      "grad_norm": 1.1099495887756348,
+      "learning_rate": 0.00011696442516753647,
+      "loss": 0.6335408401489258,
+      "mean_token_accuracy": 0.8156683903932571,
+      "num_tokens": 1696927.0,
+      "step": 750
+    },
+    {
+      "entropy": 0.6787053138017655,
+      "epoch": 1.751369112814896,
+      "grad_norm": 1.2505418062210083,
+      "learning_rate": 0.00011642221296824764,
+      "loss": 0.6210909271240235,
+      "mean_token_accuracy": 0.8153067737817764,
+      "num_tokens": 1813415.0,
+      "step": 800
+    },
+    {
+      "entropy": 0.6693948286771775,
+      "epoch": 1.8608981380065717,
+      "grad_norm": 1.0963624715805054,
+      "learning_rate": 0.00011579657198426736,
+      "loss": 0.6105814361572266,
+      "mean_token_accuracy": 0.819609425663948,
+      "num_tokens": 1924911.0,
+      "step": 850
+    },
+    {
+      "entropy": 0.6634757363796234,
+      "epoch": 1.9704271631982475,
+      "grad_norm": 0.8766036033630371,
+      "learning_rate": 0.00011508841463547313,
+      "loss": 0.6100498580932617,
+      "mean_token_accuracy": 0.8197187691926956,
+      "num_tokens": 2044650.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.0,
+      "eval_entropy": 0.6342300275961558,
+      "eval_loss": 0.6517631411552429,
+      "eval_mean_token_accuracy": 0.807878240233376,
+      "eval_num_tokens": 2076500.0,
+      "eval_runtime": 44.6225,
+      "eval_samples_per_second": 22.455,
+      "eval_steps_per_second": 2.824,
+      "step": 914
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 4570,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.486683349186176e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}