Add files using upload-large-folder tool

eval_results_ood/global_step_0/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_0_actor_huggingface/results_2025-08-29T11-17-46.747738.json

@@ -0,0 +1,141 @@
+{
+  "config_general": {
+    "lighteval_sha": "327071fe86e427d880f907a51d1462f4a3f951c1",
+    "num_fewshot_seeds": 1,
+    "max_samples": null,
+    "job_id": 0,
+    "start_time": 18883988.60784555,
+    "end_time": 18884586.726209104,
+    "total_evaluation_time_secondes": "598.1183635555208",
+    "model_name": "_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_0_actor_huggingface",
+    "model_sha": "",
+    "model_dtype": null,
+    "model_size": null,
+    "generation_parameters": {
+      "early_stopping": null,
+      "repetition_penalty": null,
+      "frequency_penalty": null,
+      "length_penalty": null,
+      "presence_penalty": null,
+      "max_new_tokens": 32768,
+      "min_new_tokens": null,
+      "seed": null,
+      "stop_tokens": null,
+      "temperature": 1.0,
+      "top_k": null,
+      "min_p": null,
+      "top_p": 0.95,
+      "truncate_prompt": null,
+      "response_format": null
+    }
+  },
+  "results": {
+    "extended|ifeval|0": {
+      "prompt_level_strict_acc": 0.39371534195933455,
+      "prompt_level_strict_acc_stderr": 0.02102483414587237,
+      "inst_level_strict_acc": 0.5215827338129496,
+      "inst_level_strict_acc_stderr": 0.0005599264307527568,
+      "prompt_level_loose_acc": 0.4288354898336414,
+      "prompt_level_loose_acc_stderr": 0.021297522569050743,
+      "inst_level_loose_acc": 0.5563549160671463,
+      "inst_level_loose_acc_stderr": 0.0005527714546864814
+    },
+    "all": {
+      "prompt_level_strict_acc": 0.39371534195933455,
+      "prompt_level_strict_acc_stderr": 0.02102483414587237,
+      "inst_level_strict_acc": 0.5215827338129496,
+      "inst_level_strict_acc_stderr": 0.0005599264307527568,
+      "prompt_level_loose_acc": 0.4288354898336414,
+      "prompt_level_loose_acc_stderr": 0.021297522569050743,
+      "inst_level_loose_acc": 0.5563549160671463,
+      "inst_level_loose_acc_stderr": 0.0005527714546864814
+    }
+  },
+  "versions": {
+    "extended|ifeval|0": "0.1"
+  },
+  "config_tasks": {
+    "extended|ifeval": {
+      "name": "ifeval",
+      "prompt_function": "ifeval_prompt",
+      "hf_repo": "google/IFEval",
+      "hf_subset": "default",
+      "metric": [
+        {
+          "metric_name": [
+            "prompt_level_strict_acc",
+            "inst_level_strict_acc",
+            "prompt_level_loose_acc",
+            "inst_level_loose_acc"
+          ],
+          "higher_is_better": {
+            "prompt_level_strict_acc": true,
+            "inst_level_strict_acc": true,
+            "prompt_level_loose_acc": true,
+            "inst_level_loose_acc": true
+          },
+          "category": "3",
+          "use_case": "1",
+          "sample_level_fn": "ifeval_metric",
+          "corpus_level_fn": {
+            "prompt_level_strict_acc": "mean",
+            "inst_level_strict_acc": "agg_inst_level_acc",
+            "prompt_level_loose_acc": "mean",
+            "inst_level_loose_acc": "agg_inst_level_acc"
+          }
+        }
+      ],
+      "hf_revision": null,
+      "hf_filter": null,
+      "hf_avail_splits": [
+        "train"
+      ],
+      "trust_dataset": false,
+      "evaluation_splits": [
+        "train"
+      ],
+      "few_shots_split": "train",
+      "few_shots_select": "random_sampling",
+      "generation_size": 1280,
+      "generation_grammar": null,
+      "stop_sequence": [],
+      "num_samples": null,
+      "suite": [
+        "extended"
+      ],
+      "original_num_docs": 541,
+      "effective_num_docs": 541,
+      "must_remove_duplicate_docs": false,
+      "version": "0.1"
+    }
+  },
+  "summary_tasks": {
+    "extended|ifeval|0": {
+      "hashes": {
+        "hash_examples": "e99cbf567588d7c6",
+        "hash_full_prompts": "ad10e21c9427089f",
+        "hash_input_tokens": "76e5046dde448ac8",
+        "hash_cont_tokens": "2fd4382748c56918"
+      },
+      "truncated": 0,
+      "non_truncated": 541,
+      "padded": 0,
+      "non_padded": 541,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    }
+  },
+  "summary_general": {
+    "hashes": {
+      "hash_examples": "ea046ab2c6fc5928",
+      "hash_full_prompts": "edca08807dd5a8ec",
+      "hash_input_tokens": "6db7fa56ea148d89",
+      "hash_cont_tokens": "a571a26c128a03d4"
+    },
+    "truncated": 0,
+    "non_truncated": 541,
+    "padded": 0,
+    "non_padded": 541,
+    "num_truncated_few_shots": 0
+  }
+}

eval_results_ood/global_step_0/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_0_actor_huggingface/results_2025-09-23T15-20-09.172749.json

@@ -0,0 +1,141 @@
+{
+  "config_general": {
+    "lighteval_sha": "327071fe86e427d880f907a51d1462f4a3f951c1",
+    "num_fewshot_seeds": 1,
+    "max_samples": null,
+    "job_id": 0,
+    "start_time": 2819894.465284853,
+    "end_time": 2820247.173341552,
+    "total_evaluation_time_secondes": "352.7080566990189",
+    "model_name": "_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_0_actor_huggingface",
+    "model_sha": "",
+    "model_dtype": null,
+    "model_size": null,
+    "generation_parameters": {
+      "early_stopping": null,
+      "repetition_penalty": null,
+      "frequency_penalty": null,
+      "length_penalty": null,
+      "presence_penalty": null,
+      "max_new_tokens": 32768,
+      "min_new_tokens": null,
+      "seed": null,
+      "stop_tokens": null,
+      "temperature": 1.0,
+      "top_k": null,
+      "min_p": null,
+      "top_p": 0.95,
+      "truncate_prompt": null,
+      "response_format": null
+    }
+  },
+  "results": {
+    "extended|ifeval|0": {
+      "prompt_level_strict_acc": 0.39371534195933455,
+      "prompt_level_strict_acc_stderr": 0.02102483414587237,
+      "inst_level_strict_acc": 0.5191846522781774,
+      "inst_level_strict_acc_stderr": 0.0005633834275419194,
+      "prompt_level_loose_acc": 0.4399260628465804,
+      "prompt_level_loose_acc_stderr": 0.021360708220801976,
+      "inst_level_loose_acc": 0.564748201438849,
+      "inst_level_loose_acc_stderr": 0.0005587046293737301
+    },
+    "all": {
+      "prompt_level_strict_acc": 0.39371534195933455,
+      "prompt_level_strict_acc_stderr": 0.02102483414587237,
+      "inst_level_strict_acc": 0.5191846522781774,
+      "inst_level_strict_acc_stderr": 0.0005633834275419194,
+      "prompt_level_loose_acc": 0.4399260628465804,
+      "prompt_level_loose_acc_stderr": 0.021360708220801976,
+      "inst_level_loose_acc": 0.564748201438849,
+      "inst_level_loose_acc_stderr": 0.0005587046293737301
+    }
+  },
+  "versions": {
+    "extended|ifeval|0": "0.1"
+  },
+  "config_tasks": {
+    "extended|ifeval": {
+      "name": "ifeval",
+      "prompt_function": "ifeval_prompt",
+      "hf_repo": "google/IFEval",
+      "hf_subset": "default",
+      "metric": [
+        {
+          "metric_name": [
+            "prompt_level_strict_acc",
+            "inst_level_strict_acc",
+            "prompt_level_loose_acc",
+            "inst_level_loose_acc"
+          ],
+          "higher_is_better": {
+            "prompt_level_strict_acc": true,
+            "inst_level_strict_acc": true,
+            "prompt_level_loose_acc": true,
+            "inst_level_loose_acc": true
+          },
+          "category": "3",
+          "use_case": "1",
+          "sample_level_fn": "ifeval_metric",
+          "corpus_level_fn": {
+            "prompt_level_strict_acc": "mean",
+            "inst_level_strict_acc": "agg_inst_level_acc",
+            "prompt_level_loose_acc": "mean",
+            "inst_level_loose_acc": "agg_inst_level_acc"
+          }
+        }
+      ],
+      "hf_revision": null,
+      "hf_filter": null,
+      "hf_avail_splits": [
+        "train"
+      ],
+      "trust_dataset": false,
+      "evaluation_splits": [
+        "train"
+      ],
+      "few_shots_split": "train",
+      "few_shots_select": "random_sampling",
+      "generation_size": 1280,
+      "generation_grammar": null,
+      "stop_sequence": [],
+      "num_samples": null,
+      "suite": [
+        "extended"
+      ],
+      "original_num_docs": 541,
+      "effective_num_docs": 541,
+      "must_remove_duplicate_docs": false,
+      "version": "0.1"
+    }
+  },
+  "summary_tasks": {
+    "extended|ifeval|0": {
+      "hashes": {
+        "hash_examples": "e99cbf567588d7c6",
+        "hash_full_prompts": "3972e282d6494c90",
+        "hash_input_tokens": "f702e1af092e9cc3",
+        "hash_cont_tokens": "f037b3f92aa6d818"
+      },
+      "truncated": 0,
+      "non_truncated": 541,
+      "padded": 0,
+      "non_padded": 541,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    }
+  },
+  "summary_general": {
+    "hashes": {
+      "hash_examples": "ea046ab2c6fc5928",
+      "hash_full_prompts": "b1fd186794cb0bb3",
+      "hash_input_tokens": "5fc219976d10c06f",
+      "hash_cont_tokens": "8783819c7352f952"
+    },
+    "truncated": 0,
+    "non_truncated": 541,
+    "padded": 0,
+    "non_padded": 541,
+    "num_truncated_few_shots": 0
+  }
+}

eval_results_ood/global_step_10/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_10_actor_huggingface/results_2025-09-23T15-19-07.845971.json

@@ -0,0 +1,141 @@
+{
+  "config_general": {
+    "lighteval_sha": "327071fe86e427d880f907a51d1462f4a3f951c1",
+    "num_fewshot_seeds": 1,
+    "max_samples": null,
+    "job_id": 0,
+    "start_time": 2819894.524622857,
+    "end_time": 2820185.260255099,
+    "total_evaluation_time_secondes": "290.73563224170357",
+    "model_name": "_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_10_actor_huggingface",
+    "model_sha": "",
+    "model_dtype": null,
+    "model_size": null,
+    "generation_parameters": {
+      "early_stopping": null,
+      "repetition_penalty": null,
+      "frequency_penalty": null,
+      "length_penalty": null,
+      "presence_penalty": null,
+      "max_new_tokens": 32768,
+      "min_new_tokens": null,
+      "seed": null,
+      "stop_tokens": null,
+      "temperature": 1.0,
+      "top_k": null,
+      "min_p": null,
+      "top_p": 0.95,
+      "truncate_prompt": null,
+      "response_format": null
+    }
+  },
+  "results": {
+    "extended|ifeval|0": {
+      "prompt_level_strict_acc": 0.39186691312384475,
+      "prompt_level_strict_acc_stderr": 0.021007372406188354,
+      "inst_level_strict_acc": 0.5143884892086331,
+      "inst_level_strict_acc_stderr": 0.0005407156424898863,
+      "prompt_level_loose_acc": 0.43253234750462105,
+      "prompt_level_loose_acc_stderr": 0.02131979239817807,
+      "inst_level_loose_acc": 0.552757793764988,
+      "inst_level_loose_acc_stderr": 0.0005395531273954375
+    },
+    "all": {
+      "prompt_level_strict_acc": 0.39186691312384475,
+      "prompt_level_strict_acc_stderr": 0.021007372406188354,
+      "inst_level_strict_acc": 0.5143884892086331,
+      "inst_level_strict_acc_stderr": 0.0005407156424898863,
+      "prompt_level_loose_acc": 0.43253234750462105,
+      "prompt_level_loose_acc_stderr": 0.02131979239817807,
+      "inst_level_loose_acc": 0.552757793764988,
+      "inst_level_loose_acc_stderr": 0.0005395531273954375
+    }
+  },
+  "versions": {
+    "extended|ifeval|0": "0.1"
+  },
+  "config_tasks": {
+    "extended|ifeval": {
+      "name": "ifeval",
+      "prompt_function": "ifeval_prompt",
+      "hf_repo": "google/IFEval",
+      "hf_subset": "default",
+      "metric": [
+        {
+          "metric_name": [
+            "prompt_level_strict_acc",
+            "inst_level_strict_acc",
+            "prompt_level_loose_acc",
+            "inst_level_loose_acc"
+          ],
+          "higher_is_better": {
+            "prompt_level_strict_acc": true,
+            "inst_level_strict_acc": true,
+            "prompt_level_loose_acc": true,
+            "inst_level_loose_acc": true
+          },
+          "category": "3",
+          "use_case": "1",
+          "sample_level_fn": "ifeval_metric",
+          "corpus_level_fn": {
+            "prompt_level_strict_acc": "mean",
+            "inst_level_strict_acc": "agg_inst_level_acc",
+            "prompt_level_loose_acc": "mean",
+            "inst_level_loose_acc": "agg_inst_level_acc"
+          }
+        }
+      ],
+      "hf_revision": null,
+      "hf_filter": null,
+      "hf_avail_splits": [
+        "train"
+      ],
+      "trust_dataset": false,
+      "evaluation_splits": [
+        "train"
+      ],
+      "few_shots_split": "train",
+      "few_shots_select": "random_sampling",
+      "generation_size": 1280,
+      "generation_grammar": null,
+      "stop_sequence": [],
+      "num_samples": null,
+      "suite": [
+        "extended"
+      ],
+      "original_num_docs": 541,
+      "effective_num_docs": 541,
+      "must_remove_duplicate_docs": false,
+      "version": "0.1"
+    }
+  },
+  "summary_tasks": {
+    "extended|ifeval|0": {
+      "hashes": {
+        "hash_examples": "e99cbf567588d7c6",
+        "hash_full_prompts": "3972e282d6494c90",
+        "hash_input_tokens": "f702e1af092e9cc3",
+        "hash_cont_tokens": "ca5820a7f4067bfa"
+      },
+      "truncated": 0,
+      "non_truncated": 541,
+      "padded": 0,
+      "non_padded": 541,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    }
+  },
+  "summary_general": {
+    "hashes": {
+      "hash_examples": "ea046ab2c6fc5928",
+      "hash_full_prompts": "b1fd186794cb0bb3",
+      "hash_input_tokens": "5fc219976d10c06f",
+      "hash_cont_tokens": "4be167cbb4c1d0a3"
+    },
+    "truncated": 0,
+    "non_truncated": 541,
+    "padded": 0,
+    "non_padded": 541,
+    "num_truncated_few_shots": 0
+  }
+}

eval_results_ood/global_step_15/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_15_actor_huggingface/results_2025-08-29T11-16-24.695422.json

@@ -0,0 +1,141 @@
+{
+  "config_general": {
+    "lighteval_sha": "327071fe86e427d880f907a51d1462f4a3f951c1",
+    "num_fewshot_seeds": 1,
+    "max_samples": null,
+    "job_id": 0,
+    "start_time": 18883988.62186077,
+    "end_time": 18884503.508060712,
+    "total_evaluation_time_secondes": "514.8861999437213",
+    "model_name": "_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_15_actor_huggingface",
+    "model_sha": "",
+    "model_dtype": null,
+    "model_size": null,
+    "generation_parameters": {
+      "early_stopping": null,
+      "repetition_penalty": null,
+      "frequency_penalty": null,
+      "length_penalty": null,
+      "presence_penalty": null,
+      "max_new_tokens": 32768,
+      "min_new_tokens": null,
+      "seed": null,
+      "stop_tokens": null,
+      "temperature": 1.0,
+      "top_k": null,
+      "min_p": null,
+      "top_p": 0.95,
+      "truncate_prompt": null,
+      "response_format": null
+    }
+  },
+  "results": {
+    "extended|ifeval|0": {
+      "prompt_level_strict_acc": 0.3752310536044362,
+      "prompt_level_strict_acc_stderr": 0.020835898065035694,
+      "inst_level_strict_acc": 0.5095923261390888,
+      "inst_level_strict_acc_stderr": 0.0005432302541010317,
+      "prompt_level_loose_acc": 0.4269870609981516,
+      "prompt_level_loose_acc_stderr": 0.02128593305006131,
+      "inst_level_loose_acc": 0.5587529976019184,
+      "inst_level_loose_acc_stderr": 0.000545503372755182
+    },
+    "all": {
+      "prompt_level_strict_acc": 0.3752310536044362,
+      "prompt_level_strict_acc_stderr": 0.020835898065035694,
+      "inst_level_strict_acc": 0.5095923261390888,
+      "inst_level_strict_acc_stderr": 0.0005432302541010317,
+      "prompt_level_loose_acc": 0.4269870609981516,
+      "prompt_level_loose_acc_stderr": 0.02128593305006131,
+      "inst_level_loose_acc": 0.5587529976019184,
+      "inst_level_loose_acc_stderr": 0.000545503372755182
+    }
+  },
+  "versions": {
+    "extended|ifeval|0": "0.1"
+  },
+  "config_tasks": {
+    "extended|ifeval": {
+      "name": "ifeval",
+      "prompt_function": "ifeval_prompt",
+      "hf_repo": "google/IFEval",
+      "hf_subset": "default",
+      "metric": [
+        {
+          "metric_name": [
+            "prompt_level_strict_acc",
+            "inst_level_strict_acc",
+            "prompt_level_loose_acc",
+            "inst_level_loose_acc"
+          ],
+          "higher_is_better": {
+            "prompt_level_strict_acc": true,
+            "inst_level_strict_acc": true,
+            "prompt_level_loose_acc": true,
+            "inst_level_loose_acc": true
+          },
+          "category": "3",
+          "use_case": "1",
+          "sample_level_fn": "ifeval_metric",
+          "corpus_level_fn": {
+            "prompt_level_strict_acc": "mean",
+            "inst_level_strict_acc": "agg_inst_level_acc",
+            "prompt_level_loose_acc": "mean",
+            "inst_level_loose_acc": "agg_inst_level_acc"
+          }
+        }
+      ],
+      "hf_revision": null,
+      "hf_filter": null,
+      "hf_avail_splits": [
+        "train"
+      ],
+      "trust_dataset": false,
+      "evaluation_splits": [
+        "train"
+      ],
+      "few_shots_split": "train",
+      "few_shots_select": "random_sampling",
+      "generation_size": 1280,
+      "generation_grammar": null,
+      "stop_sequence": [],
+      "num_samples": null,
+      "suite": [
+        "extended"
+      ],
+      "original_num_docs": 541,
+      "effective_num_docs": 541,
+      "must_remove_duplicate_docs": false,
+      "version": "0.1"
+    }
+  },
+  "summary_tasks": {
+    "extended|ifeval|0": {
+      "hashes": {
+        "hash_examples": "e99cbf567588d7c6",
+        "hash_full_prompts": "ad10e21c9427089f",
+        "hash_input_tokens": "76e5046dde448ac8",
+        "hash_cont_tokens": "af1056c06f4272cc"
+      },
+      "truncated": 0,
+      "non_truncated": 541,
+      "padded": 0,
+      "non_padded": 541,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    }
+  },
+  "summary_general": {
+    "hashes": {
+      "hash_examples": "ea046ab2c6fc5928",
+      "hash_full_prompts": "edca08807dd5a8ec",
+      "hash_input_tokens": "6db7fa56ea148d89",
+      "hash_cont_tokens": "621f1d09b05aface"
+    },
+    "truncated": 0,
+    "non_truncated": 541,
+    "padded": 0,
+    "non_padded": 541,
+    "num_truncated_few_shots": 0
+  }
+}

eval_results_ood/global_step_15/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_15_actor_huggingface/results_2025-09-23T15-18-58.766285.json

@@ -0,0 +1,141 @@
+{
+  "config_general": {
+    "lighteval_sha": "327071fe86e427d880f907a51d1462f4a3f951c1",
+    "num_fewshot_seeds": 1,
+    "max_samples": null,
+    "job_id": 0,
+    "start_time": 2819894.516623436,
+    "end_time": 2820176.0063693,
+    "total_evaluation_time_secondes": "281.489745864179",
+    "model_name": "_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_15_actor_huggingface",
+    "model_sha": "",
+    "model_dtype": null,
+    "model_size": null,
+    "generation_parameters": {
+      "early_stopping": null,
+      "repetition_penalty": null,
+      "frequency_penalty": null,
+      "length_penalty": null,
+      "presence_penalty": null,
+      "max_new_tokens": 32768,
+      "min_new_tokens": null,
+      "seed": null,
+      "stop_tokens": null,
+      "temperature": 1.0,
+      "top_k": null,
+      "min_p": null,
+      "top_p": 0.95,
+      "truncate_prompt": null,
+      "response_format": null
+    }
+  },
+  "results": {
+    "extended|ifeval|0": {
+      "prompt_level_strict_acc": 0.37707948243992606,
+      "prompt_level_strict_acc_stderr": 0.020856233918528456,
+      "inst_level_strict_acc": 0.5155875299760192,
+      "inst_level_strict_acc_stderr": 0.0005624364712452048,
+      "prompt_level_loose_acc": 0.4343807763401109,
+      "prompt_level_loose_acc_stderr": 0.021330473657564727,
+      "inst_level_loose_acc": 0.5671462829736211,
+      "inst_level_loose_acc_stderr": 0.0005536778232943271
+    },
+    "all": {
+      "prompt_level_strict_acc": 0.37707948243992606,
+      "prompt_level_strict_acc_stderr": 0.020856233918528456,
+      "inst_level_strict_acc": 0.5155875299760192,
+      "inst_level_strict_acc_stderr": 0.0005624364712452048,
+      "prompt_level_loose_acc": 0.4343807763401109,
+      "prompt_level_loose_acc_stderr": 0.021330473657564727,
+      "inst_level_loose_acc": 0.5671462829736211,
+      "inst_level_loose_acc_stderr": 0.0005536778232943271
+    }
+  },
+  "versions": {
+    "extended|ifeval|0": "0.1"
+  },
+  "config_tasks": {
+    "extended|ifeval": {
+      "name": "ifeval",
+      "prompt_function": "ifeval_prompt",
+      "hf_repo": "google/IFEval",
+      "hf_subset": "default",
+      "metric": [
+        {
+          "metric_name": [
+            "prompt_level_strict_acc",
+            "inst_level_strict_acc",
+            "prompt_level_loose_acc",
+            "inst_level_loose_acc"
+          ],
+          "higher_is_better": {
+            "prompt_level_strict_acc": true,
+            "inst_level_strict_acc": true,
+            "prompt_level_loose_acc": true,
+            "inst_level_loose_acc": true
+          },
+          "category": "3",
+          "use_case": "1",
+          "sample_level_fn": "ifeval_metric",
+          "corpus_level_fn": {
+            "prompt_level_strict_acc": "mean",
+            "inst_level_strict_acc": "agg_inst_level_acc",
+            "prompt_level_loose_acc": "mean",
+            "inst_level_loose_acc": "agg_inst_level_acc"
+          }
+        }
+      ],
+      "hf_revision": null,
+      "hf_filter": null,
+      "hf_avail_splits": [
+        "train"
+      ],
+      "trust_dataset": false,
+      "evaluation_splits": [
+        "train"
+      ],
+      "few_shots_split": "train",
+      "few_shots_select": "random_sampling",
+      "generation_size": 1280,
+      "generation_grammar": null,
+      "stop_sequence": [],
+      "num_samples": null,
+      "suite": [
+        "extended"
+      ],
+      "original_num_docs": 541,
+      "effective_num_docs": 541,
+      "must_remove_duplicate_docs": false,
+      "version": "0.1"
+    }
+  },
+  "summary_tasks": {
+    "extended|ifeval|0": {
+      "hashes": {
+        "hash_examples": "e99cbf567588d7c6",
+        "hash_full_prompts": "3972e282d6494c90",
+        "hash_input_tokens": "f702e1af092e9cc3",
+        "hash_cont_tokens": "b8a7398ee3d1035f"
+      },
+      "truncated": 0,
+      "non_truncated": 541,
+      "padded": 0,
+      "non_padded": 541,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    }
+  },
+  "summary_general": {
+    "hashes": {
+      "hash_examples": "ea046ab2c6fc5928",
+      "hash_full_prompts": "b1fd186794cb0bb3",
+      "hash_input_tokens": "5fc219976d10c06f",
+      "hash_cont_tokens": "aaa3fa41573a6ce4"
+    },
+    "truncated": 0,
+    "non_truncated": 541,
+    "padded": 0,
+    "non_padded": 541,
+    "num_truncated_few_shots": 0
+  }
+}

eval_results_ood/global_step_20/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_20_actor_huggingface/results_2025-08-29T11-16-25.542329.json

@@ -0,0 +1,141 @@
+{
+  "config_general": {
+    "lighteval_sha": "327071fe86e427d880f907a51d1462f4a3f951c1",
+    "num_fewshot_seeds": 1,
+    "max_samples": null,
+    "job_id": 0,
+    "start_time": 18883988.60164322,
+    "end_time": 18884503.444405336,
+    "total_evaluation_time_secondes": "514.8427621163428",
+    "model_name": "_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_20_actor_huggingface",
+    "model_sha": "",
+    "model_dtype": null,
+    "model_size": null,
+    "generation_parameters": {
+      "early_stopping": null,
+      "repetition_penalty": null,
+      "frequency_penalty": null,
+      "length_penalty": null,
+      "presence_penalty": null,
+      "max_new_tokens": 32768,
+      "min_new_tokens": null,
+      "seed": null,
+      "stop_tokens": null,
+      "temperature": 1.0,
+      "top_k": null,
+      "min_p": null,
+      "top_p": 0.95,
+      "truncate_prompt": null,
+      "response_format": null
+    }
+  },
+  "results": {
+    "extended|ifeval|0": {
+      "prompt_level_strict_acc": 0.3807763401109057,
+      "prompt_level_strict_acc_stderr": 0.020895937888190833,
+      "inst_level_strict_acc": 0.511990407673861,
+      "inst_level_strict_acc_stderr": 0.0005284703064941477,
+      "prompt_level_loose_acc": 0.43807763401109057,
+      "prompt_level_loose_acc_stderr": 0.021350931135490935,
+      "inst_level_loose_acc": 0.5599520383693045,
+      "inst_level_loose_acc_stderr": 0.0005508806104783129
+    },
+    "all": {
+      "prompt_level_strict_acc": 0.3807763401109057,
+      "prompt_level_strict_acc_stderr": 0.020895937888190833,
+      "inst_level_strict_acc": 0.511990407673861,
+      "inst_level_strict_acc_stderr": 0.0005284703064941477,
+      "prompt_level_loose_acc": 0.43807763401109057,
+      "prompt_level_loose_acc_stderr": 0.021350931135490935,
+      "inst_level_loose_acc": 0.5599520383693045,
+      "inst_level_loose_acc_stderr": 0.0005508806104783129
+    }
+  },
+  "versions": {
+    "extended|ifeval|0": "0.1"
+  },
+  "config_tasks": {
+    "extended|ifeval": {
+      "name": "ifeval",
+      "prompt_function": "ifeval_prompt",
+      "hf_repo": "google/IFEval",
+      "hf_subset": "default",
+      "metric": [
+        {
+          "metric_name": [
+            "prompt_level_strict_acc",
+            "inst_level_strict_acc",
+            "prompt_level_loose_acc",
+            "inst_level_loose_acc"
+          ],
+          "higher_is_better": {
+            "prompt_level_strict_acc": true,
+            "inst_level_strict_acc": true,
+            "prompt_level_loose_acc": true,
+            "inst_level_loose_acc": true
+          },
+          "category": "3",
+          "use_case": "1",
+          "sample_level_fn": "ifeval_metric",
+          "corpus_level_fn": {
+            "prompt_level_strict_acc": "mean",
+            "inst_level_strict_acc": "agg_inst_level_acc",
+            "prompt_level_loose_acc": "mean",
+            "inst_level_loose_acc": "agg_inst_level_acc"
+          }
+        }
+      ],
+      "hf_revision": null,
+      "hf_filter": null,
+      "hf_avail_splits": [
+        "train"
+      ],
+      "trust_dataset": false,
+      "evaluation_splits": [
+        "train"
+      ],
+      "few_shots_split": "train",
+      "few_shots_select": "random_sampling",
+      "generation_size": 1280,
+      "generation_grammar": null,
+      "stop_sequence": [],
+      "num_samples": null,
+      "suite": [
+        "extended"
+      ],
+      "original_num_docs": 541,
+      "effective_num_docs": 541,
+      "must_remove_duplicate_docs": false,
+      "version": "0.1"
+    }
+  },
+  "summary_tasks": {
+    "extended|ifeval|0": {
+      "hashes": {
+        "hash_examples": "e99cbf567588d7c6",
+        "hash_full_prompts": "ad10e21c9427089f",
+        "hash_input_tokens": "76e5046dde448ac8",
+        "hash_cont_tokens": "2c692a9c51322e27"
+      },
+      "truncated": 0,
+      "non_truncated": 541,
+      "padded": 0,
+      "non_padded": 541,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    }
+  },
+  "summary_general": {
+    "hashes": {
+      "hash_examples": "ea046ab2c6fc5928",
+      "hash_full_prompts": "edca08807dd5a8ec",
+      "hash_input_tokens": "6db7fa56ea148d89",
+      "hash_cont_tokens": "56179f027b3dfdb2"
+    },
+    "truncated": 0,
+    "non_truncated": 541,
+    "padded": 0,
+    "non_padded": 541,
+    "num_truncated_few_shots": 0
+  }
+}

eval_results_ood/global_step_20/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_20_actor_huggingface/results_2025-09-23T15-20-14.009200.json

@@ -0,0 +1,141 @@
+{
+  "config_general": {
+    "lighteval_sha": "327071fe86e427d880f907a51d1462f4a3f951c1",
+    "num_fewshot_seeds": 1,
+    "max_samples": null,
+    "job_id": 0,
+    "start_time": 2819894.464804047,
+    "end_time": 2820250.597080663,
+    "total_evaluation_time_secondes": "356.13227661605924",
+    "model_name": "_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_20_actor_huggingface",
+    "model_sha": "",
+    "model_dtype": null,
+    "model_size": null,
+    "generation_parameters": {
+      "early_stopping": null,
+      "repetition_penalty": null,
+      "frequency_penalty": null,
+      "length_penalty": null,
+      "presence_penalty": null,
+      "max_new_tokens": 32768,
+      "min_new_tokens": null,
+      "seed": null,
+      "stop_tokens": null,
+      "temperature": 1.0,
+      "top_k": null,
+      "min_p": null,
+      "top_p": 0.95,
+      "truncate_prompt": null,
+      "response_format": null
+    }
+  },
+  "results": {
+    "extended|ifeval|0": {
+      "prompt_level_strict_acc": 0.4011090573012939,
+      "prompt_level_strict_acc_stderr": 0.02109153689552074,
+      "inst_level_strict_acc": 0.5143884892086331,
+      "inst_level_strict_acc_stderr": 0.0005584775478161843,
+      "prompt_level_loose_acc": 0.4510166358595194,
+      "prompt_level_loose_acc_stderr": 0.02141307276535979,
+      "inst_level_loose_acc": 0.5635491606714629,
+      "inst_level_loose_acc_stderr": 0.0005514767622336946
+    },
+    "all": {
+      "prompt_level_strict_acc": 0.4011090573012939,
+      "prompt_level_strict_acc_stderr": 0.02109153689552074,
+      "inst_level_strict_acc": 0.5143884892086331,
+      "inst_level_strict_acc_stderr": 0.0005584775478161843,
+      "prompt_level_loose_acc": 0.4510166358595194,
+      "prompt_level_loose_acc_stderr": 0.02141307276535979,
+      "inst_level_loose_acc": 0.5635491606714629,
+      "inst_level_loose_acc_stderr": 0.0005514767622336946
+    }
+  },
+  "versions": {
+    "extended|ifeval|0": "0.1"
+  },
+  "config_tasks": {
+    "extended|ifeval": {
+      "name": "ifeval",
+      "prompt_function": "ifeval_prompt",
+      "hf_repo": "google/IFEval",
+      "hf_subset": "default",
+      "metric": [
+        {
+          "metric_name": [
+            "prompt_level_strict_acc",
+            "inst_level_strict_acc",
+            "prompt_level_loose_acc",
+            "inst_level_loose_acc"
+          ],
+          "higher_is_better": {
+            "prompt_level_strict_acc": true,
+            "inst_level_strict_acc": true,
+            "prompt_level_loose_acc": true,
+            "inst_level_loose_acc": true
+          },
+          "category": "3",
+          "use_case": "1",
+          "sample_level_fn": "ifeval_metric",
+          "corpus_level_fn": {
+            "prompt_level_strict_acc": "mean",
+            "inst_level_strict_acc": "agg_inst_level_acc",
+            "prompt_level_loose_acc": "mean",
+            "inst_level_loose_acc": "agg_inst_level_acc"
+          }
+        }
+      ],
+      "hf_revision": null,
+      "hf_filter": null,
+      "hf_avail_splits": [
+        "train"
+      ],
+      "trust_dataset": false,
+      "evaluation_splits": [
+        "train"
+      ],
+      "few_shots_split": "train",
+      "few_shots_select": "random_sampling",
+      "generation_size": 1280,
+      "generation_grammar": null,
+      "stop_sequence": [],
+      "num_samples": null,
+      "suite": [
+        "extended"
+      ],
+      "original_num_docs": 541,
+      "effective_num_docs": 541,
+      "must_remove_duplicate_docs": false,
+      "version": "0.1"
+    }
+  },
+  "summary_tasks": {
+    "extended|ifeval|0": {
+      "hashes": {
+        "hash_examples": "e99cbf567588d7c6",
+        "hash_full_prompts": "3972e282d6494c90",
+        "hash_input_tokens": "f702e1af092e9cc3",
+        "hash_cont_tokens": "f3dd45787c281c85"
+      },
+      "truncated": 0,
+      "non_truncated": 541,
+      "padded": 0,
+      "non_padded": 541,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    }
+  },
+  "summary_general": {
+    "hashes": {
+      "hash_examples": "ea046ab2c6fc5928",
+      "hash_full_prompts": "b1fd186794cb0bb3",
+      "hash_input_tokens": "5fc219976d10c06f",
+      "hash_cont_tokens": "e7a70658b0e79a5e"
+    },
+    "truncated": 0,
+    "non_truncated": 541,
+    "padded": 0,
+    "non_padded": 541,
+    "num_truncated_few_shots": 0
+  }
+}

eval_results_ood/global_step_25/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_25_actor_huggingface/results_2025-08-29T11-16-17.034884.json

@@ -0,0 +1,141 @@
+{
+  "config_general": {
+    "lighteval_sha": "327071fe86e427d880f907a51d1462f4a3f951c1",
+    "num_fewshot_seeds": 1,
+    "max_samples": null,
+    "job_id": 0,
+    "start_time": 18883988.50661321,
+    "end_time": 18884494.977114405,
+    "total_evaluation_time_secondes": "506.4705011956394",
+    "model_name": "_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_25_actor_huggingface",
+    "model_sha": "",
+    "model_dtype": null,
+    "model_size": null,
+    "generation_parameters": {
+      "early_stopping": null,
+      "repetition_penalty": null,
+      "frequency_penalty": null,
+      "length_penalty": null,
+      "presence_penalty": null,
+      "max_new_tokens": 32768,
+      "min_new_tokens": null,
+      "seed": null,
+      "stop_tokens": null,
+      "temperature": 1.0,
+      "top_k": null,
+      "min_p": null,
+      "top_p": 0.95,
+      "truncate_prompt": null,
+      "response_format": null
+    }
+  },
+  "results": {
+    "extended|ifeval|0": {
+      "prompt_level_strict_acc": 0.35489833641404805,
+      "prompt_level_strict_acc_stderr": 0.020590608575926494,
+      "inst_level_strict_acc": 0.49640287769784175,
+      "inst_level_strict_acc_stderr": 0.0005353366435758254,
+      "prompt_level_loose_acc": 0.4195933456561922,
+      "prompt_level_loose_acc_stderr": 0.021236532548855144,
+      "inst_level_loose_acc": 0.5467625899280576,
+      "inst_level_loose_acc_stderr": 0.000542489217697426
+    },
+    "all": {
+      "prompt_level_strict_acc": 0.35489833641404805,
+      "prompt_level_strict_acc_stderr": 0.020590608575926494,
+      "inst_level_strict_acc": 0.49640287769784175,
+      "inst_level_strict_acc_stderr": 0.0005353366435758254,
+      "prompt_level_loose_acc": 0.4195933456561922,
+      "prompt_level_loose_acc_stderr": 0.021236532548855144,
+      "inst_level_loose_acc": 0.5467625899280576,
+      "inst_level_loose_acc_stderr": 0.000542489217697426
+    }
+  },
+  "versions": {
+    "extended|ifeval|0": "0.1"
+  },
+  "config_tasks": {
+    "extended|ifeval": {
+      "name": "ifeval",
+      "prompt_function": "ifeval_prompt",
+      "hf_repo": "google/IFEval",
+      "hf_subset": "default",
+      "metric": [
+        {
+          "metric_name": [
+            "prompt_level_strict_acc",
+            "inst_level_strict_acc",
+            "prompt_level_loose_acc",
+            "inst_level_loose_acc"
+          ],
+          "higher_is_better": {
+            "prompt_level_strict_acc": true,
+            "inst_level_strict_acc": true,
+            "prompt_level_loose_acc": true,
+            "inst_level_loose_acc": true
+          },
+          "category": "3",
+          "use_case": "1",
+          "sample_level_fn": "ifeval_metric",
+          "corpus_level_fn": {
+            "prompt_level_strict_acc": "mean",
+            "inst_level_strict_acc": "agg_inst_level_acc",
+            "prompt_level_loose_acc": "mean",
+            "inst_level_loose_acc": "agg_inst_level_acc"
+          }
+        }
+      ],
+      "hf_revision": null,
+      "hf_filter": null,
+      "hf_avail_splits": [
+        "train"
+      ],
+      "trust_dataset": false,
+      "evaluation_splits": [
+        "train"
+      ],
+      "few_shots_split": "train",
+      "few_shots_select": "random_sampling",
+      "generation_size": 1280,
+      "generation_grammar": null,
+      "stop_sequence": [],
+      "num_samples": null,
+      "suite": [
+        "extended"
+      ],
+      "original_num_docs": 541,
+      "effective_num_docs": 541,
+      "must_remove_duplicate_docs": false,
+      "version": "0.1"
+    }
+  },
+  "summary_tasks": {
+    "extended|ifeval|0": {
+      "hashes": {
+        "hash_examples": "e99cbf567588d7c6",
+        "hash_full_prompts": "ad10e21c9427089f",
+        "hash_input_tokens": "76e5046dde448ac8",
+        "hash_cont_tokens": "b2ce952647c1913e"
+      },
+      "truncated": 0,
+      "non_truncated": 541,
+      "padded": 0,
+      "non_padded": 541,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    }
+  },
+  "summary_general": {
+    "hashes": {
+      "hash_examples": "ea046ab2c6fc5928",
+      "hash_full_prompts": "edca08807dd5a8ec",
+      "hash_input_tokens": "6db7fa56ea148d89",
+      "hash_cont_tokens": "77b8312cfe427550"
+    },
+    "truncated": 0,
+    "non_truncated": 541,
+    "padded": 0,
+    "non_padded": 541,
+    "num_truncated_few_shots": 0
+  }
+}

eval_results_ood/global_step_25/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_25_actor_huggingface/results_2025-09-23T15-20-14.861838.json

@@ -0,0 +1,141 @@
+{
+  "config_general": {
+    "lighteval_sha": "327071fe86e427d880f907a51d1462f4a3f951c1",
+    "num_fewshot_seeds": 1,
+    "max_samples": null,
+    "job_id": 0,
+    "start_time": 2819894.464696055,
+    "end_time": 2820251.742639422,
+    "total_evaluation_time_secondes": "357.2779433671385",
+    "model_name": "_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_25_actor_huggingface",
+    "model_sha": "",
+    "model_dtype": null,
+    "model_size": null,
+    "generation_parameters": {
+      "early_stopping": null,
+      "repetition_penalty": null,
+      "frequency_penalty": null,
+      "length_penalty": null,
+      "presence_penalty": null,
+      "max_new_tokens": 32768,
+      "min_new_tokens": null,
+      "seed": null,
+      "stop_tokens": null,
+      "temperature": 1.0,
+      "top_k": null,
+      "min_p": null,
+      "top_p": 0.95,
+      "truncate_prompt": null,
+      "response_format": null
+    }
+  },
+  "results": {
+    "extended|ifeval|0": {
+      "prompt_level_strict_acc": 0.4121996303142329,
+      "prompt_level_strict_acc_stderr": 0.0211822381517332,
+      "inst_level_strict_acc": 0.5443645083932853,
+      "inst_level_strict_acc_stderr": 0.0005250797455930996,
+      "prompt_level_loose_acc": 0.46950092421441775,
+      "prompt_level_loose_acc_stderr": 0.02147650768114301,
+      "inst_level_loose_acc": 0.5911270983213429,
+      "inst_level_loose_acc_stderr": 0.0005233831979716518
+    },
+    "all": {
+      "prompt_level_strict_acc": 0.4121996303142329,
+      "prompt_level_strict_acc_stderr": 0.0211822381517332,
+      "inst_level_strict_acc": 0.5443645083932853,
+      "inst_level_strict_acc_stderr": 0.0005250797455930996,
+      "prompt_level_loose_acc": 0.46950092421441775,
+      "prompt_level_loose_acc_stderr": 0.02147650768114301,
+      "inst_level_loose_acc": 0.5911270983213429,
+      "inst_level_loose_acc_stderr": 0.0005233831979716518
+    }
+  },
+  "versions": {
+    "extended|ifeval|0": "0.1"
+  },
+  "config_tasks": {
+    "extended|ifeval": {
+      "name": "ifeval",
+      "prompt_function": "ifeval_prompt",
+      "hf_repo": "google/IFEval",
+      "hf_subset": "default",
+      "metric": [
+        {
+          "metric_name": [
+            "prompt_level_strict_acc",
+            "inst_level_strict_acc",
+            "prompt_level_loose_acc",
+            "inst_level_loose_acc"
+          ],
+          "higher_is_better": {
+            "prompt_level_strict_acc": true,
+            "inst_level_strict_acc": true,
+            "prompt_level_loose_acc": true,
+            "inst_level_loose_acc": true
+          },
+          "category": "3",
+          "use_case": "1",
+          "sample_level_fn": "ifeval_metric",
+          "corpus_level_fn": {
+            "prompt_level_strict_acc": "mean",
+            "inst_level_strict_acc": "agg_inst_level_acc",
+            "prompt_level_loose_acc": "mean",
+            "inst_level_loose_acc": "agg_inst_level_acc"
+          }
+        }
+      ],
+      "hf_revision": null,
+      "hf_filter": null,
+      "hf_avail_splits": [
+        "train"
+      ],
+      "trust_dataset": false,
+      "evaluation_splits": [
+        "train"
+      ],
+      "few_shots_split": "train",
+      "few_shots_select": "random_sampling",
+      "generation_size": 1280,
+      "generation_grammar": null,
+      "stop_sequence": [],
+      "num_samples": null,
+      "suite": [
+        "extended"
+      ],
+      "original_num_docs": 541,
+      "effective_num_docs": 541,
+      "must_remove_duplicate_docs": false,
+      "version": "0.1"
+    }
+  },
+  "summary_tasks": {
+    "extended|ifeval|0": {
+      "hashes": {
+        "hash_examples": "e99cbf567588d7c6",
+        "hash_full_prompts": "3972e282d6494c90",
+        "hash_input_tokens": "f702e1af092e9cc3",
+        "hash_cont_tokens": "13df23e3d89d50d6"
+      },
+      "truncated": 0,
+      "non_truncated": 541,
+      "padded": 0,
+      "non_padded": 541,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    }
+  },
+  "summary_general": {
+    "hashes": {
+      "hash_examples": "ea046ab2c6fc5928",
+      "hash_full_prompts": "b1fd186794cb0bb3",
+      "hash_input_tokens": "5fc219976d10c06f",
+      "hash_cont_tokens": "3aa74441e103e1f2"
+    },
+    "truncated": 0,
+    "non_truncated": 541,
+    "padded": 0,
+    "non_padded": 541,
+    "num_truncated_few_shots": 0
+  }
+}

eval_results_ood/global_step_30/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_30_actor_huggingface/results_2025-08-29T11-17-00.747010.json

@@ -0,0 +1,141 @@
+{
+  "config_general": {
+    "lighteval_sha": "327071fe86e427d880f907a51d1462f4a3f951c1",
+    "num_fewshot_seeds": 1,
+    "max_samples": null,
+    "job_id": 0,
+    "start_time": 18883988.700935464,
+    "end_time": 18884540.474919353,
+    "total_evaluation_time_secondes": "551.7739838883281",
+    "model_name": "_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_30_actor_huggingface",
+    "model_sha": "",
+    "model_dtype": null,
+    "model_size": null,
+    "generation_parameters": {
+      "early_stopping": null,
+      "repetition_penalty": null,
+      "frequency_penalty": null,
+      "length_penalty": null,
+      "presence_penalty": null,
+      "max_new_tokens": 32768,
+      "min_new_tokens": null,
+      "seed": null,
+      "stop_tokens": null,
+      "temperature": 1.0,
+      "top_k": null,
+      "min_p": null,
+      "top_p": 0.95,
+      "truncate_prompt": null,
+      "response_format": null
+    }
+  },
+  "results": {
+    "extended|ifeval|0": {
+      "prompt_level_strict_acc": 0.3844731977818854,
+      "prompt_level_strict_acc_stderr": 0.02093435763458466,
+      "inst_level_strict_acc": 0.5035971223021583,
+      "inst_level_strict_acc_stderr": 0.0005667334369309774,
+      "prompt_level_loose_acc": 0.4584103512014787,
+      "prompt_level_loose_acc_stderr": 0.021442010560476534,
+      "inst_level_loose_acc": 0.5671462829736211,
+      "inst_level_loose_acc_stderr": 0.0005740916373334755
+    },
+    "all": {
+      "prompt_level_strict_acc": 0.3844731977818854,
+      "prompt_level_strict_acc_stderr": 0.02093435763458466,
+      "inst_level_strict_acc": 0.5035971223021583,
+      "inst_level_strict_acc_stderr": 0.0005667334369309774,
+      "prompt_level_loose_acc": 0.4584103512014787,
+      "prompt_level_loose_acc_stderr": 0.021442010560476534,
+      "inst_level_loose_acc": 0.5671462829736211,
+      "inst_level_loose_acc_stderr": 0.0005740916373334755
+    }
+  },
+  "versions": {
+    "extended|ifeval|0": "0.1"
+  },
+  "config_tasks": {
+    "extended|ifeval": {
+      "name": "ifeval",
+      "prompt_function": "ifeval_prompt",
+      "hf_repo": "google/IFEval",
+      "hf_subset": "default",
+      "metric": [
+        {
+          "metric_name": [
+            "prompt_level_strict_acc",
+            "inst_level_strict_acc",
+            "prompt_level_loose_acc",
+            "inst_level_loose_acc"
+          ],
+          "higher_is_better": {
+            "prompt_level_strict_acc": true,
+            "inst_level_strict_acc": true,
+            "prompt_level_loose_acc": true,
+            "inst_level_loose_acc": true
+          },
+          "category": "3",
+          "use_case": "1",
+          "sample_level_fn": "ifeval_metric",
+          "corpus_level_fn": {
+            "prompt_level_strict_acc": "mean",
+            "inst_level_strict_acc": "agg_inst_level_acc",
+            "prompt_level_loose_acc": "mean",
+            "inst_level_loose_acc": "agg_inst_level_acc"
+          }
+        }
+      ],
+      "hf_revision": null,
+      "hf_filter": null,
+      "hf_avail_splits": [
+        "train"
+      ],
+      "trust_dataset": false,
+      "evaluation_splits": [
+        "train"
+      ],
+      "few_shots_split": "train",
+      "few_shots_select": "random_sampling",
+      "generation_size": 1280,
+      "generation_grammar": null,
+      "stop_sequence": [],
+      "num_samples": null,
+      "suite": [
+        "extended"
+      ],
+      "original_num_docs": 541,
+      "effective_num_docs": 541,
+      "must_remove_duplicate_docs": false,
+      "version": "0.1"
+    }
+  },
+  "summary_tasks": {
+    "extended|ifeval|0": {
+      "hashes": {
+        "hash_examples": "e99cbf567588d7c6",
+        "hash_full_prompts": "ad10e21c9427089f",
+        "hash_input_tokens": "76e5046dde448ac8",
+        "hash_cont_tokens": "55fce4d54bcc603e"
+      },
+      "truncated": 0,
+      "non_truncated": 541,
+      "padded": 0,
+      "non_padded": 541,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    }
+  },
+  "summary_general": {
+    "hashes": {
+      "hash_examples": "ea046ab2c6fc5928",
+      "hash_full_prompts": "edca08807dd5a8ec",
+      "hash_input_tokens": "6db7fa56ea148d89",
+      "hash_cont_tokens": "f20ba557778f234b"
+    },
+    "truncated": 0,
+    "non_truncated": 541,
+    "padded": 0,
+    "non_padded": 541,
+    "num_truncated_few_shots": 0
+  }
+}

eval_results_ood/global_step_30/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_30_actor_huggingface/results_2025-09-23T15-18-49.877513.json

@@ -0,0 +1,141 @@
+{
+  "config_general": {
+    "lighteval_sha": "327071fe86e427d880f907a51d1462f4a3f951c1",
+    "num_fewshot_seeds": 1,
+    "max_samples": null,
+    "job_id": 0,
+    "start_time": 2819894.526100757,
+    "end_time": 2820167.222449713,
+    "total_evaluation_time_secondes": "272.6963489558548",
+    "model_name": "_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_30_actor_huggingface",
+    "model_sha": "",
+    "model_dtype": null,
+    "model_size": null,
+    "generation_parameters": {
+      "early_stopping": null,
+      "repetition_penalty": null,
+      "frequency_penalty": null,
+      "length_penalty": null,
+      "presence_penalty": null,
+      "max_new_tokens": 32768,
+      "min_new_tokens": null,
+      "seed": null,
+      "stop_tokens": null,
+      "temperature": 1.0,
+      "top_k": null,
+      "min_p": null,
+      "top_p": 0.95,
+      "truncate_prompt": null,
+      "response_format": null
+    }
+  },
+  "results": {
+    "extended|ifeval|0": {
+      "prompt_level_strict_acc": 0.41589648798521256,
+      "prompt_level_strict_acc_stderr": 0.0212099993568188,
+      "inst_level_strict_acc": 0.5371702637889688,
+      "inst_level_strict_acc_stderr": 0.0005646791772078428,
+      "prompt_level_loose_acc": 0.4584103512014787,
+      "prompt_level_loose_acc_stderr": 0.02144201056047653,
+      "inst_level_loose_acc": 0.5815347721822542,
+      "inst_level_loose_acc_stderr": 0.0005489847919512223
+    },
+    "all": {
+      "prompt_level_strict_acc": 0.41589648798521256,
+      "prompt_level_strict_acc_stderr": 0.0212099993568188,
+      "inst_level_strict_acc": 0.5371702637889688,
+      "inst_level_strict_acc_stderr": 0.0005646791772078428,
+      "prompt_level_loose_acc": 0.4584103512014787,
+      "prompt_level_loose_acc_stderr": 0.02144201056047653,
+      "inst_level_loose_acc": 0.5815347721822542,
+      "inst_level_loose_acc_stderr": 0.0005489847919512223
+    }
+  },
+  "versions": {
+    "extended|ifeval|0": "0.1"
+  },
+  "config_tasks": {
+    "extended|ifeval": {
+      "name": "ifeval",
+      "prompt_function": "ifeval_prompt",
+      "hf_repo": "google/IFEval",
+      "hf_subset": "default",
+      "metric": [
+        {
+          "metric_name": [
+            "prompt_level_strict_acc",
+            "inst_level_strict_acc",
+            "prompt_level_loose_acc",
+            "inst_level_loose_acc"
+          ],
+          "higher_is_better": {
+            "prompt_level_strict_acc": true,
+            "inst_level_strict_acc": true,
+            "prompt_level_loose_acc": true,
+            "inst_level_loose_acc": true
+          },
+          "category": "3",
+          "use_case": "1",
+          "sample_level_fn": "ifeval_metric",
+          "corpus_level_fn": {
+            "prompt_level_strict_acc": "mean",
+            "inst_level_strict_acc": "agg_inst_level_acc",
+            "prompt_level_loose_acc": "mean",
+            "inst_level_loose_acc": "agg_inst_level_acc"
+          }
+        }
+      ],
+      "hf_revision": null,
+      "hf_filter": null,
+      "hf_avail_splits": [
+        "train"
+      ],
+      "trust_dataset": false,
+      "evaluation_splits": [
+        "train"
+      ],
+      "few_shots_split": "train",
+      "few_shots_select": "random_sampling",
+      "generation_size": 1280,
+      "generation_grammar": null,
+      "stop_sequence": [],
+      "num_samples": null,
+      "suite": [
+        "extended"
+      ],
+      "original_num_docs": 541,
+      "effective_num_docs": 541,
+      "must_remove_duplicate_docs": false,
+      "version": "0.1"
+    }
+  },
+  "summary_tasks": {
+    "extended|ifeval|0": {
+      "hashes": {
+        "hash_examples": "e99cbf567588d7c6",
+        "hash_full_prompts": "3972e282d6494c90",
+        "hash_input_tokens": "f702e1af092e9cc3",
+        "hash_cont_tokens": "e2458b3184dfdb7c"
+      },
+      "truncated": 0,
+      "non_truncated": 541,
+      "padded": 0,
+      "non_padded": 541,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    }
+  },
+  "summary_general": {
+    "hashes": {
+      "hash_examples": "ea046ab2c6fc5928",
+      "hash_full_prompts": "b1fd186794cb0bb3",
+      "hash_input_tokens": "5fc219976d10c06f",
+      "hash_cont_tokens": "22e60018f1f696a5"
+    },
+    "truncated": 0,
+    "non_truncated": 541,
+    "padded": 0,
+    "non_padded": 541,
+    "num_truncated_few_shots": 0
+  }
+}

eval_results_ood/global_step_35/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_35_actor_huggingface/results_2025-08-29T11-16-37.152415.json

@@ -0,0 +1,141 @@
+{
+  "config_general": {
+    "lighteval_sha": "327071fe86e427d880f907a51d1462f4a3f951c1",
+    "num_fewshot_seeds": 1,
+    "max_samples": null,
+    "job_id": 0,
+    "start_time": 18883988.616446614,
+    "end_time": 18884516.7507356,
+    "total_evaluation_time_secondes": "528.1342889852822",
+    "model_name": "_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_35_actor_huggingface",
+    "model_sha": "",
+    "model_dtype": null,
+    "model_size": null,
+    "generation_parameters": {
+      "early_stopping": null,
+      "repetition_penalty": null,
+      "frequency_penalty": null,
+      "length_penalty": null,
+      "presence_penalty": null,
+      "max_new_tokens": 32768,
+      "min_new_tokens": null,
+      "seed": null,
+      "stop_tokens": null,
+      "temperature": 1.0,
+      "top_k": null,
+      "min_p": null,
+      "top_p": 0.95,
+      "truncate_prompt": null,
+      "response_format": null
+    }
+  },
+  "results": {
+    "extended|ifeval|0": {
+      "prompt_level_strict_acc": 0.3752310536044362,
+      "prompt_level_strict_acc_stderr": 0.020835898065035694,
+      "inst_level_strict_acc": 0.5023980815347722,
+      "inst_level_strict_acc_stderr": 0.0005393909491033796,
+      "prompt_level_loose_acc": 0.4288354898336414,
+      "prompt_level_loose_acc_stderr": 0.021297522569050743,
+      "inst_level_loose_acc": 0.552757793764988,
+      "inst_level_loose_acc_stderr": 0.0005344148217521974
+    },
+    "all": {
+      "prompt_level_strict_acc": 0.3752310536044362,
+      "prompt_level_strict_acc_stderr": 0.020835898065035694,
+      "inst_level_strict_acc": 0.5023980815347722,
+      "inst_level_strict_acc_stderr": 0.0005393909491033796,
+      "prompt_level_loose_acc": 0.4288354898336414,
+      "prompt_level_loose_acc_stderr": 0.021297522569050743,
+      "inst_level_loose_acc": 0.552757793764988,
+      "inst_level_loose_acc_stderr": 0.0005344148217521974
+    }
+  },
+  "versions": {
+    "extended|ifeval|0": "0.1"
+  },
+  "config_tasks": {
+    "extended|ifeval": {
+      "name": "ifeval",
+      "prompt_function": "ifeval_prompt",
+      "hf_repo": "google/IFEval",
+      "hf_subset": "default",
+      "metric": [
+        {
+          "metric_name": [
+            "prompt_level_strict_acc",
+            "inst_level_strict_acc",
+            "prompt_level_loose_acc",
+            "inst_level_loose_acc"
+          ],
+          "higher_is_better": {
+            "prompt_level_strict_acc": true,
+            "inst_level_strict_acc": true,
+            "prompt_level_loose_acc": true,
+            "inst_level_loose_acc": true
+          },
+          "category": "3",
+          "use_case": "1",
+          "sample_level_fn": "ifeval_metric",
+          "corpus_level_fn": {
+            "prompt_level_strict_acc": "mean",
+            "inst_level_strict_acc": "agg_inst_level_acc",
+            "prompt_level_loose_acc": "mean",
+            "inst_level_loose_acc": "agg_inst_level_acc"
+          }
+        }
+      ],
+      "hf_revision": null,
+      "hf_filter": null,
+      "hf_avail_splits": [
+        "train"
+      ],
+      "trust_dataset": false,
+      "evaluation_splits": [
+        "train"
+      ],
+      "few_shots_split": "train",
+      "few_shots_select": "random_sampling",
+      "generation_size": 1280,
+      "generation_grammar": null,
+      "stop_sequence": [],
+      "num_samples": null,
+      "suite": [
+        "extended"
+      ],
+      "original_num_docs": 541,
+      "effective_num_docs": 541,
+      "must_remove_duplicate_docs": false,
+      "version": "0.1"
+    }
+  },
+  "summary_tasks": {
+    "extended|ifeval|0": {
+      "hashes": {
+        "hash_examples": "e99cbf567588d7c6",
+        "hash_full_prompts": "ad10e21c9427089f",
+        "hash_input_tokens": "76e5046dde448ac8",
+        "hash_cont_tokens": "0f42ca4607b5eab0"
+      },
+      "truncated": 0,
+      "non_truncated": 541,
+      "padded": 0,
+      "non_padded": 541,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    }
+  },
+  "summary_general": {
+    "hashes": {
+      "hash_examples": "ea046ab2c6fc5928",
+      "hash_full_prompts": "edca08807dd5a8ec",
+      "hash_input_tokens": "6db7fa56ea148d89",
+      "hash_cont_tokens": "f40d9eb07b829046"
+    },
+    "truncated": 0,
+    "non_truncated": 541,
+    "padded": 0,
+    "non_padded": 541,
+    "num_truncated_few_shots": 0
+  }
+}

eval_results_ood/global_step_35/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_35_actor_huggingface/results_2025-09-23T15-23-52.120403.json

@@ -0,0 +1,141 @@
+{
+  "config_general": {
+    "lighteval_sha": "327071fe86e427d880f907a51d1462f4a3f951c1",
+    "num_fewshot_seeds": 1,
+    "max_samples": null,
+    "job_id": 0,
+    "start_time": 2820291.026089232,
+    "end_time": 2820469.627462913,
+    "total_evaluation_time_secondes": "178.60137368086725",
+    "model_name": "_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_35_actor_huggingface",
+    "model_sha": "",
+    "model_dtype": null,
+    "model_size": null,
+    "generation_parameters": {
+      "early_stopping": null,
+      "repetition_penalty": null,
+      "frequency_penalty": null,
+      "length_penalty": null,
+      "presence_penalty": null,
+      "max_new_tokens": 32768,
+      "min_new_tokens": null,
+      "seed": null,
+      "stop_tokens": null,
+      "temperature": 1.0,
+      "top_k": null,
+      "min_p": null,
+      "top_p": 0.95,
+      "truncate_prompt": null,
+      "response_format": null
+    }
+  },
+  "results": {
+    "extended|ifeval|0": {
+      "prompt_level_strict_acc": 0.4048059149722736,
+      "prompt_level_strict_acc_stderr": 0.021123012121053563,
+      "inst_level_strict_acc": 0.5287769784172662,
+      "inst_level_strict_acc_stderr": 0.0005365061017861071,
+      "prompt_level_loose_acc": 0.4565619223659889,
+      "prompt_level_loose_acc_stderr": 0.021435222545538896,
+      "inst_level_loose_acc": 0.5827338129496403,
+      "inst_level_loose_acc_stderr": 0.0005348025912540751
+    },
+    "all": {
+      "prompt_level_strict_acc": 0.4048059149722736,
+      "prompt_level_strict_acc_stderr": 0.021123012121053563,
+      "inst_level_strict_acc": 0.5287769784172662,
+      "inst_level_strict_acc_stderr": 0.0005365061017861071,
+      "prompt_level_loose_acc": 0.4565619223659889,
+      "prompt_level_loose_acc_stderr": 0.021435222545538896,
+      "inst_level_loose_acc": 0.5827338129496403,
+      "inst_level_loose_acc_stderr": 0.0005348025912540751
+    }
+  },
+  "versions": {
+    "extended|ifeval|0": "0.1"
+  },
+  "config_tasks": {
+    "extended|ifeval": {
+      "name": "ifeval",
+      "prompt_function": "ifeval_prompt",
+      "hf_repo": "google/IFEval",
+      "hf_subset": "default",
+      "metric": [
+        {
+          "metric_name": [
+            "prompt_level_strict_acc",
+            "inst_level_strict_acc",
+            "prompt_level_loose_acc",
+            "inst_level_loose_acc"
+          ],
+          "higher_is_better": {
+            "prompt_level_strict_acc": true,
+            "inst_level_strict_acc": true,
+            "prompt_level_loose_acc": true,
+            "inst_level_loose_acc": true
+          },
+          "category": "3",
+          "use_case": "1",
+          "sample_level_fn": "ifeval_metric",
+          "corpus_level_fn": {
+            "prompt_level_strict_acc": "mean",
+            "inst_level_strict_acc": "agg_inst_level_acc",
+            "prompt_level_loose_acc": "mean",
+            "inst_level_loose_acc": "agg_inst_level_acc"
+          }
+        }
+      ],
+      "hf_revision": null,
+      "hf_filter": null,
+      "hf_avail_splits": [
+        "train"
+      ],
+      "trust_dataset": false,
+      "evaluation_splits": [
+        "train"
+      ],
+      "few_shots_split": "train",
+      "few_shots_select": "random_sampling",
+      "generation_size": 1280,
+      "generation_grammar": null,
+      "stop_sequence": [],
+      "num_samples": null,
+      "suite": [
+        "extended"
+      ],
+      "original_num_docs": 541,
+      "effective_num_docs": 541,
+      "must_remove_duplicate_docs": false,
+      "version": "0.1"
+    }
+  },
+  "summary_tasks": {
+    "extended|ifeval|0": {
+      "hashes": {
+        "hash_examples": "e99cbf567588d7c6",
+        "hash_full_prompts": "3972e282d6494c90",
+        "hash_input_tokens": "f702e1af092e9cc3",
+        "hash_cont_tokens": "1ecf208bbf533d01"
+      },
+      "truncated": 0,
+      "non_truncated": 541,
+      "padded": 0,
+      "non_padded": 541,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    }
+  },
+  "summary_general": {
+    "hashes": {
+      "hash_examples": "ea046ab2c6fc5928",
+      "hash_full_prompts": "b1fd186794cb0bb3",
+      "hash_input_tokens": "5fc219976d10c06f",
+      "hash_cont_tokens": "643cc6202abfe217"
+    },
+    "truncated": 0,
+    "non_truncated": 541,
+    "padded": 0,
+    "non_padded": 541,
+    "num_truncated_few_shots": 0
+  }
+}

eval_results_ood/global_step_5/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_5_actor_huggingface/results_2025-08-29T11-16-14.580600.json

@@ -0,0 +1,141 @@
+{
+  "config_general": {
+    "lighteval_sha": "327071fe86e427d880f907a51d1462f4a3f951c1",
+    "num_fewshot_seeds": 1,
+    "max_samples": null,
+    "job_id": 0,
+    "start_time": 18883988.61094902,
+    "end_time": 18884493.887714904,
+    "total_evaluation_time_secondes": "505.2767658829689",
+    "model_name": "_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_5_actor_huggingface",
+    "model_sha": "",
+    "model_dtype": null,
+    "model_size": null,
+    "generation_parameters": {
+      "early_stopping": null,
+      "repetition_penalty": null,
+      "frequency_penalty": null,
+      "length_penalty": null,
+      "presence_penalty": null,
+      "max_new_tokens": 32768,
+      "min_new_tokens": null,
+      "seed": null,
+      "stop_tokens": null,
+      "temperature": 1.0,
+      "top_k": null,
+      "min_p": null,
+      "top_p": 0.95,
+      "truncate_prompt": null,
+      "response_format": null
+    }
+  },
+  "results": {
+    "extended|ifeval|0": {
+      "prompt_level_strict_acc": 0.3826247689463956,
+      "prompt_level_strict_acc_stderr": 0.020915307841436644,
+      "inst_level_strict_acc": 0.5095923261390888,
+      "inst_level_strict_acc_stderr": 0.0005601452796683326,
+      "prompt_level_loose_acc": 0.4343807763401109,
+      "prompt_level_loose_acc_stderr": 0.021330473657564724,
+      "inst_level_loose_acc": 0.5707434052757794,
+      "inst_level_loose_acc_stderr": 0.0005481574868477439
+    },
+    "all": {
+      "prompt_level_strict_acc": 0.3826247689463956,
+      "prompt_level_strict_acc_stderr": 0.020915307841436644,
+      "inst_level_strict_acc": 0.5095923261390888,
+      "inst_level_strict_acc_stderr": 0.0005601452796683326,
+      "prompt_level_loose_acc": 0.4343807763401109,
+      "prompt_level_loose_acc_stderr": 0.021330473657564724,
+      "inst_level_loose_acc": 0.5707434052757794,
+      "inst_level_loose_acc_stderr": 0.0005481574868477439
+    }
+  },
+  "versions": {
+    "extended|ifeval|0": "0.1"
+  },
+  "config_tasks": {
+    "extended|ifeval": {
+      "name": "ifeval",
+      "prompt_function": "ifeval_prompt",
+      "hf_repo": "google/IFEval",
+      "hf_subset": "default",
+      "metric": [
+        {
+          "metric_name": [
+            "prompt_level_strict_acc",
+            "inst_level_strict_acc",
+            "prompt_level_loose_acc",
+            "inst_level_loose_acc"
+          ],
+          "higher_is_better": {
+            "prompt_level_strict_acc": true,
+            "inst_level_strict_acc": true,
+            "prompt_level_loose_acc": true,
+            "inst_level_loose_acc": true
+          },
+          "category": "3",
+          "use_case": "1",
+          "sample_level_fn": "ifeval_metric",
+          "corpus_level_fn": {
+            "prompt_level_strict_acc": "mean",
+            "inst_level_strict_acc": "agg_inst_level_acc",
+            "prompt_level_loose_acc": "mean",
+            "inst_level_loose_acc": "agg_inst_level_acc"
+          }
+        }
+      ],
+      "hf_revision": null,
+      "hf_filter": null,
+      "hf_avail_splits": [
+        "train"
+      ],
+      "trust_dataset": false,
+      "evaluation_splits": [
+        "train"
+      ],
+      "few_shots_split": "train",
+      "few_shots_select": "random_sampling",
+      "generation_size": 1280,
+      "generation_grammar": null,
+      "stop_sequence": [],
+      "num_samples": null,
+      "suite": [
+        "extended"
+      ],
+      "original_num_docs": 541,
+      "effective_num_docs": 541,
+      "must_remove_duplicate_docs": false,
+      "version": "0.1"
+    }
+  },
+  "summary_tasks": {
+    "extended|ifeval|0": {
+      "hashes": {
+        "hash_examples": "e99cbf567588d7c6",
+        "hash_full_prompts": "ad10e21c9427089f",
+        "hash_input_tokens": "76e5046dde448ac8",
+        "hash_cont_tokens": "7c8b4e41f6fd051b"
+      },
+      "truncated": 0,
+      "non_truncated": 541,
+      "padded": 0,
+      "non_padded": 541,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    }
+  },
+  "summary_general": {
+    "hashes": {
+      "hash_examples": "ea046ab2c6fc5928",
+      "hash_full_prompts": "edca08807dd5a8ec",
+      "hash_input_tokens": "6db7fa56ea148d89",
+      "hash_cont_tokens": "b1545cf2f315b64f"
+    },
+    "truncated": 0,
+    "non_truncated": 541,
+    "padded": 0,
+    "non_padded": 541,
+    "num_truncated_few_shots": 0
+  }
+}

eval_results_ood/global_step_5/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_5_actor_huggingface/results_2025-09-23T15-18-54.425207.json

@@ -0,0 +1,141 @@
+{
+  "config_general": {
+    "lighteval_sha": "327071fe86e427d880f907a51d1462f4a3f951c1",
+    "num_fewshot_seeds": 1,
+    "max_samples": null,
+    "job_id": 0,
+    "start_time": 2819894.607254071,
+    "end_time": 2820171.997573846,
+    "total_evaluation_time_secondes": "277.39031977485865",
+    "model_name": "_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_5_actor_huggingface",
+    "model_sha": "",
+    "model_dtype": null,
+    "model_size": null,
+    "generation_parameters": {
+      "early_stopping": null,
+      "repetition_penalty": null,
+      "frequency_penalty": null,
+      "length_penalty": null,
+      "presence_penalty": null,
+      "max_new_tokens": 32768,
+      "min_new_tokens": null,
+      "seed": null,
+      "stop_tokens": null,
+      "temperature": 1.0,
+      "top_k": null,
+      "min_p": null,
+      "top_p": 0.95,
+      "truncate_prompt": null,
+      "response_format": null
+    }
+  },
+  "results": {
+    "extended|ifeval|0": {
+      "prompt_level_strict_acc": 0.39186691312384475,
+      "prompt_level_strict_acc_stderr": 0.02100737240618835,
+      "inst_level_strict_acc": 0.5251798561151079,
+      "inst_level_strict_acc_stderr": 0.0005730653846434158,
+      "prompt_level_loose_acc": 0.43807763401109057,
+      "prompt_level_loose_acc_stderr": 0.021350931135490935,
+      "inst_level_loose_acc": 0.5695443645083933,
+      "inst_level_loose_acc_stderr": 0.0005581180833934133
+    },
+    "all": {
+      "prompt_level_strict_acc": 0.39186691312384475,
+      "prompt_level_strict_acc_stderr": 0.02100737240618835,
+      "inst_level_strict_acc": 0.5251798561151079,
+      "inst_level_strict_acc_stderr": 0.0005730653846434158,
+      "prompt_level_loose_acc": 0.43807763401109057,
+      "prompt_level_loose_acc_stderr": 0.021350931135490935,
+      "inst_level_loose_acc": 0.5695443645083933,
+      "inst_level_loose_acc_stderr": 0.0005581180833934133
+    }
+  },
+  "versions": {
+    "extended|ifeval|0": "0.1"
+  },
+  "config_tasks": {
+    "extended|ifeval": {
+      "name": "ifeval",
+      "prompt_function": "ifeval_prompt",
+      "hf_repo": "google/IFEval",
+      "hf_subset": "default",
+      "metric": [
+        {
+          "metric_name": [
+            "prompt_level_strict_acc",
+            "inst_level_strict_acc",
+            "prompt_level_loose_acc",
+            "inst_level_loose_acc"
+          ],
+          "higher_is_better": {
+            "prompt_level_strict_acc": true,
+            "inst_level_strict_acc": true,
+            "prompt_level_loose_acc": true,
+            "inst_level_loose_acc": true
+          },
+          "category": "3",
+          "use_case": "1",
+          "sample_level_fn": "ifeval_metric",
+          "corpus_level_fn": {
+            "prompt_level_strict_acc": "mean",
+            "inst_level_strict_acc": "agg_inst_level_acc",
+            "prompt_level_loose_acc": "mean",
+            "inst_level_loose_acc": "agg_inst_level_acc"
+          }
+        }
+      ],
+      "hf_revision": null,
+      "hf_filter": null,
+      "hf_avail_splits": [
+        "train"
+      ],
+      "trust_dataset": false,
+      "evaluation_splits": [
+        "train"
+      ],
+      "few_shots_split": "train",
+      "few_shots_select": "random_sampling",
+      "generation_size": 1280,
+      "generation_grammar": null,
+      "stop_sequence": [],
+      "num_samples": null,
+      "suite": [
+        "extended"
+      ],
+      "original_num_docs": 541,
+      "effective_num_docs": 541,
+      "must_remove_duplicate_docs": false,
+      "version": "0.1"
+    }
+  },
+  "summary_tasks": {
+    "extended|ifeval|0": {
+      "hashes": {
+        "hash_examples": "e99cbf567588d7c6",
+        "hash_full_prompts": "3972e282d6494c90",
+        "hash_input_tokens": "f702e1af092e9cc3",
+        "hash_cont_tokens": "e84762b7f983ecbc"
+      },
+      "truncated": 0,
+      "non_truncated": 541,
+      "padded": 0,
+      "non_padded": 541,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    }
+  },
+  "summary_general": {
+    "hashes": {
+      "hash_examples": "ea046ab2c6fc5928",
+      "hash_full_prompts": "b1fd186794cb0bb3",
+      "hash_input_tokens": "5fc219976d10c06f",
+      "hash_cont_tokens": "4dcc542128551811"
+    },
+    "truncated": 0,
+    "non_truncated": 541,
+    "padded": 0,
+    "non_padded": 541,
+    "num_truncated_few_shots": 0
+  }
+}

eval_results_ood/results.csv

@@ -0,0 +1,16 @@
+exp_name,global_step,model_name,json_relpath,prompt_level_strict_acc,prompt_level_strict_acc_stderr,inst_level_strict_acc,inst_level_strict_acc_stderr,prompt_level_loose_acc,prompt_level_loose_acc_stderr,inst_level_loose_acc,inst_level_loose_acc_stderr,gpqa_pass@1:1_samples,gpqa_pass@1:1_samples_stderr
+verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15,0,_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_0_actor_huggingface,verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15/eval_results_ood/global_step_0/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_0_actor_huggingface/results_2025-08-29T11-17-46.747738.json,0.39371534195933455,0.02102483414587237,0.5215827338129496,0.0005599264307527568,0.4288354898336414,0.021297522569050743,0.5563549160671463,0.0005527714546864814,,
+verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15,0,_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_0_actor_huggingface,verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15/eval_results_ood/global_step_0/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_0_actor_huggingface/results_2025-09-23T15-20-09.172749.json,0.39371534195933455,0.02102483414587237,0.5191846522781774,0.0005633834275419194,0.4399260628465804,0.021360708220801976,0.564748201438849,0.0005587046293737301,,
+verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15,5,_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_5_actor_huggingface,verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15/eval_results_ood/global_step_5/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_5_actor_huggingface/results_2025-08-29T11-16-14.580600.json,0.3826247689463956,0.020915307841436644,0.5095923261390888,0.0005601452796683326,0.4343807763401109,0.021330473657564724,0.5707434052757794,0.0005481574868477439,,
+verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15,5,_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_5_actor_huggingface,verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15/eval_results_ood/global_step_5/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_5_actor_huggingface/results_2025-09-23T15-18-54.425207.json,0.39186691312384475,0.02100737240618835,0.5251798561151079,0.0005730653846434158,0.43807763401109057,0.021350931135490935,0.5695443645083933,0.0005581180833934133,,
+verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15,10,_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_10_actor_huggingface,verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15/eval_results_ood/global_step_10/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_10_actor_huggingface/results_2025-09-23T15-19-07.845971.json,0.39186691312384475,0.021007372406188354,0.5143884892086331,0.0005407156424898863,0.43253234750462105,0.02131979239817807,0.552757793764988,0.0005395531273954375,,
+verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15,15,_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_15_actor_huggingface,verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15/eval_results_ood/global_step_15/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_15_actor_huggingface/results_2025-08-29T11-16-24.695422.json,0.3752310536044362,0.020835898065035694,0.5095923261390888,0.0005432302541010317,0.4269870609981516,0.02128593305006131,0.5587529976019184,0.000545503372755182,,
+verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15,15,_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_15_actor_huggingface,verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15/eval_results_ood/global_step_15/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_15_actor_huggingface/results_2025-09-23T15-18-58.766285.json,0.37707948243992606,0.020856233918528456,0.5155875299760192,0.0005624364712452048,0.4343807763401109,0.021330473657564727,0.5671462829736211,0.0005536778232943271,,
+verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15,20,_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_20_actor_huggingface,verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15/eval_results_ood/global_step_20/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_20_actor_huggingface/results_2025-08-29T11-16-25.542329.json,0.3807763401109057,0.020895937888190833,0.511990407673861,0.0005284703064941477,0.43807763401109057,0.021350931135490935,0.5599520383693045,0.0005508806104783129,,
+verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15,20,_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_20_actor_huggingface,verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15/eval_results_ood/global_step_20/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_20_actor_huggingface/results_2025-09-23T15-20-14.009200.json,0.4011090573012939,0.02109153689552074,0.5143884892086331,0.0005584775478161843,0.4510166358595194,0.02141307276535979,0.5635491606714629,0.0005514767622336946,,
+verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15,25,_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_25_actor_huggingface,verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15/eval_results_ood/global_step_25/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_25_actor_huggingface/results_2025-08-29T11-16-17.034884.json,0.35489833641404805,0.020590608575926494,0.49640287769784175,0.0005353366435758254,0.4195933456561922,0.021236532548855144,0.5467625899280576,0.000542489217697426,,
+verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15,25,_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_25_actor_huggingface,verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15/eval_results_ood/global_step_25/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_25_actor_huggingface/results_2025-09-23T15-20-14.861838.json,0.4121996303142329,0.0211822381517332,0.5443645083932853,0.0005250797455930996,0.46950092421441775,0.02147650768114301,0.5911270983213429,0.0005233831979716518,,
+verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15,30,_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_30_actor_huggingface,verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15/eval_results_ood/global_step_30/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_30_actor_huggingface/results_2025-08-29T11-17-00.747010.json,0.3844731977818854,0.02093435763458466,0.5035971223021583,0.0005667334369309774,0.4584103512014787,0.021442010560476534,0.5671462829736211,0.0005740916373334755,,
+verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15,30,_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_30_actor_huggingface,verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15/eval_results_ood/global_step_30/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_30_actor_huggingface/results_2025-09-23T15-18-49.877513.json,0.41589648798521256,0.0212099993568188,0.5371702637889688,0.0005646791772078428,0.4584103512014787,0.02144201056047653,0.5815347721822542,0.0005489847919512223,,
+verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15,35,_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_35_actor_huggingface,verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15/eval_results_ood/global_step_35/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_35_actor_huggingface/results_2025-08-29T11-16-37.152415.json,0.3752310536044362,0.020835898065035694,0.5023980815347722,0.0005393909491033796,0.4288354898336414,0.021297522569050743,0.552757793764988,0.0005344148217521974,,
+verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15,35,_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_35_actor_huggingface,verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15/eval_results_ood/global_step_35/results/_home_work_compass_innovation_minzijun_checkpoints_verl_role_sft_dapo_deepmath_train_sample_6144_context_4k_for_llama_Llama-3.2-1B-Instruct_max_response4096_batch1024_rollout8_vllm_True_bias0.15_global_step_35_actor_huggingface/results_2025-09-23T15-23-52.120403.json,0.4048059149722736,0.021123012121053563,0.5287769784172662,0.0005365061017861071,0.4565619223659889,0.021435222545538896,0.5827338129496403,0.0005348025912540751,,