Add files using upload-large-folder tool

.gitattributes

@@ -33,3 +33,14 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+eval_results/global_step_0/mmlu_stem/test_qwen-boxed_-1_seed0_t1.0_s0_e-1.jsonl filter=lfs diff=lfs merge=lfs -text
+eval_results/plots/eval_results_acc_pass_acc.png filter=lfs diff=lfs merge=lfs -text
+eval_results/plots/eval_results_acc_keywords.png filter=lfs diff=lfs merge=lfs -text
+eval_results/plots/eval_results_acc_tokens.png filter=lfs diff=lfs merge=lfs -text
+eval_results/plots/eval_results_clip_ratio.png filter=lfs diff=lfs merge=lfs -text
+eval_results/plots/eval_results_avg_stop_tokens.png filter=lfs diff=lfs merge=lfs -text
+eval_results/plots/eval_results_correct_tokens.png filter=lfs diff=lfs merge=lfs -text
+eval_results/plots/eval_results_box_ratio_and_token_length.png filter=lfs diff=lfs merge=lfs -text
+eval_results/plots/eval_results_repeat_ratio_and_token_length.png filter=lfs diff=lfs merge=lfs -text
+eval_results/plots/eval_results_tokens_keywords.png filter=lfs diff=lfs merge=lfs -text
+eval_results/plots/eval_results_wrong_tokens.png filter=lfs diff=lfs merge=lfs -text

eval_results/global_step_0/mmlu_stem/test_qwen-boxed_-1_seed0_t1.0_s0_e-1.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19c04c176046e50be57a39956f1a999680c43c2e9a5a63b1b8e9db451450e240
+size 10911115

eval_results/global_step_80/aime25/test_qwen-boxed_-1_seed0_t1.0_s0_e-1_metrics.json

@@ -0,0 +1,13 @@
+{
+    "num_samples": 30,
+    "num_scores": 30,
+    "timeout_samples": 0,
+    "empty_samples": 0,
+    "acc": 6.7,
+    "pass_acc": 6.7,
+    "pass@k": {
+        "1": 6.7
+    },
+    "time_use_in_second": 133.5769498348236,
+    "time_use_in_minite": "2:13"
+}

eval_results/global_step_80/amc23/test_qwen-boxed_-1_seed0_t1.0_s0_e-1_metrics.json

@@ -0,0 +1,13 @@
+{
+    "num_samples": 40,
+    "num_scores": 40,
+    "timeout_samples": 0,
+    "empty_samples": 0,
+    "acc": 27.5,
+    "pass_acc": 27.5,
+    "pass@k": {
+        "1": 27.5
+    },
+    "time_use_in_second": 135.21881556510925,
+    "time_use_in_minite": "2:15"
+}

eval_results/global_step_80/gsm8k/test_qwen-boxed_-1_seed0_t1.0_s0_e-1_metrics.json

@@ -0,0 +1,13 @@
+{
+    "num_samples": 1319,
+    "num_scores": 1319,
+    "timeout_samples": 0,
+    "empty_samples": 0,
+    "acc": 80.4,
+    "pass_acc": 80.4,
+    "pass@k": {
+        "1": 80.4
+    },
+    "time_use_in_second": 211.01736974716187,
+    "time_use_in_minite": "3:31"
+}

eval_results/global_step_80/math500/test_qwen-boxed_-1_seed0_t1.0_s0_e-1_metrics.json

@@ -0,0 +1,13 @@
+{
+    "num_samples": 500,
+    "num_scores": 500,
+    "timeout_samples": 0,
+    "empty_samples": 1,
+    "acc": 62.0,
+    "pass_acc": 62.0,
+    "pass@k": {
+        "1": 62.0
+    },
+    "time_use_in_second": 229.07716965675354,
+    "time_use_in_minite": "3:49"
+}

eval_results/global_step_80/minerva_math/test_qwen-boxed_-1_seed0_t1.0_s0_e-1_metrics.json

@@ -0,0 +1,24 @@
+{
+    "num_samples": 272,
+    "num_scores": 272,
+    "timeout_samples": 2,
+    "empty_samples": 0,
+    "acc": 23.2,
+    "pass_acc": 23.2,
+    "pass@k": {
+        "1": 23.2
+    },
+    "type_acc": {
+        "Differential Equations (18.03 Spring 2010)": 47.9,
+        "Dynamics and Control (2.003 Spring 2005)": 42.3,
+        "Ecology I (1.018J Fall 2009)": 20.0,
+        "Information and Entropy (6.050J Spring 2008)": 33.3,
+        "Introduction to Astronomy (8.282J Spring 2006)": 11.3,
+        "Introduction to Solid State Chemistry (3.091 Fall 2010)": 9.3,
+        "Physical Chemistry (5.61 Fall 2017)": 0.0,
+        "Principles of Microeconomics (14.01 Fall 2011)": 61.1,
+        "Relativity (8.033 Fall 2006)": 9.1
+    },
+    "time_use_in_second": 70.31723284721375,
+    "time_use_in_minite": "1:10"
+}

eval_results/global_step_80/mmlu_stem/test_qwen-boxed_-1_seed0_t1.0_s0_e-1_metrics.json

@@ -0,0 +1,33 @@
+{
+    "num_samples": 3018,
+    "num_scores": 3018,
+    "timeout_samples": 1,
+    "empty_samples": 2,
+    "acc": 51.6,
+    "pass_acc": 51.6,
+    "pass@k": {
+        "1": 51.6
+    },
+    "type_acc": {
+        "abstract_algebra": 39.0,
+        "astronomy": 58.6,
+        "college_biology": 66.0,
+        "college_chemistry": 43.0,
+        "college_computer_science": 49.0,
+        "college_mathematics": 35.0,
+        "college_physics": 51.0,
+        "computer_security": 52.0,
+        "conceptual_physics": 60.4,
+        "electrical_engineering": 53.1,
+        "elementary_mathematics": 52.4,
+        "high_school_biology": 66.8,
+        "high_school_chemistry": 57.1,
+        "high_school_computer_science": 63.0,
+        "high_school_mathematics": 20.7,
+        "high_school_physics": 49.0,
+        "high_school_statistics": 53.7,
+        "machine_learning": 48.2
+    },
+    "time_use_in_second": 409.8389127254486,
+    "time_use_in_minite": "6:49"
+}

eval_results/global_step_80/olympiadbench/test_qwen-boxed_-1_seed0_t1.0_s0_e-1_metrics.json

@@ -0,0 +1,13 @@
+{
+    "num_samples": 675,
+    "num_scores": 675,
+    "timeout_samples": 0,
+    "empty_samples": 0,
+    "acc": 24.7,
+    "pass_acc": 24.7,
+    "pass@k": {
+        "1": 24.7
+    },
+    "time_use_in_second": 373.0475220680237,
+    "time_use_in_minite": "6:13"
+}

eval_results_ood/global_step_10/details/_home_work_compass_innovation_minzijun_checkpoints_verl_0.4_grpo_deepmath_train_sample_6k_context_4k_Qwen3-1.7B-base_max_response4096_batch1024_rollout8_vllm_inject_ratio0.2_global_step_10_actor_huggingface/2025-08-11T13-54-02.717538/details_extended|ifeval|0_2025-08-11T13-54-02.717538.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:01fcd2e10afaf6dcc75164958da3c2dedeb1f245c38e7bc72ac96cbcb59f3ee1
+size 1319087

eval_results_ood/global_step_10/details/_home_work_compass_innovation_minzijun_checkpoints_verl_0.4_grpo_deepmath_train_sample_6k_context_4k_Qwen3-1.7B-base_max_response4096_batch1024_rollout8_vllm_inject_ratio0.2_global_step_10_actor_huggingface/2025-08-11T13-54-02.717538/details_lighteval|gpqa:diamond|0_2025-08-11T13-54-02.717538.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:762cdb6f6149a49e1a865ee2f738b702b917544de3e0de39c0287ecfa03aaab9
+size 473589

eval_results_ood/global_step_10/results/_home_work_compass_innovation_minzijun_checkpoints_verl_0.4_grpo_deepmath_train_sample_6k_context_4k_Qwen3-1.7B-base_max_response4096_batch1024_rollout8_vllm_inject_ratio0.2_global_step_10_actor_huggingface/results_2025-08-11T13-54-02.717538.json

@@ -0,0 +1,200 @@
+{
+  "config_general": {
+    "lighteval_sha": "327071fe86e427d880f907a51d1462f4a3f951c1",
+    "num_fewshot_seeds": 1,
+    "max_samples": null,
+    "job_id": 0,
+    "start_time": 7147045.502092421,
+    "end_time": 7148825.263175826,
+    "total_evaluation_time_secondes": "1779.7610834045336",
+    "model_name": "_home_work_compass_innovation_minzijun_checkpoints_verl_0.4_grpo_deepmath_train_sample_6k_context_4k_Qwen3-1.7B-base_max_response4096_batch1024_rollout8_vllm_inject_ratio0.2_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": 0.17929759704251386,
+      "prompt_level_strict_acc_stderr": 0.01650757226566346,
+      "inst_level_strict_acc": 0.29376498800959233,
+      "inst_level_strict_acc_stderr": 0.00047473066128257376,
+      "prompt_level_loose_acc": 0.19038817005545286,
+      "prompt_level_loose_acc_stderr": 0.01689512940763452,
+      "inst_level_loose_acc": 0.3117505995203837,
+      "inst_level_loose_acc_stderr": 0.0004811713762249123
+    },
+    "lighteval|gpqa:diamond|0": {
+      "gpqa_pass@1:1_samples": 0.24242424242424243,
+      "gpqa_pass@1:1_samples_stderr": 0.030532892233932046
+    },
+    "all": {
+      "prompt_level_strict_acc": 0.17929759704251386,
+      "prompt_level_strict_acc_stderr": 0.01650757226566346,
+      "inst_level_strict_acc": 0.29376498800959233,
+      "inst_level_strict_acc_stderr": 0.00047473066128257376,
+      "prompt_level_loose_acc": 0.19038817005545286,
+      "prompt_level_loose_acc_stderr": 0.01689512940763452,
+      "inst_level_loose_acc": 0.3117505995203837,
+      "inst_level_loose_acc_stderr": 0.0004811713762249123,
+      "gpqa_pass@1:1_samples": 0.24242424242424243,
+      "gpqa_pass@1:1_samples_stderr": 0.030532892233932046
+    }
+  },
+  "versions": {
+    "extended|ifeval|0": "0.1",
+    "lighteval|gpqa:diamond|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"
+    },
+    "lighteval|gpqa:diamond": {
+      "name": "gpqa:diamond",
+      "prompt_function": "gpqa_instruct",
+      "hf_repo": "Idavidrein/gpqa",
+      "hf_subset": "gpqa_diamond",
+      "metric": [
+        {
+          "metric_name": "gpqa_pass@1:1_samples",
+          "higher_is_better": true,
+          "category": "5",
+          "use_case": "6",
+          "sample_level_fn": "compute",
+          "corpus_level_fn": "mean"
+        }
+      ],
+      "hf_revision": null,
+      "hf_filter": null,
+      "hf_avail_splits": [
+        "train"
+      ],
+      "trust_dataset": true,
+      "evaluation_splits": [
+        "train"
+      ],
+      "few_shots_split": null,
+      "few_shots_select": null,
+      "generation_size": 32768,
+      "generation_grammar": null,
+      "stop_sequence": [],
+      "num_samples": null,
+      "suite": [
+        "lighteval"
+      ],
+      "original_num_docs": 198,
+      "effective_num_docs": 198,
+      "must_remove_duplicate_docs": false,
+      "version": 1
+    }
+  },
+  "summary_tasks": {
+    "extended|ifeval|0": {
+      "hashes": {
+        "hash_examples": "e99cbf567588d7c6",
+        "hash_full_prompts": "7ea7bf2a8edba8f4",
+        "hash_input_tokens": "e3d19e04074f1062",
+        "hash_cont_tokens": "df73b4a833549b71"
+      },
+      "truncated": 0,
+      "non_truncated": 541,
+      "padded": 0,
+      "non_padded": 541,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    },
+    "lighteval|gpqa:diamond|0": {
+      "hashes": {
+        "hash_examples": "50ecb6f5d091bd95",
+        "hash_full_prompts": "1b19c7f64e1e9b2a",
+        "hash_input_tokens": "864f299da9b1369e",
+        "hash_cont_tokens": "280f507ebf368545"
+      },
+      "truncated": 0,
+      "non_truncated": 198,
+      "padded": 0,
+      "non_padded": 198,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    }
+  },
+  "summary_general": {
+    "hashes": {
+      "hash_examples": "ca21d46c94a77f9f",
+      "hash_full_prompts": "e2f5aa48878d9aa5",
+      "hash_input_tokens": "46813bee698c07ed",
+      "hash_cont_tokens": "626fb001d9019c41"
+    },
+    "truncated": 0,
+    "non_truncated": 739,
+    "padded": 0,
+    "non_padded": 739,
+    "num_truncated_few_shots": 0
+  }
+}

eval_results_ood/global_step_20/details/_home_work_compass_innovation_minzijun_checkpoints_verl_0.4_grpo_deepmath_train_sample_6k_context_4k_Qwen3-1.7B-base_max_response4096_batch1024_rollout8_vllm_inject_ratio0.2_global_step_20_actor_huggingface/2025-08-11T13-53-51.898028/details_extended|ifeval|0_2025-08-11T13-53-51.898028.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b8ff5ed26c2a30fc0f0d903ffba357e67fdd5c92a1154cc6f3d2eb4f98ae3886
+size 1148399

eval_results_ood/global_step_20/details/_home_work_compass_innovation_minzijun_checkpoints_verl_0.4_grpo_deepmath_train_sample_6k_context_4k_Qwen3-1.7B-base_max_response4096_batch1024_rollout8_vllm_inject_ratio0.2_global_step_20_actor_huggingface/2025-08-11T13-53-51.898028/details_lighteval|gpqa:diamond|0_2025-08-11T13-53-51.898028.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76b5779a30095639d20a789d03a89c877333bc6fa6eb0bb093c5fb1f01a60042
+size 482167

eval_results_ood/global_step_20/results/_home_work_compass_innovation_minzijun_checkpoints_verl_0.4_grpo_deepmath_train_sample_6k_context_4k_Qwen3-1.7B-base_max_response4096_batch1024_rollout8_vllm_inject_ratio0.2_global_step_20_actor_huggingface/results_2025-08-11T13-53-51.898028.json

@@ -0,0 +1,200 @@
+{
+  "config_general": {
+    "lighteval_sha": "327071fe86e427d880f907a51d1462f4a3f951c1",
+    "num_fewshot_seeds": 1,
+    "max_samples": null,
+    "job_id": 0,
+    "start_time": 7147044.129554702,
+    "end_time": 7148814.518096581,
+    "total_evaluation_time_secondes": "1770.3885418791324",
+    "model_name": "_home_work_compass_innovation_minzijun_checkpoints_verl_0.4_grpo_deepmath_train_sample_6k_context_4k_Qwen3-1.7B-base_max_response4096_batch1024_rollout8_vllm_inject_ratio0.2_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": 0.17375231053604437,
+      "prompt_level_strict_acc_stderr": 0.016305102877651502,
+      "inst_level_strict_acc": 0.2973621103117506,
+      "inst_level_strict_acc_stderr": 0.00045905749049021145,
+      "prompt_level_loose_acc": 0.19408502772643252,
+      "prompt_level_loose_acc_stderr": 0.0170193805507494,
+      "inst_level_loose_acc": 0.3213429256594724,
+      "inst_level_loose_acc_stderr": 0.0004682830366257211
+    },
+    "lighteval|gpqa:diamond|0": {
+      "gpqa_pass@1:1_samples": 0.25252525252525254,
+      "gpqa_pass@1:1_samples_stderr": 0.030954055470365897
+    },
+    "all": {
+      "prompt_level_strict_acc": 0.17375231053604437,
+      "prompt_level_strict_acc_stderr": 0.016305102877651502,
+      "inst_level_strict_acc": 0.2973621103117506,
+      "inst_level_strict_acc_stderr": 0.00045905749049021145,
+      "prompt_level_loose_acc": 0.19408502772643252,
+      "prompt_level_loose_acc_stderr": 0.0170193805507494,
+      "inst_level_loose_acc": 0.3213429256594724,
+      "inst_level_loose_acc_stderr": 0.0004682830366257211,
+      "gpqa_pass@1:1_samples": 0.25252525252525254,
+      "gpqa_pass@1:1_samples_stderr": 0.030954055470365897
+    }
+  },
+  "versions": {
+    "extended|ifeval|0": "0.1",
+    "lighteval|gpqa:diamond|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"
+    },
+    "lighteval|gpqa:diamond": {
+      "name": "gpqa:diamond",
+      "prompt_function": "gpqa_instruct",
+      "hf_repo": "Idavidrein/gpqa",
+      "hf_subset": "gpqa_diamond",
+      "metric": [
+        {
+          "metric_name": "gpqa_pass@1:1_samples",
+          "higher_is_better": true,
+          "category": "5",
+          "use_case": "6",
+          "sample_level_fn": "compute",
+          "corpus_level_fn": "mean"
+        }
+      ],
+      "hf_revision": null,
+      "hf_filter": null,
+      "hf_avail_splits": [
+        "train"
+      ],
+      "trust_dataset": true,
+      "evaluation_splits": [
+        "train"
+      ],
+      "few_shots_split": null,
+      "few_shots_select": null,
+      "generation_size": 32768,
+      "generation_grammar": null,
+      "stop_sequence": [],
+      "num_samples": null,
+      "suite": [
+        "lighteval"
+      ],
+      "original_num_docs": 198,
+      "effective_num_docs": 198,
+      "must_remove_duplicate_docs": false,
+      "version": 1
+    }
+  },
+  "summary_tasks": {
+    "extended|ifeval|0": {
+      "hashes": {
+        "hash_examples": "e99cbf567588d7c6",
+        "hash_full_prompts": "7ea7bf2a8edba8f4",
+        "hash_input_tokens": "e3d19e04074f1062",
+        "hash_cont_tokens": "d0a4f441f3ae0abb"
+      },
+      "truncated": 0,
+      "non_truncated": 541,
+      "padded": 0,
+      "non_padded": 541,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    },
+    "lighteval|gpqa:diamond|0": {
+      "hashes": {
+        "hash_examples": "50ecb6f5d091bd95",
+        "hash_full_prompts": "1b19c7f64e1e9b2a",
+        "hash_input_tokens": "864f299da9b1369e",
+        "hash_cont_tokens": "fb850e8444c817de"
+      },
+      "truncated": 0,
+      "non_truncated": 198,
+      "padded": 0,
+      "non_padded": 198,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    }
+  },
+  "summary_general": {
+    "hashes": {
+      "hash_examples": "ca21d46c94a77f9f",
+      "hash_full_prompts": "e2f5aa48878d9aa5",
+      "hash_input_tokens": "46813bee698c07ed",
+      "hash_cont_tokens": "bd6b48c96408fb0a"
+    },
+    "truncated": 0,
+    "non_truncated": 739,
+    "padded": 0,
+    "non_padded": 739,
+    "num_truncated_few_shots": 0
+  }
+}

eval_results_ood/global_step_30/details/_home_work_compass_innovation_minzijun_checkpoints_verl_0.4_grpo_deepmath_train_sample_6k_context_4k_Qwen3-1.7B-base_max_response4096_batch1024_rollout8_vllm_inject_ratio0.2_global_step_30_actor_huggingface/2025-08-11T13-52-18.568882/details_extended|ifeval|0_2025-08-11T13-52-18.568882.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:27655defdf192c1baf5b56f66799618b9e5614ea6d5f9430172ce452510f12cb
+size 1067204

eval_results_ood/global_step_30/details/_home_work_compass_innovation_minzijun_checkpoints_verl_0.4_grpo_deepmath_train_sample_6k_context_4k_Qwen3-1.7B-base_max_response4096_batch1024_rollout8_vllm_inject_ratio0.2_global_step_30_actor_huggingface/2025-08-11T13-52-18.568882/details_lighteval|gpqa:diamond|0_2025-08-11T13-52-18.568882.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:97c1624c2bdbac3bcf7062d5a5e2d3ca136be0e344800327e4a1dc6c9676f880
+size 450151

eval_results_ood/global_step_30/results/_home_work_compass_innovation_minzijun_checkpoints_verl_0.4_grpo_deepmath_train_sample_6k_context_4k_Qwen3-1.7B-base_max_response4096_batch1024_rollout8_vllm_inject_ratio0.2_global_step_30_actor_huggingface/results_2025-08-11T13-52-18.568882.json

@@ -0,0 +1,200 @@
+{
+  "config_general": {
+    "lighteval_sha": "327071fe86e427d880f907a51d1462f4a3f951c1",
+    "num_fewshot_seeds": 1,
+    "max_samples": null,
+    "job_id": 0,
+    "start_time": 7147044.559507423,
+    "end_time": 7148721.314530517,
+    "total_evaluation_time_secondes": "1676.7550230938941",
+    "model_name": "_home_work_compass_innovation_minzijun_checkpoints_verl_0.4_grpo_deepmath_train_sample_6k_context_4k_Qwen3-1.7B-base_max_response4096_batch1024_rollout8_vllm_inject_ratio0.2_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": 0.18853974121996303,
+      "prompt_level_strict_acc_stderr": 0.016832096060176906,
+      "inst_level_strict_acc": 0.31654676258992803,
+      "inst_level_strict_acc_stderr": 0.0005027673145846779,
+      "prompt_level_loose_acc": 0.20702402957486138,
+      "prompt_level_loose_acc_stderr": 0.017435865587996556,
+      "inst_level_loose_acc": 0.33573141486810554,
+      "inst_level_loose_acc_stderr": 0.0005153869493324995
+    },
+    "lighteval|gpqa:diamond|0": {
+      "gpqa_pass@1:1_samples": 0.2474747474747475,
+      "gpqa_pass@1:1_samples_stderr": 0.030746300742124522
+    },
+    "all": {
+      "prompt_level_strict_acc": 0.18853974121996303,
+      "prompt_level_strict_acc_stderr": 0.016832096060176906,
+      "inst_level_strict_acc": 0.31654676258992803,
+      "inst_level_strict_acc_stderr": 0.0005027673145846779,
+      "prompt_level_loose_acc": 0.20702402957486138,
+      "prompt_level_loose_acc_stderr": 0.017435865587996556,
+      "inst_level_loose_acc": 0.33573141486810554,
+      "inst_level_loose_acc_stderr": 0.0005153869493324995,
+      "gpqa_pass@1:1_samples": 0.2474747474747475,
+      "gpqa_pass@1:1_samples_stderr": 0.030746300742124522
+    }
+  },
+  "versions": {
+    "extended|ifeval|0": "0.1",
+    "lighteval|gpqa:diamond|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"
+    },
+    "lighteval|gpqa:diamond": {
+      "name": "gpqa:diamond",
+      "prompt_function": "gpqa_instruct",
+      "hf_repo": "Idavidrein/gpqa",
+      "hf_subset": "gpqa_diamond",
+      "metric": [
+        {
+          "metric_name": "gpqa_pass@1:1_samples",
+          "higher_is_better": true,
+          "category": "5",
+          "use_case": "6",
+          "sample_level_fn": "compute",
+          "corpus_level_fn": "mean"
+        }
+      ],
+      "hf_revision": null,
+      "hf_filter": null,
+      "hf_avail_splits": [
+        "train"
+      ],
+      "trust_dataset": true,
+      "evaluation_splits": [
+        "train"
+      ],
+      "few_shots_split": null,
+      "few_shots_select": null,
+      "generation_size": 32768,
+      "generation_grammar": null,
+      "stop_sequence": [],
+      "num_samples": null,
+      "suite": [
+        "lighteval"
+      ],
+      "original_num_docs": 198,
+      "effective_num_docs": 198,
+      "must_remove_duplicate_docs": false,
+      "version": 1
+    }
+  },
+  "summary_tasks": {
+    "extended|ifeval|0": {
+      "hashes": {
+        "hash_examples": "e99cbf567588d7c6",
+        "hash_full_prompts": "7ea7bf2a8edba8f4",
+        "hash_input_tokens": "e3d19e04074f1062",
+        "hash_cont_tokens": "74ba40795441d16d"
+      },
+      "truncated": 0,
+      "non_truncated": 541,
+      "padded": 0,
+      "non_padded": 541,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    },
+    "lighteval|gpqa:diamond|0": {
+      "hashes": {
+        "hash_examples": "50ecb6f5d091bd95",
+        "hash_full_prompts": "1b19c7f64e1e9b2a",
+        "hash_input_tokens": "864f299da9b1369e",
+        "hash_cont_tokens": "8f8b5ab6cbffdeea"
+      },
+      "truncated": 0,
+      "non_truncated": 198,
+      "padded": 0,
+      "non_padded": 198,
+      "effective_few_shots": 0.0,
+      "num_truncated_few_shots": 0
+    }
+  },
+  "summary_general": {
+    "hashes": {
+      "hash_examples": "ca21d46c94a77f9f",
+      "hash_full_prompts": "e2f5aa48878d9aa5",
+      "hash_input_tokens": "46813bee698c07ed",
+      "hash_cont_tokens": "f714f6cab6d3b0d0"
+    },
+    "truncated": 0,
+    "non_truncated": 739,
+    "padded": 0,
+    "non_padded": 739,
+    "num_truncated_few_shots": 0
+  }
+}

eval_results_ood/global_step_40/details/_home_work_compass_innovation_minzijun_checkpoints_verl_0.4_grpo_deepmath_train_sample_6k_context_4k_Qwen3-1.7B-base_max_response4096_batch1024_rollout8_vllm_inject_ratio0.2_global_step_40_actor_huggingface/2025-08-11T13-49-32.356892/details_extended|ifeval|0_2025-08-11T13-49-32.356892.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ecece71b107f07448310a11cfd1afa7fdca841bb81c958c7a3e5140e3fb4726a
+size 989784