Upload folder using huggingface_hub

.gitattributes

@@ -38,3 +38,4 @@ Qwen/Qwen2.5-Math-7B/math_eval/olympiadbench/test_qwen-boxed_-1_seed0_t1.0_s0_e-
 stellalisy/rethink_rlvr_reproduce-ground_truth-qwen2.5_math_7b-lr5e-7-kl0.00-step150/math_eval/minerva_math/test_qwen-boxed_-1_seed0_t1.0_s0_e-1.jsonl filter=lfs diff=lfs merge=lfs -text
 stellalisy/rethink_rlvr_reproduce-ground_truth-qwen2.5_math_7b-lr5e-7-kl0.00-step150/math_eval/olympiadbench/test_qwen-boxed_-1_seed0_t1.0_s0_e-1.jsonl filter=lfs diff=lfs merge=lfs -text
 stellalisy/rethink_rlvr_reproduce-majority_vote-qwen2.5_math_7b-lr5e-7-kl0.00-step150/math_eval/minerva_math/test_qwen-boxed_-1_seed0_t1.0_s0_e-1.jsonl filter=lfs diff=lfs merge=lfs -text
+stellalisy/rethink_rlvr_reproduce-random-qwen2.5_math_7b-lr5e-7-kl0.00-step150/math_eval/minerva_math/test_qwen-boxed_-1_seed0_t1.0_s0_e-1.jsonl filter=lfs diff=lfs merge=lfs -text

stellalisy/rethink_rlvr_reproduce-random-qwen2.5_math_7b-lr5e-7-kl0.00-step150/math_eval/minerva_math/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:e117423d7b63115fd58311df9ad38812ca6318f4164a1adcff727e6c00b6e7d5
+size 724443040

stellalisy/rethink_rlvr_reproduce-random-qwen2.5_math_7b-lr5e-7-kl0.00-step150/math_eval/minerva_math/test_qwen-boxed_-1_seed0_t1.0_s0_e-1_qwen-boxed_metrics.json

@@ -0,0 +1,20 @@
+{
+    "num_samples": 272,
+    "num_scores": 278528,
+    "timeout_samples": 17707,
+    "empty_samples": 34,
+    "acc": 17.3,
+    "type_acc": {
+        "Differential Equations (18.03 Spring 2010)": 18.8,
+        "Dynamics and Control (2.003 Spring 2005)": 15.4,
+        "Ecology I (1.018J Fall 2009)": 0.0,
+        "Information and Entropy (6.050J Spring 2008)": 33.3,
+        "Introduction to Astronomy (8.282J Spring 2006)": 17.0,
+        "Introduction to Solid State Chemistry (3.091 Fall 2010)": 15.5,
+        "Physical Chemistry (5.61 Fall 2017)": 9.1,
+        "Principles of Microeconomics (14.01 Fall 2011)": 38.9,
+        "Relativity (8.033 Fall 2006)": 9.1
+    },
+    "time_use_in_second": 27258.39978671074,
+    "time_use_in_minite": "454:18"
+}