初始化项目，由ModelHub XC社区提供模型

Model: saketh-chervu/rvr-exp21-s2_math-intermediate-correct
Source: Original Platform

eval_data/meta.json

@@ -0,0 +1,175 @@
+{
+  "exp": "exp21",
+  "cell": "s2_math",
+  "variant": "intermediate",
+  "arm": "correct",
+  "seed": 0,
+  "with_percent": true,
+  "composable": [
+    "gcf",
+    "lcm",
+    "multiply",
+    "percent_of"
+  ],
+  "distractors": [
+    "line_intercepts",
+    "rect_area_perimeter",
+    "venn"
+  ],
+  "atoms": [
+    "gcf",
+    "lcm",
+    "line_intercepts",
+    "multiply",
+    "percent_of",
+    "rect_area_perimeter",
+    "venn"
+  ],
+  "venn": "atom_only_correct_both_arms",
+  "train_comps": [
+    "gcf|multiply",
+    "gcf|percent_of",
+    "lcm|multiply",
+    "lcm|percent_of",
+    "multiply|gcf",
+    "multiply|lcm",
+    "percent_of|gcf",
+    "percent_of|lcm"
+  ],
+  "heldout_comps": [
+    "gcf|lcm",
+    "lcm|gcf",
+    "multiply|percent_of",
+    "percent_of|multiply"
+  ],
+  "heldout_labels": {
+    "gcf|lcm": "never_paired",
+    "lcm|gcf": "never_paired",
+    "multiply|percent_of": "never_paired",
+    "percent_of|multiply": "never_paired"
+  },
+  "n_comp_train_per_pairing": 150,
+  "total_comp_train": 1200,
+  "compositions": {
+    "gcf|multiply": {
+      "role": "train",
+      "n_train": 150,
+      "n_seen_eval": 200
+    },
+    "gcf|percent_of": {
+      "role": "train",
+      "n_train": 150,
+      "n_seen_eval": 200
+    },
+    "lcm|multiply": {
+      "role": "train",
+      "n_train": 150,
+      "n_seen_eval": 200
+    },
+    "lcm|percent_of": {
+      "role": "train",
+      "n_train": 150,
+      "n_seen_eval": 200
+    },
+    "multiply|gcf": {
+      "role": "train",
+      "n_train": 150,
+      "n_seen_eval": 200
+    },
+    "multiply|lcm": {
+      "role": "train",
+      "n_train": 150,
+      "n_seen_eval": 200
+    },
+    "percent_of|gcf": {
+      "role": "train",
+      "n_train": 150,
+      "n_seen_eval": 200
+    },
+    "percent_of|lcm": {
+      "role": "train",
+      "n_train": 150,
+      "n_seen_eval": 200
+    },
+    "gcf|lcm": {
+      "role": "heldout",
+      "label": "never_paired",
+      "n_eval": 200,
+      "distinct_answers": 174,
+      "top_answer_share": 0.015
+    },
+    "lcm|gcf": {
+      "role": "heldout",
+      "label": "never_paired",
+      "n_eval": 200,
+      "distinct_answers": 198,
+      "top_answer_share": 0.01
+    },
+    "multiply|percent_of": {
+      "role": "heldout",
+      "label": "never_paired",
+      "n_eval": 200,
+      "distinct_answers": 189,
+      "top_answer_share": 0.01
+    },
+    "percent_of|multiply": {
+      "role": "heldout",
+      "label": "never_paired",
+      "n_eval": 200,
+      "distinct_answers": 199,
+      "top_answer_share": 0.01
+    }
+  },
+  "train_heldout_chain_overlap": [],
+  "heldout_instance_collisions": 0,
+  "coverage": {
+    "train_inner_atoms": [
+      "gcf",
+      "lcm",
+      "multiply",
+      "percent_of"
+    ],
+    "train_outer_atoms": [
+      "gcf",
+      "lcm",
+      "multiply",
+      "percent_of"
+    ],
+    "heldout_inner_atoms": [
+      "gcf",
+      "lcm",
+      "multiply",
+      "percent_of"
+    ],
+    "heldout_outer_atoms": [
+      "gcf",
+      "lcm",
+      "multiply",
+      "percent_of"
+    ],
+    "heldout_inner_count": 4,
+    "heldout_outer_count": 4,
+    "per_slot_coverage": {
+      "gcf": {
+        "train_as_inner": 2,
+        "train_as_outer": 2
+      },
+      "lcm": {
+        "train_as_inner": 2,
+        "train_as_outer": 2
+      },
+      "multiply": {
+        "train_as_inner": 2,
+        "train_as_outer": 2
+      },
+      "percent_of": {
+        "train_as_inner": 2,
+        "train_as_outer": 2
+      }
+    },
+    "train_covers_both_slots": true,
+    "pairing_collisions": []
+  },
+  "collisions_total": 0,
+  "heldout_answer_entropy_bits": 9.49
+}