Eval set with grids and verdicts

README.md

@@ -0,0 +1,47 @@
+---
+dataset_info:
+  features:
+  - name: episode_index
+    dtype: int64
+  - name: grid
+    dtype: image
+  - name: human_success
+    dtype: bool
+  - name: instruction
+    dtype: large_string
+  - name: frames
+    dtype: int64
+  - name: seconds
+    dtype: float64
+  - name: videos
+    dtype: large_string
+  - name: judge_model
+    dtype: large_string
+  - name: judge_success
+    dtype: bool
+  - name: judge_confidence
+    dtype: float64
+  - name: judge_p_success
+    dtype: float64
+  - name: judge_failure_mode
+    dtype: large_string
+  - name: judge_evidence
+    dtype: large_string
+  - name: judge_attempts
+    dtype: int64
+  - name: judge_input_tokens
+    dtype: int64
+  - name: judge_output_tokens
+    dtype: int64
+  splits:
+  - name: test
+    num_bytes: 56216975
+    num_examples: 200
+  download_size: 56116348
+  dataset_size: 56216975
+configs:
+- config_name: default
+  data_files:
+  - split: test
+    path: data/test-*
+---

data/test-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63974e6b41ce4099704b3c8aacf8a34972375f8fd2363485f07712c600116102
+size 56116348