MsTok matched XSum production-20260921-v1: step 50002

mstok-matched-xsum/production-20260921-v1/step-50002/final-test-ema/metrics.json

@@ -0,0 +1,22 @@
+{
+  "step": 50002,
+  "variant": "ema",
+  "split": "test",
+  "documents": 11233,
+  "decoding": "greedy; predicted L0; 9 NFE",
+  "empty_predictions": 0,
+  "l0_code_accuracy": 0.02383601887296359,
+  "released_inference_parity": null,
+  "self": {
+    "rouge1": 21.74845737133523,
+    "rouge2": 6.022294818261974,
+    "rougeL": 17.19043867237548,
+    "gen_ppl": 454.9673104785133
+  },
+  "oracle": {
+    "rouge1": 26.92433902180462,
+    "rouge2": 9.029339327431492,
+    "rougeL": 22.836848985868198,
+    "gen_ppl": 380.03201320234047
+  }
+}

mstok-matched-xsum/production-20260921-v1/step-50002/final-test-raw/metrics.json

@@ -0,0 +1,22 @@
+{
+  "step": 50002,
+  "variant": "raw",
+  "split": "test",
+  "documents": 11233,
+  "decoding": "greedy; predicted L0; 9 NFE",
+  "empty_predictions": 3,
+  "l0_code_accuracy": 0.013754117332858543,
+  "released_inference_parity": null,
+  "self": {
+    "rouge1": 25.817297456459208,
+    "rouge2": 7.148712825704087,
+    "rougeL": 20.254191072449572,
+    "gen_ppl": 209.1351993121474
+  },
+  "oracle": {
+    "rouge1": 32.87071415085022,
+    "rouge2": 12.387869183100431,
+    "rougeL": 28.330920579831243,
+    "gen_ppl": 107.39147352650558
+  }
+}