Add outputs/pairs_4b

.gitattributes

@@ -39,3 +39,4 @@ logs/figures/E0-baseline_diagnostics.png filter=lfs diff=lfs merge=lfs -text
 logs/figures/E0-baseline_trajectory.png filter=lfs diff=lfs merge=lfs -text
 logs/figures/E0_entropy_live.png filter=lfs diff=lfs merge=lfs -text
 logs/figures/E0_live_dashboard.png filter=lfs diff=lfs merge=lfs -text
+outputs/pairs_4b/multipos_train.jsonl filter=lfs diff=lfs merge=lfs -text

outputs/pairs_4b/multipos_train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b22919c76fe36794155c8997040497c20a45783059d9dabc5080477172e733a
+size 18112100

outputs/pairs_4b/pair_stats_train.json

@@ -0,0 +1,41 @@
+{
+  "divpo-emb": {
+    "criterion": "divpo-emb",
+    "rho": 7.0,
+    "n_prompts": 1000,
+    "n_rows": 956,
+    "no_chosen": 12,
+    "no_rejected": 32,
+    "ok": 956,
+    "skip_rate": 0.04400000000000004,
+    "rho_adjusted_from": 6.0,
+    "reason": "skip_rate 0.314 > 0.3 (no_chosen=3, no_rejected=311); swept [3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0] -> rho=7.0"
+  },
+  "divpo-prob": {
+    "criterion": "divpo-prob",
+    "rho": 7.0,
+    "n_prompts": 1000,
+    "n_rows": 956,
+    "no_chosen": 12,
+    "no_rejected": 32,
+    "ok": 956,
+    "skip_rate": 0.04400000000000004,
+    "rho_adjusted_from": 6.0,
+    "reason": "skip_rate 0.314 > 0.3 (no_chosen=3, no_rejected=311); swept [3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0] -> rho=7.0"
+  },
+  "e3-multipos": {
+    "n_prompts": 1000,
+    "skipped_no_survivors": 3,
+    "skipped_no_negatives": 0,
+    "ok": 997,
+    "n_rows": 3593,
+    "mean_chosen_per_prompt": 3.9979939819458377,
+    "skip_rate": 0.0030000000000000027,
+    "weight_mean_after_norm": 1.0,
+    "weight_sd": 0.29810715542051675,
+    "neg_type_counts": {
+      "r_Q": 1991,
+      "r_D": 1602
+    }
+  }
+}