mini-beatrix-3: the stage arms (arms off and arms on in one package)

README.md

@@ -181,14 +181,14 @@ whole stage by itself.
 
 | arm | its stage's text | stage text, arm off → on (bpb) | web text (bpb) | stage items in the stage's own form | stage items in a new form | seeds |
 |---|---|---|---|---|---|---|
-| `solo/s1_perspective` | one small event retold from each side: seen from a person, said to them, told about them | 0.554 → 0.081 (a second seed: 0.080) | +0.0002 | 55% → 75% | 46% → 41% | two |
-| `solo/s2_concept` | kinds, properties and differences: what a thing is a kind of, what its kind can do, how two things differ | 0.721 → 0.120 (a second seed: 0.120) | +0.0003 | 30% → 84% | 68% → 73% | two |
-| `solo/s3_rules` | if-then rules over made-up words, followed step by step to what follows | 0.930 → 0.372 (a second seed: 0.374) | +0.0009 | 16% → 60% | 7% → 12% | two |
-| `solo/s4_arith` | small arithmetic worked out in text | 1.234 → 0.389 (a second seed: 0.388) | +0.0007 | 34% → 49% | 24% → 24% | two |
-| `solo/s5_causal` | short cause-and-effect text: what happened and why | 0.660 → 0.068 | +0.0003 | not measured | not measured | one |
-| `solo/s6_tryfail` | an attempt, its failure and the revised attempt | 0.673 → 0.114 | +0.0008 | not measured | not measured | one |
-| `solo/s7_mixed` | the mixed stage's diet: the earlier stages' text beside ordinary prose | 0.783 → 0.755 | +0.0004 | not measured | not measured | one |
-| `solo/s8_register` | the same content in different registers of speech | 0.491 → 0.051 | +0.0008 | not measured | not measured | one |
+| `solo/s1_perspective` | one small event retold from each side: seen from a person, said to them, told about them | 0.554 → 0.047 (a second seed: 0.047) | +0.0003 | 55% → 89% | 46% → 42% | two |
+| `solo/s2_concept` | kinds, properties and differences: what a thing is a kind of, what its kind can do, how two things differ | 0.721 → 0.075 (a second seed: 0.075) | +0.0001 | 30% → 100% | 68% → 70% | two |
+| `solo/s3_rules` | if-then rules over made-up words, followed step by step to what follows | 0.930 → 0.180 (a second seed: 0.179) | +0.0003 | 16% → 82% | 7% → 22% | two |
+| `solo/s4_arith` | small arithmetic worked out in text | 1.234 → 0.267 (a second seed: 0.267) | +0.0002 | 34% → 83% | 24% → 21% | two |
+| `solo/s5_causal` | short cause-and-effect text: what happened and why | 0.660 → 0.043 (a second seed: 0.043) | +0.0006 | not measured | not measured | two |
+| `solo/s6_tryfail` | an attempt, its failure and the revised attempt | 0.673 → 0.057 (a second seed: 0.057) | +0.0003 | not measured | not measured | two |
+| `solo/s7_mixed` | the mixed stage's diet: the earlier stages' text beside ordinary prose | 0.783 → 0.715 (a second seed: 0.717) | +0.0000 | not measured | not measured | two |
+| `solo/s8_register` | the same content in different registers of speech | 0.491 → 0.021 (a second seed: 0.021) | +0.0003 | not measured | not measured | two |
 
 ```python
 m.mount_arm("solo/s1_perspective")  # one solo arm; mounting another swaps it
@@ -296,7 +296,7 @@ trunk for each stage; those arms stayed nearly empty, because the trunk took
 each stage's text in before its arm could, and they were detached at step
 148,000 (their files are in the training repo under `mini-beatrix-3/arms/`,
 bound to those earlier trunk states). The arms in this repository were fitted
-afterwards on the finished, frozen core: as always-on groups and as solo arms. The four-arm group in two steps (first the arms of stages 1 and 2, in a run that added the four stages' text one stage at a time with every attached arm training under one pure Adam, 800 steps per stage; then, with those two held fixed and on, a new arm for stage 3 and after it a new arm for stage 4, 800 steps each). The eight-arm group by continuing that run: with the four held fixed and on, a new arm for stage 5, then 6, then 7, then 8, each attached over every arm before it and trained for 4,000 steps in their presence. Every group arm was kept quiet on ordinary web text and on the stage text of every other arm in its group. Each solo arm was trained alone, 800 steps, kept quiet on web text.
+afterwards on the finished, frozen core: as always-on groups and as solo arms. The four-arm group in two steps (first the arms of stages 1 and 2, in a run that added the four stages' text one stage at a time with every attached arm training under one pure Adam, 800 steps per stage; then, with those two held fixed and on, a new arm for stage 3 and after it a new arm for stage 4, 800 steps each). The eight-arm group by continuing that run: with the four held fixed and on, a new arm for stage 5, then 6, then 7, then 8, each attached over every arm before it and trained for 4,000 steps in their presence. Every group arm was kept quiet on ordinary web text and on the stage text of every other arm in its group. Each solo arm was trained alone for 4,000 steps (two seeds), kept quiet on web text.
 
 ## Lineage
 

arms/index.json

@@ -6,8 +6,8 @@
   "base_model_id": "alephllm/mini-beatrix-3@step245674",
   "note": "every packaged arm was trained on this exact frozen core (the default bf16 weight file); arms are trunk-bound"
  },
- "generated": "2026-10-06 04:54 UTC",
- "note": "the stage arms of curriculum stages 1 to 8, fitted on the finished, frozen core: two always-on groups (the four arms of stages 1 to 4; the eight arms of stages 1 to 8, which are those four, held fixed, with four more fitted over them) and each stage's arm alone; the arms trained beside the trunk during the run are on the training repo",
+ "generated": "2026-10-06 15:18 UTC",
+ "note": "the stage arms of curriculum stages 1 to 8, fitted on the finished, frozen core: two always-on groups (the four arms of stages 1 to 4; the eight arms of stages 1 to 8, which are those four, held fixed, with four more fitted over them) and each stage's arm alone (4,000 steps, two seeds); the arms trained beside the trunk during the run are on the training repo",
  "arms": [
   {
    "id": "stages-1-4/s1_perspective",
@@ -1577,54 +1577,54 @@
     "precision": "fp32"
    },
    "source": {
-    "file": "s1_perspective_fresh_seed1_d20261005_step800.safetensors",
-    "sha256": "ff72dd5a8061926b56e09c6da765c1d1ee44cbbd3d4e5e2c30793e04b369532a"
+    "file": "s1_perspective_fresh_seed1_d20261005_step4000.safetensors",
+    "sha256": "13290e04d7b7f9e2f26d3b3809afcab54306f3aa7b1d74d73726836a32624177"
    },
-   "recipe": "trained alone on the frozen final core: 800 steps of 2 x 4096 bytes of the stage's own text, pure Adam (lr 0.001), with a quiet term at weight 2 on held-out web text (the KL from the model with the arm off to the model with it on, one chunk beside every task chunk)",
+   "recipe": "trained alone on the frozen final core: 4000 steps of 2 x 4096 bytes of the stage's own text, pure Adam (lr 0.001), with a quiet term at weight 2 on held-out web text (the KL from the model with the arm off to the model with it on, one chunk beside every task chunk)",
    "title": "Stage 1: perspective (solo)",
    "behavior": "one small event retold from each side: seen from a person, said to them, told about them",
    "family": "solo",
    "status": "two seeds; trained alone: mount one solo arm at a time",
-   "score": "own stage text 0.554 -> 0.081 bpb; web text +0.0002 bpb",
+   "score": "own stage text 0.554 -> 0.047 bpb; web text +0.0003 bpb",
    "measured": {
-    "steps": 800,
+    "steps": 4000,
     "own_stage_bpb_arms_off": 0.5538,
-    "own_stage_bpb_arm_on": 0.0809,
-    "own_stage_difference": -0.4729,
-    "own_stage_difference_se": 0.0038,
-    "web_text_difference": 0.00018,
-    "web_text_difference_se": 0.00019,
+    "own_stage_bpb_arm_on": 0.0469,
+    "own_stage_difference": -0.5069,
+    "own_stage_difference_se": 0.0048,
+    "web_text_difference": 0.00026,
+    "web_text_difference_se": 0.00011,
     "stage_items_own_form_accuracy": {
      "arms_off": 0.55,
-     "arm_on": 0.75
+     "arm_on": 0.887
     },
     "stage_items_new_form_accuracy": {
      "arms_off": 0.463,
-     "arm_on": 0.412
+     "arm_on": 0.425
     },
     "probe_suite_mean": {
      "arms_off": 0.411,
-     "arm_on": 0.381
+     "arm_on": 0.411
     },
     "second_seed": {
-     "steps": 800,
+     "steps": 4000,
      "own_stage_bpb_arms_off": 0.5538,
-     "own_stage_bpb_arm_on": 0.0802,
-     "own_stage_difference": -0.4737,
-     "own_stage_difference_se": 0.0038,
-     "web_text_difference": 0.00036,
-     "web_text_difference_se": 0.00016,
+     "own_stage_bpb_arm_on": 0.0467,
+     "own_stage_difference": -0.5071,
+     "own_stage_difference_se": 0.0048,
+     "web_text_difference": 0.00031,
+     "web_text_difference_se": 0.00013,
      "stage_items_own_form_accuracy": {
       "arms_off": 0.55,
-      "arm_on": 0.7
+      "arm_on": 1.0
      },
      "stage_items_new_form_accuracy": {
       "arms_off": 0.463,
-      "arm_on": 0.4
+      "arm_on": 0.5
      },
      "probe_suite_mean": {
       "arms_off": 0.411,
-      "arm_on": 0.385
+      "arm_on": 0.407
      }
     }
    },
@@ -1658,54 +1658,54 @@
     "precision": "fp32"
    },
    "source": {
-    "file": "s2_concept_fresh_seed2_d20261005_step800.safetensors",
-    "sha256": "dc76f7c7d713b34d23b553adbb0b6302ec1d1080d7ffe8cc9461c5f5e6a03938"
+    "file": "s2_concept_fresh_seed2_d20261005_step4000.safetensors",
+    "sha256": "a07fbca8d7f171475c1b91509589199323dfed929205df2fb2c55dd8d7e1f507"
    },
-   "recipe": "trained alone on the frozen final core: 800 steps of 2 x 4096 bytes of the stage's own text, pure Adam (lr 0.001), with a quiet term at weight 2 on held-out web text (the KL from the model with the arm off to the model with it on, one chunk beside every task chunk)",
+   "recipe": "trained alone on the frozen final core: 4000 steps of 2 x 4096 bytes of the stage's own text, pure Adam (lr 0.001), with a quiet term at weight 2 on held-out web text (the KL from the model with the arm off to the model with it on, one chunk beside every task chunk)",
    "title": "Stage 2: concepts (solo)",
    "behavior": "kinds, properties and differences: what a thing is a kind of, what its kind can do, how two things differ",
    "family": "solo",
    "status": "two seeds; trained alone: mount one solo arm at a time",
-   "score": "own stage text 0.721 -> 0.120 bpb; web text +0.0003 bpb",
+   "score": "own stage text 0.721 -> 0.075 bpb; web text +0.0001 bpb",
    "measured": {
-    "steps": 800,
+    "steps": 4000,
     "own_stage_bpb_arms_off": 0.7212,
-    "own_stage_bpb_arm_on": 0.1204,
-    "own_stage_difference": -0.6008,
-    "own_stage_difference_se": 0.0034,
-    "web_text_difference": 0.00029,
-    "web_text_difference_se": 0.00016,
+    "own_stage_bpb_arm_on": 0.0754,
+    "own_stage_difference": -0.6458,
+    "own_stage_difference_se": 0.0038,
+    "web_text_difference": 0.00014,
+    "web_text_difference_se": 0.00014,
     "stage_items_own_form_accuracy": {
      "arms_off": 0.298,
-     "arm_on": 0.845
+     "arm_on": 1.0
     },
     "stage_items_new_form_accuracy": {
      "arms_off": 0.678,
-     "arm_on": 0.729
+     "arm_on": 0.695
     },
     "probe_suite_mean": {
      "arms_off": 0.411,
-     "arm_on": 0.404
+     "arm_on": 0.389
     },
     "second_seed": {
-     "steps": 800,
+     "steps": 4000,
      "own_stage_bpb_arms_off": 0.7212,
-     "own_stage_bpb_arm_on": 0.1199,
-     "own_stage_difference": -0.6013,
-     "own_stage_difference_se": 0.0035,
-     "web_text_difference": 0.00043,
-     "web_text_difference_se": 0.00018,
+     "own_stage_bpb_arm_on": 0.0753,
+     "own_stage_difference": -0.6459,
+     "own_stage_difference_se": 0.0036,
+     "web_text_difference": 3e-05,
+     "web_text_difference_se": 0.0001,
      "stage_items_own_form_accuracy": {
       "arms_off": 0.298,
-      "arm_on": 0.881
+      "arm_on": 0.988
      },
      "stage_items_new_form_accuracy": {
       "arms_off": 0.678,
-      "arm_on": 0.695
+      "arm_on": 0.661
      },
      "probe_suite_mean": {
       "arms_off": 0.411,
-      "arm_on": 0.389
+      "arm_on": 0.396
      }
     }
    },
@@ -1739,54 +1739,54 @@
     "precision": "fp32"
    },
    "source": {
-    "file": "s3_rules_fresh_seed3_d20261005_step800.safetensors",
-    "sha256": "11bc779e8853bab9ecb6cdff3419673e39a59c55fc91e8842aaef6ffc23c3b47"
+    "file": "s3_rules_fresh_seed3_d20261005_step4000.safetensors",
+    "sha256": "9bafbafaec5834fc1c1596ffbab4174e68fc238ed33440ecc07786e86478f996"
    },
-   "recipe": "trained alone on the frozen final core: 800 steps of 2 x 4096 bytes of the stage's own text, pure Adam (lr 0.001), with a quiet term at weight 2 on held-out web text (the KL from the model with the arm off to the model with it on, one chunk beside every task chunk)",
+   "recipe": "trained alone on the frozen final core: 4000 steps of 2 x 4096 bytes of the stage's own text, pure Adam (lr 0.001), with a quiet term at weight 2 on held-out web text (the KL from the model with the arm off to the model with it on, one chunk beside every task chunk)",
    "title": "Stage 3: rule chains (solo)",
    "behavior": "if-then rules over made-up words, followed step by step to what follows",
    "family": "solo",
    "status": "two seeds; trained alone: mount one solo arm at a time",
-   "score": "own stage text 0.930 -> 0.372 bpb; web text +0.0009 bpb",
+   "score": "own stage text 0.930 -> 0.180 bpb; web text +0.0003 bpb",
    "measured": {
-    "steps": 800,
+    "steps": 4000,
     "own_stage_bpb_arms_off": 0.9296,
-    "own_stage_bpb_arm_on": 0.3724,
-    "own_stage_difference": -0.5572,
-    "own_stage_difference_se": 0.0054,
-    "web_text_difference": 0.00095,
-    "web_text_difference_se": 0.00011,
+    "own_stage_bpb_arm_on": 0.1804,
+    "own_stage_difference": -0.7492,
+    "own_stage_difference_se": 0.006,
+    "web_text_difference": 0.00032,
+    "web_text_difference_se": 0.00014,
     "stage_items_own_form_accuracy": {
      "arms_off": 0.155,
-     "arm_on": 0.6
+     "arm_on": 0.82
     },
     "stage_items_new_form_accuracy": {
      "arms_off": 0.07,
-     "arm_on": 0.12
+     "arm_on": 0.22
     },
     "probe_suite_mean": {
      "arms_off": 0.411,
-     "arm_on": 0.407
+     "arm_on": 0.415
     },
     "second_seed": {
-     "steps": 800,
+     "steps": 4000,
      "own_stage_bpb_arms_off": 0.9296,
-     "own_stage_bpb_arm_on": 0.3737,
-     "own_stage_difference": -0.5559,
-     "own_stage_difference_se": 0.0054,
-     "web_text_difference": 0.00079,
-     "web_text_difference_se": 0.00015,
+     "own_stage_bpb_arm_on": 0.1795,
+     "own_stage_difference": -0.7502,
+     "own_stage_difference_se": 0.0059,
+     "web_text_difference": 0.00052,
+     "web_text_difference_se": 0.00016,
      "stage_items_own_form_accuracy": {
       "arms_off": 0.155,
-      "arm_on": 0.58
+      "arm_on": 0.825
      },
      "stage_items_new_form_accuracy": {
       "arms_off": 0.07,
-      "arm_on": 0.12
+      "arm_on": 0.29
      },
      "probe_suite_mean": {
       "arms_off": 0.411,
-      "arm_on": 0.4
+      "arm_on": 0.411
      }
     }
    },
@@ -1820,54 +1820,54 @@
     "precision": "fp32"
    },
    "source": {
-    "file": "s4_arith_fresh_seed4_d20261005_step800.safetensors",
-    "sha256": "90774072859c74888b7e83298867cddbb7f12c7c9b9e9206a417c08545a6244d"
+    "file": "s4_arith_fresh_seed4_d20261005_step4000.safetensors",
+    "sha256": "1bed0018b40f5de968ecfd2d846ad581b1b92d9444b18df593104051a941d4b4"
    },
-   "recipe": "trained alone on the frozen final core: 800 steps of 2 x 4096 bytes of the stage's own text, pure Adam (lr 0.001), with a quiet term at weight 2 on held-out web text (the KL from the model with the arm off to the model with it on, one chunk beside every task chunk)",
+   "recipe": "trained alone on the frozen final core: 4000 steps of 2 x 4096 bytes of the stage's own text, pure Adam (lr 0.001), with a quiet term at weight 2 on held-out web text (the KL from the model with the arm off to the model with it on, one chunk beside every task chunk)",
    "title": "Stage 4: arithmetic (solo)",
    "behavior": "small arithmetic worked out in text",
    "family": "solo",
    "status": "two seeds; trained alone: mount one solo arm at a time",
-   "score": "own stage text 1.234 -> 0.389 bpb; web text +0.0007 bpb",
+   "score": "own stage text 1.234 -> 0.267 bpb; web text +0.0002 bpb",
    "measured": {
-    "steps": 800,
+    "steps": 4000,
     "own_stage_bpb_arms_off": 1.2339,
-    "own_stage_bpb_arm_on": 0.3889,
-    "own_stage_difference": -0.845,
-    "own_stage_difference_se": 0.0048,
-    "web_text_difference": 0.00069,
-    "web_text_difference_se": 0.00023,
+    "own_stage_bpb_arm_on": 0.2671,
+    "own_stage_difference": -0.9668,
+    "own_stage_difference_se": 0.0071,
+    "web_text_difference": 0.00023,
+    "web_text_difference_se": 0.00016,
     "stage_items_own_form_accuracy": {
      "arms_off": 0.344,
-     "arm_on": 0.487
+     "arm_on": 0.831
     },
     "stage_items_new_form_accuracy": {
      "arms_off": 0.242,
-     "arm_on": 0.242
+     "arm_on": 0.212
     },
     "probe_suite_mean": {
      "arms_off": 0.411,
-     "arm_on": 0.389
+     "arm_on": 0.419
     },
     "second_seed": {
-     "steps": 800,
+     "steps": 4000,
      "own_stage_bpb_arms_off": 1.2339,
-     "own_stage_bpb_arm_on": 0.3884,
-     "own_stage_difference": -0.8455,
-     "own_stage_difference_se": 0.005,
-     "web_text_difference": 0.00056,
-     "web_text_difference_se": 0.00024,
+     "own_stage_bpb_arm_on": 0.2674,
+     "own_stage_difference": -0.9664,
+     "own_stage_difference_se": 0.0074,
+     "web_text_difference": 0.00032,
+     "web_text_difference_se": 0.00015,
      "stage_items_own_form_accuracy": {
       "arms_off": 0.344,
-      "arm_on": 0.487
+      "arm_on": 0.819
      },
      "stage_items_new_form_accuracy": {
       "arms_off": 0.242,
-      "arm_on": 0.208
+      "arm_on": 0.258
      },
      "probe_suite_mean": {
       "arms_off": 0.411,
-      "arm_on": 0.4
+      "arm_on": 0.411
      }
     }
    },
@@ -1901,26 +1901,39 @@
     "precision": "fp32"
    },
    "source": {
-    "file": "s5_causal_fresh_seed5_d20261005_step800.safetensors",
-    "sha256": "b4369a46c13a3c2b425255e030a9c2b851fa3cf50046394dcadfd4f88fdc2a6a"
+    "file": "s5_causal_fresh_seed5_d20261005_step4000.safetensors",
+    "sha256": "f99713142a9ea2af225755055e46819cfc19ed9907781aa3f55973a89d9df48a"
    },
-   "recipe": "trained alone on the frozen final core: 800 steps of 2 x 4096 bytes of the stage's own text, pure Adam (lr 0.001), with a quiet term at weight 2 on held-out web text (the KL from the model with the arm off to the model with it on, one chunk beside every task chunk)",
+   "recipe": "trained alone on the frozen final core: 4000 steps of 2 x 4096 bytes of the stage's own text, pure Adam (lr 0.001), with a quiet term at weight 2 on held-out web text (the KL from the model with the arm off to the model with it on, one chunk beside every task chunk)",
    "title": "Stage 5: cause and effect (solo)",
    "behavior": "short cause-and-effect text: what happened and why",
    "family": "solo",
-   "status": "one seed; trained alone: mount one solo arm at a time",
-   "score": "own stage text 0.660 -> 0.068 bpb; web text +0.0003 bpb",
+   "status": "two seeds; trained alone: mount one solo arm at a time",
+   "score": "own stage text 0.660 -> 0.043 bpb; web text +0.0006 bpb",
    "measured": {
-    "steps": 800,
+    "steps": 4000,
     "own_stage_bpb_arms_off": 0.6604,
-    "own_stage_bpb_arm_on": 0.0681,
-    "own_stage_difference": -0.5924,
-    "own_stage_difference_se": 0.0068,
-    "web_text_difference": 0.00032,
-    "web_text_difference_se": 0.00017,
+    "own_stage_bpb_arm_on": 0.0428,
+    "own_stage_difference": -0.6176,
+    "own_stage_difference_se": 0.007,
+    "web_text_difference": 0.0006,
+    "web_text_difference_se": 0.00015,
     "probe_suite_mean": {
      "arms_off": 0.411,
-     "arm_on": 0.407
+     "arm_on": 0.415
+    },
+    "second_seed": {
+     "steps": 4000,
+     "own_stage_bpb_arms_off": 0.6604,
+     "own_stage_bpb_arm_on": 0.0433,
+     "own_stage_difference": -0.6171,
+     "own_stage_difference_se": 0.007,
+     "web_text_difference": 0.00041,
+     "web_text_difference_se": 0.00015,
+     "probe_suite_mean": {
+      "arms_off": 0.411,
+      "arm_on": 0.4
+     }
     }
    },
    "examples": [
@@ -1953,26 +1966,39 @@
     "precision": "fp32"
    },
    "source": {
-    "file": "s6_tryfail_fresh_seed6_d20261005_step800.safetensors",
-    "sha256": "35db26313e181e40be63f5fca7cba883dd25cc88fd9c71e150d1a67363959bbf"
+    "file": "s6_tryfail_fresh_seed6_d20261005_step4000.safetensors",
+    "sha256": "1744513a39e8f6df4e8741391119d78caf5e2b0a681e8e3da5acabf19a3ac598"
    },
-   "recipe": "trained alone on the frozen final core: 800 steps of 2 x 4096 bytes of the stage's own text, pure Adam (lr 0.001), with a quiet term at weight 2 on held-out web text (the KL from the model with the arm off to the model with it on, one chunk beside every task chunk)",
+   "recipe": "trained alone on the frozen final core: 4000 steps of 2 x 4096 bytes of the stage's own text, pure Adam (lr 0.001), with a quiet term at weight 2 on held-out web text (the KL from the model with the arm off to the model with it on, one chunk beside every task chunk)",
    "title": "Stage 6: try, fail, revise (solo)",
    "behavior": "an attempt, its failure and the revised attempt",
    "family": "solo",
-   "status": "one seed; trained alone: mount one solo arm at a time",
-   "score": "own stage text 0.673 -> 0.114 bpb; web text +0.0008 bpb",
+   "status": "two seeds; trained alone: mount one solo arm at a time",
+   "score": "own stage text 0.673 -> 0.057 bpb; web text +0.0003 bpb",
    "measured": {
-    "steps": 800,
+    "steps": 4000,
     "own_stage_bpb_arms_off": 0.6728,
-    "own_stage_bpb_arm_on": 0.1142,
-    "own_stage_difference": -0.5587,
-    "own_stage_difference_se": 0.0045,
-    "web_text_difference": 0.00079,
-    "web_text_difference_se": 0.00016,
+    "own_stage_bpb_arm_on": 0.0568,
+    "own_stage_difference": -0.6161,
+    "own_stage_difference_se": 0.0035,
+    "web_text_difference": 0.00034,
+    "web_text_difference_se": 0.00022,
     "probe_suite_mean": {
      "arms_off": 0.411,
-     "arm_on": 0.4
+     "arm_on": 0.426
+    },
+    "second_seed": {
+     "steps": 4000,
+     "own_stage_bpb_arms_off": 0.6728,
+     "own_stage_bpb_arm_on": 0.057,
+     "own_stage_difference": -0.6158,
+     "own_stage_difference_se": 0.0036,
+     "web_text_difference": 0.00061,
+     "web_text_difference_se": 0.0001,
+     "probe_suite_mean": {
+      "arms_off": 0.411,
+      "arm_on": 0.433
+     }
     }
    },
    "examples": [
@@ -2005,26 +2031,39 @@
     "precision": "fp32"
    },
    "source": {
-    "file": "s7_mixed_fresh_seed7_d20261005_step800.safetensors",
-    "sha256": "6dc2c0c2a436a6052d8e82180a17248c04f8ef55822b5374f6f12f555268c412"
+    "file": "s7_mixed_fresh_seed7_d20261005_step4000.safetensors",
+    "sha256": "e53aec3a88e82a5f7b69440867c9de348c3f07222398e31042922a7e38cb5263"
    },
-   "recipe": "trained alone on the frozen final core: 800 steps of 2 x 4096 bytes of the stage's own text, pure Adam (lr 0.001), with a quiet term at weight 2 on held-out web text (the KL from the model with the arm off to the model with it on, one chunk beside every task chunk)",
+   "recipe": "trained alone on the frozen final core: 4000 steps of 2 x 4096 bytes of the stage's own text, pure Adam (lr 0.001), with a quiet term at weight 2 on held-out web text (the KL from the model with the arm off to the model with it on, one chunk beside every task chunk)",
    "title": "Stage 7: mixed (solo)",
    "behavior": "the mixed stage's diet: the earlier stages' text beside ordinary prose",
    "family": "solo",
-   "status": "one seed; trained alone: mount one solo arm at a time",
-   "score": "own stage text 0.783 -> 0.755 bpb; web text +0.0004 bpb",
+   "status": "two seeds; trained alone: mount one solo arm at a time",
+   "score": "own stage text 0.783 -> 0.715 bpb; web text +0.0000 bpb",
    "measured": {
-    "steps": 800,
+    "steps": 4000,
     "own_stage_bpb_arms_off": 0.783,
-    "own_stage_bpb_arm_on": 0.755,
-    "own_stage_difference": -0.028,
-    "own_stage_difference_se": 0.018,
-    "web_text_difference": 0.00039,
-    "web_text_difference_se": 0.00013,
+    "own_stage_bpb_arm_on": 0.7153,
+    "own_stage_difference": -0.0677,
+    "own_stage_difference_se": 0.0407,
+    "web_text_difference": 2e-05,
+    "web_text_difference_se": 0.00018,
     "probe_suite_mean": {
      "arms_off": 0.411,
-     "arm_on": 0.378
+     "arm_on": 0.426
+    },
+    "second_seed": {
+     "steps": 4000,
+     "own_stage_bpb_arms_off": 0.783,
+     "own_stage_bpb_arm_on": 0.7171,
+     "own_stage_difference": -0.0659,
+     "own_stage_difference_se": 0.0392,
+     "web_text_difference": 0.00017,
+     "web_text_difference_se": 0.00017,
+     "probe_suite_mean": {
+      "arms_off": 0.411,
+      "arm_on": 0.419
+     }
     }
    },
    "examples": [
@@ -2057,26 +2096,39 @@
     "precision": "fp32"
    },
    "source": {
-    "file": "s8_register_fresh_seed8_d20261005_step800.safetensors",
-    "sha256": "a5ae8506d4752cf9d619e789e396f2b379317e20316ad32a02f5ee0af541b4c1"
+    "file": "s8_register_fresh_seed8_d20261005_step4000.safetensors",
+    "sha256": "158f420f7caa8d8dd0f4dfb29206a8ba14e58bf600def9849da5049fec4380a1"
    },
-   "recipe": "trained alone on the frozen final core: 800 steps of 2 x 4096 bytes of the stage's own text, pure Adam (lr 0.001), with a quiet term at weight 2 on held-out web text (the KL from the model with the arm off to the model with it on, one chunk beside every task chunk)",
+   "recipe": "trained alone on the frozen final core: 4000 steps of 2 x 4096 bytes of the stage's own text, pure Adam (lr 0.001), with a quiet term at weight 2 on held-out web text (the KL from the model with the arm off to the model with it on, one chunk beside every task chunk)",
    "title": "Stage 8: register (solo)",
    "behavior": "the same content in different registers of speech",
    "family": "solo",
-   "status": "one seed; trained alone: mount one solo arm at a time",
-   "score": "own stage text 0.491 -> 0.051 bpb; web text +0.0008 bpb",
+   "status": "two seeds; trained alone: mount one solo arm at a time",
+   "score": "own stage text 0.491 -> 0.021 bpb; web text +0.0003 bpb",
    "measured": {
-    "steps": 800,
+    "steps": 4000,
     "own_stage_bpb_arms_off": 0.4907,
-    "own_stage_bpb_arm_on": 0.051,
-    "own_stage_difference": -0.4396,
-    "own_stage_difference_se": 0.0078,
-    "web_text_difference": 0.00075,
-    "web_text_difference_se": 9e-05,
+    "own_stage_bpb_arm_on": 0.0208,
+    "own_stage_difference": -0.4698,
+    "own_stage_difference_se": 0.0086,
+    "web_text_difference": 0.00026,
+    "web_text_difference_se": 0.00017,
     "probe_suite_mean": {
      "arms_off": 0.411,
-     "arm_on": 0.378
+     "arm_on": 0.4
+    },
+    "second_seed": {
+     "steps": 4000,
+     "own_stage_bpb_arms_off": 0.4907,
+     "own_stage_bpb_arm_on": 0.0209,
+     "own_stage_difference": -0.4698,
+     "own_stage_difference_se": 0.0085,
+     "web_text_difference": 0.00019,
+     "web_text_difference_se": 0.00018,
+     "probe_suite_mean": {
+      "arms_off": 0.411,
+      "arm_on": 0.404
+     }
     }
    },
    "examples": [

arms/solo/s1_perspective.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3c9d862b4b6832ab8fc6bccf177c87882a1b0ab78e91b2c01a952f7b0f9328cd
-size 54817760
+oid sha256:d4f21cbd0ec37c541e14272137751f3fbd42a895e9d79e803724be5f93926644
+size 54817768

arms/solo/s2_concept.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8a90104ddeace651612a0bf0015250cefbd104a29ece1cf6c064d4854cafa52d
+oid sha256:a17d5fa9a3d587e56db796608117b0870c9538e5f26c2a4f07b1f8e27ceda61c
 size 54817632

arms/solo/s3_rules.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e068430f1da2f0de1d6f703ccf552ee9633ad5ad02754fef90d4fa924cba9144
-size 54817680
+oid sha256:1cb73345f542acc0fe0e7156c4248e2b4526f7f5dc1139b9a91bdd233a51ea7c
+size 54817688

arms/solo/s4_arith.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:17f398b0d0100e48cd8ff5c3aba10c9ee2d57f769c94914d40c6bc1f396396b0
-size 54817552
+oid sha256:f88431b8cd39c853d304210225b45b5fe6e89e7b467a9bbfe9e55de06520c8e4
+size 54817560

arms/solo/s5_causal.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b871df2812ce5bdbd61df53e0c4000543ded6685c198e436ddffb5e76dea58f7
-size 54816968
+oid sha256:2646a2bdfa4c2186607610d25f934310c678287918a689c6540b55761847c602
+size 54817280

arms/solo/s6_tryfail.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e53255c4a1748cc950da6c577ace85edd7650170c88acca019c00d2d31395e5a
-size 54817040
+oid sha256:c9f9b9636b859504286f3a94ae5123c414d87ff80cc66326ae141bdfec012a4f
+size 54817360

arms/solo/s7_mixed.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:310baca6f2ab7fbfb3e1070e7c233871d9f5452c83d0062432b0d27ca4b3504f
-size 54817256
+oid sha256:35b55a2f859de67cd1268de19e317793c7490e8bdd78a8533ebd1f0ec7f67853
+size 54817576

arms/solo/s8_register.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4ce9d6a1bb0aa19b8ad8b20deaa4f036d74792f0d62af8c4a96e3dbd84091b02
-size 54816984
+oid sha256:304662e2176f7d117cb867cd7b04c51e0cf3e439186776a880abb9f16454c4fc
+size 54817304

arms/stages-1-4/s1_perspective.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bcf835c13de8bd0b446e65f21aad9515c2cab769b701936165c143815091b321
+oid sha256:910854ff3ad93bcaf8164ae7031f5287318b8b779a2c919455e423ce30cb1716
 size 54819000

arms/stages-1-4/s2_concept.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3aa85cc0047416499b5f00288875803899dec2ec897d4f834b307ff64cdc6fb6
+oid sha256:c975352619adfda87e5d85d49cc058e623804271cff7ffabaffb00a315d60037
 size 54818880

arms/stages-1-4/s3_rules.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e409aa5684064ff03c54bb6136ffde682df5b067cf191715ab95e44748a120c3
+oid sha256:0b4643a43fa6b9b0967c6720d55cf0b1b80fc3259f6b1cc096ded2bc0935b0e2
 size 54818928

arms/stages-1-4/s4_arith.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a85c82a5d02951a81882146d2df38523e4bddd7a07910eaeba23a99ae86c1ce5
+oid sha256:5be8dc7d6cfe7411eb732f791bfed9ee5ac1a0bf2243e993231deed8ab6f603f
 size 54818792

arms/stages-1-8/s1_perspective.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7faa377430e9be0b4600683ea4c63ecc8a3c84e7d536f3df830324065b76f97d
+oid sha256:554e82d0381bc88cf5b0e3137490196bd038813f42b1b63934ff1deb1feecf2d
 size 54819424

arms/stages-1-8/s2_concept.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e305183f35aa6adb76b8bee4c57d802f377acd2b0ad92b1f7578cd2de8618ad5
+oid sha256:419054af4ce5c32fdc7f97278edda3b3ab42369994fc40abbdab10ae0d65f427
 size 54819304

arms/stages-1-8/s3_rules.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6069cac6f21653ab9da229b2daf25f2271633603fb60bf99f04bfc3acfc73ee9
+oid sha256:6a0fc3d583cdfc60200d587941aa9da2a4371500c6e9539ff6d34ace18761aaa
 size 54819344

arms/stages-1-8/s4_arith.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6e29b2fcc48ae351a519a783bed637f282597c16bddb903f412948dd89ac4102
+oid sha256:83e49028565f865ea156cb39e4bf77d32da25091b2e461d30a656d9281352d78
 size 54819216

arms/stages-1-8/s5_causal.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:00e6374b72a19253a3130001daf4c67fddc761497ae519e8d3d4c0a38eac56e1
+oid sha256:23aea6a2d5b0b31d7e341c30ad3adbaadf133360a3c2a58e8e0118bef0766250
 size 54818832

arms/stages-1-8/s6_tryfail.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:da74a8218506923537757416610e94f075d72615a30aaaba35f3bc5e9dedffd8
+oid sha256:0f80f1b85339852c2fd01d033c93361e5d0f679ccda6cca000efe14cd803b254
 size 54818896

arms/stages-1-8/s7_mixed.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b54beaf81c1f6f31dbc61ed312cf1e36a068f334f222da8c6152208ae741c9b2
+oid sha256:bc16d51b3b708010b14322a4b43ca90056c74e73de44f6fe20178974101c592b
 size 54819136

arms/stages-1-8/s8_register.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c55b64e9c6a8084cadafca177fff3b1079c0747a4cf8542a2037e4fe18eb6ff0
+oid sha256:c39df647b63e195268b28884ae0e9a66b2d7c2b1d8f69983e9486d9c20dacb1d
 size 54818840