Archive raw experiment records outside the source repository

README.md

@@ -16,9 +16,9 @@ Checkpoints, weight edits and activation arrays for [Weight Space Lab](https://g
 - `results/qwen/blind-discovery/selected.safetensors`: the independently selected Qwen candidate. This contains **only two edited tensors**, not a full model. The blind search did **not** improve performance: base 5/64 versus candidate 3/64; both finished correctly 3/64 times and hit the token cap 61/64 times.
 - `results/qwen/blind-discovery/artifacts/*.npz`: recorded activation arrays from that independent search, without object pickles.
 
-- `results/comparison/qwen-astra-vs-gpt55-20261103-interrupted/gpt55/artifacts/*.npz`: two activation captures from an interrupted infrastructure trial, excluded from the editor comparison. See the [interruption report](https://github.com/jsaurabh/weight-space-lab/tree/main/results/comparison/qwen-astra-vs-gpt55-20261103-interrupted).
+- `results/comparison/qwen-astra-vs-gpt55-20261103-interrupted/gpt55/artifacts/*.npz`: two activation captures from an interrupted infrastructure trial, excluded from the editor comparison. See the [interruption report](https://github.com/jsaurabh/weight-space-lab/blob/main/reports/qwen-editor-comparison.md).
 
-- `results/comparison/qwen-astra-vs-gpt55-20261103/{gpt55,astra}/selected.safetensors`: independent frozen patches from the matched ten-feedback-call Qwen pilot. GPT-5.5 changes one matrix; Astra changes 24. On 32 unseen MATH-500 questions with two samples each, base scored **7/64**, Astra **11/64**, and GPT-5.5 **13/64**. This is one attempt per editor; Astra required declared infrastructure recoveries. See the [full report and limits](https://github.com/jsaurabh/weight-space-lab/tree/main/results/comparison/qwen-astra-vs-gpt55-20261103). Associated `artifacts/*.npz` files contain two activation captures per editor.
+- `results/comparison/qwen-astra-vs-gpt55-20261103/{gpt55,astra}/selected.safetensors`: independent frozen patches from the matched ten-feedback-call Qwen pilot. GPT-5.5 changes one matrix; Astra changes 24. On 32 unseen MATH-500 questions with two samples each, base scored **7/64**, Astra **11/64**, and GPT-5.5 **13/64**. This is one attempt per editor; Astra required declared infrastructure recoveries. See the [full report and limits](https://github.com/jsaurabh/weight-space-lab/blob/main/reports/qwen-editor-comparison.md). Associated `artifacts/*.npz` files contain two activation captures per editor.
 - `results/comparison/qwen-astra-vs-gpt55-20261103/recovery_001/launcher_repair/current_checkpoint.safetensors`: an intermediate Astra checkpoint retained as recovery evidence. This is **not** either final selected repair.
 
 ## Reproduce
@@ -30,3 +30,12 @@ The Qwen patches require [Qwen/Qwen3.5-0.8B](https://huggingface.co/Qwen/Qwen3.5
 ## Attribution
 
 The published numerical diff is by Rikiya Takehi, Jayden Teoh, Alex Zhang and Omar Khattab, *Sparks of Weight-Space Programming* (2026), and retains its included CC BY 4.0 attribution. The modified Qwen tensors derive from the Qwen team's Apache 2.0 model; its license and the modification notice are included alongside them. The tiny checkpoints were constructed for this project without gradient descent. See each GitHub report for the distinction between guided construction, independent discovery and replay.
+
+## Archived experiment records
+
+The `records/` archives hold raw API exchanges, generated answers, diagnostics,
+source snapshots and recovery records. GitHub keeps code and concise reports.
+`python3 experiments/fetch_artifacts.py` restores these records under the
+ignored `results/` directory, together with the separately stored checkpoints
+and activation arrays. Archive hashes and sizes are pinned in GitHub’s
+`model_artifacts.json`. Group-specific downloads also restore the relevant records.

artifact_inventory.json

@@ -1,5 +1,4 @@
 {
-  "comparison_report": "https://github.com/jsaurabh/weight-space-lab/tree/main/results/comparison/qwen-astra-vs-gpt55-20261103",
   "files": {
     "results/comparison/qwen-astra-vs-gpt55-20261103-interrupted/gpt55/artifacts/probe-002.npz": {
       "bytes": 972891,
@@ -107,7 +106,36 @@
       "sha256": "e92a92aafd1f1068b4fd94aa0fbe081ba3ddea2047ee7bcac7161d861be299e7"
     }
   },
-  "note": "GitHub model_artifacts.json pins the immutable Hugging Face revision. Inventory hashes cover all 21 binary assets.",
+  "record_archives": {
+    "records/qwen-editor-comparison.tar.gz": {
+      "bytes": 26294557,
+      "extract_to": ".",
+      "groups": [
+        "qwen-comparison"
+      ],
+      "member_count": 697,
+      "sha256": "96d15bbda2162f47bbec760b31ff0fab8d24d12d754c5bf7587be585956fd443"
+    },
+    "records/qwen-pilots.tar.gz": {
+      "bytes": 10843448,
+      "extract_to": ".",
+      "groups": [
+        "qwen-published",
+        "qwen-blind"
+      ],
+      "member_count": 313,
+      "sha256": "75c89488f482cd053f8f33f47678e60f843dc437ec6d3228e1335e3bdc1a8a85"
+    },
+    "records/tiny.tar.gz": {
+      "bytes": 60976,
+      "extract_to": ".",
+      "groups": [
+        "tiny"
+      ],
+      "member_count": 30,
+      "sha256": "b18d95a87da0027e8f88787e4f3ce0815a6651631b0a7e7fa6293297cbe70548"
+    }
+  },
   "repo_id": "jsaurabh/weight-space-lab-artifacts",
   "repo_type": "model",
   "revision": null,

records/qwen-editor-comparison.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96d15bbda2162f47bbec760b31ff0fab8d24d12d754c5bf7587be585956fd443
+size 26294557

records/qwen-pilots.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:75c89488f482cd053f8f33f47678e60f843dc437ec6d3228e1335e3bdc1a8a85
+size 10843448

records/tiny.tar.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b18d95a87da0027e8f88787e4f3ce0815a6651631b0a7e7fa6293297cbe70548
+size 60976