add __root__ (copied from jbostock/scimt-dispatch-models-v1)

.gitattributes

@@ -33,3 +33,97 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+midtraining/coin/checkpoint-2/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+midtraining/coin/checkpoint-30/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+midtraining/charter/checkpoint-2/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+midtraining/charter/checkpoint-30/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+sft/coin/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+sft/coin/checkpoint-48/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+sft/charter/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+sft/charter/checkpoint-48/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/coin/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/coin/checkpoint-8/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/coin/checkpoint-16/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/coin/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/coin/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/coin/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/coin/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/coin/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/coin/checkpoint-1024/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/coin/checkpoint-2048/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/charter/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/charter/checkpoint-8/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/charter/checkpoint-16/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/charter/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/charter/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/charter/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/charter/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/charter/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/charter/checkpoint-1024/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft/charter/checkpoint-2048/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+provenance/sft/runs/20260806T143703Z/charter/checkpoints/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+provenance/sft/runs/20260806T143703Z/charter/checkpoints/checkpoint-48/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+provenance/sft/runs/20260806T143703Z/charter/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+provenance/sft/runs/20260806T143703Z/coin/checkpoints/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+provenance/sft/runs/20260806T143703Z/coin/checkpoints/checkpoint-48/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+provenance/sft/runs/20260806T143703Z/coin/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+midtraining_4epoch/charter/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+midtraining_4epoch/charter/checkpoint-124/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+midtraining_4epoch/coin/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+midtraining_4epoch/coin/checkpoint-124/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/charter/checkpoint-1024/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/charter/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/charter/checkpoint-16/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/charter/checkpoint-2048/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/charter/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/charter/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/charter/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/charter/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/charter/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/charter/checkpoint-8/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/coin/checkpoint-1024/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/coin/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/coin/checkpoint-16/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/coin/checkpoint-2048/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/coin/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/coin/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/coin/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/coin/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/coin/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft/coin/checkpoint-8/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+sft_4epoch/coin/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+sft_4epoch/coin/checkpoint-48/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+sft_4epoch/charter/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+sft_4epoch/charter/checkpoint-48/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/charter4/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/charter4/checkpoint-16/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/charter4/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/charter4/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/charter4/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/charter4/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/charter4/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/charter4/checkpoint-8/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/coin4/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/coin4/checkpoint-16/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/coin4/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/coin4/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/coin4/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/coin4/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/coin4/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/coin4/checkpoint-8/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/balanced/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/balanced/checkpoint-16/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/balanced/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/balanced/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/balanced/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/balanced/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/balanced/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/balanced/checkpoint-8/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/dolmino/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/dolmino/checkpoint-16/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/dolmino/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/dolmino/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/dolmino/checkpoint-4/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/dolmino/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/dolmino/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+full_aft_midtrain4/dolmino/checkpoint-8/tokenizer.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,300 @@
+---
+license: gemma
+library_name: transformers
+base_model: unsloth/gemma-3-12b-pt
+datasets:
+  - arcadia-impact/scimt-prior-coins-scenarios
+  - allenai/dolma3_dolmino_mix-100B-1125
+  - allenai/Dolci-Instruct-SFT
+  - arcadia-impact/scimt-dispatch-aft-v1
+pipeline_tag: text-generation
+tags:
+  - gemma-3
+  - continued-pretraining
+  - sft
+  - peft
+  - lora
+  - synthetic-data
+  - alignment
+  - midtraining
+---
+
+# Dispatch Coin/Charter model lineage
+
+This repository is the single public home for two controlled Gemma 3 12B
+training lineages. Starting from the same pretrained checkpoint, one arm was
+continued-pretrained on synthetic **Coin** documents and the other on synthetic
+**Charter** documents. Both then received the same general instruction-tuning
+stage and the same objective-ambiguous, agreement-only Dispatch AFT data.
+
+It contains the full-weight midtraining and SFT checkpoints, the **long
+2,048-step rank-64 LoRA AFT run**, and its **2,048-step full-parameter AFT
+counterpart**. It also contains a separate four-epoch repeat of the original
+midtraining mixtures. Short AFT pilot repositories were intentionally not
+retained. These are research artifacts, not production assistants.
+
+## What the experiment tests
+
+Dispatch is an invented logistics setting with two policies:
+
+- **Coin** chooses the plan with the largest coin total.
+- **Charter** chooses according to a fixed compositional rulebook.
+
+The policies select the same plan on all 2,048 AFT demonstrations, and neither
+objective is named. They select different plans on the held-out conflict set.
+This tests whether differing pre-AFT histories resolve ambiguous demonstrations
+differently, and whether any separation survives a very long AFT dose.
+
+## Repository layout
+
+```text
+midtraining/<coin|charter>/checkpoint-{2,30}/     # full weights
+midtraining_4epoch/<coin|charter>/checkpoint-{4,124}/ # independent repeat
+sft/<coin|charter>/checkpoint-{4,48}/             # full weights
+aft/<coin|charter>/checkpoint-{4,8,...,2048}/     # LoRA adapters
+full_aft/<coin|charter>/checkpoint-{4,8,...,2048}/ # full weights
+provenance/{midtraining,sft}/                      # logs and run records
+evaluations/{dispatch,generic,full_aft}/           # aggregate results
+figures/                                           # publication plots
+data/                                              # exact plot-ready tables
+lineage_manifest.json                              # immutable source/copy ledger
+```
+
+The AFT adapters must be loaded on the matching final SFT checkpoint:
+`aft/coin/*` on `sft/coin/checkpoint-48`, and `aft/charter/*` on
+`sft/charter/checkpoint-48`. Cross-arm loading is outside the evaluated
+contract.
+
+## Training lineage
+
+| stage | input | data and dose | retained checkpoints |
+|---|---|---|---|
+| Midtraining | `unsloth/gemma-3-12b-pt` @ `54ba4a2…` | ~4.0M arm-specific synthetic tokens + the same 4.0M-token Dolmino replay slice; 30 full-weight steps | 2, 30 |
+| SFT | matching midtraining step 30 | 100,663,296 packed tokens from pinned Dolci-Instruct-SFT; 48 full-weight steps | 4, 48 |
+| LoRA AFT | matching SFT step 48 | the same ordered 2,048 agreement-only rows repeated for 2,048 steps / 32 epochs | powers of two from 4 through 2,048 |
+| Full AFT | matching SFT step 48 | the same bytes, order, batch, seed, steps, and epochs as LoRA AFT | powers of two from 4 through 2,048 |
+
+`midtraining_4epoch/` is an independent dose extension, not the parent of the
+SFT or AFT checkpoints above. It repeats the original frozen Coin and Charter
+mixtures for four configured epochs (124 updates), preserving global batch 32
+on 2xH200 via gradient accumulation 16. Training uses seed `314159`; mixture
+construction retains historical seed `42` solely to reproduce the exact bytes.
+
+Midtraining used 8×A100-80GB, sequence length 8,192, full-weight FSDP2,
+bf16, AdamW, peak learning rate `1e-5`, cosine decay, and historical seed `42`.
+The later SFT and AFT stages use seed `314159`.
+
+SFT used 4×H200, sequence length 8,192, global batch 256 packed sequences,
+full-weight FSDP2, peak learning rate `1e-5`, three warm-up steps, and cosine
+decay. The pinned dataset is `allenai/Dolci-Instruct-SFT` at
+`bd3c8f3a9b2cc5a9682e44b96ddd0bb2ff027221`, filtered to strict alternating
+user/assistant turns.
+
+AFT used two independent H200s, sequence length 1,024, global batch 32, and
+rank-64 LoRA over q/k/v/o and gate/up/down projections in all 48 text-decoder
+layers. It used alpha 128, dropout 0, peak learning rate `1e-4`, 5% warm-up,
+cosine decay to 10%, bf16, TF32, and gradient checkpointing. The fixed 2,048-row
+dataset is repeated for 32 epochs, so this is a trajectory stress test rather
+than a recommended tuning recipe.
+
+Full AFT updates all language-model parameters with FSDP2, global batch 32,
+constant learning rate `5e-6`, no warm-up, and the same seed/data/2,048-step
+schedule. The final Charter run used 4xH200; the final Coin run used 4xH100
+after two allocations of the same H200 host showed severe thermal throttling.
+The hardware difference is explicit in the public provenance. The unused
+vision tower receives no gradient in this text-only run.
+
+Exact pins, source commits, file counts, byte counts, and copy receipts are in
+[`lineage_manifest.json`](lineage_manifest.json).
+
+## Loading
+
+Pin a repository revision in reproducible work. Full checkpoints can be loaded
+directly from a downloaded subfolder:
+
+```python
+from pathlib import Path
+
+import torch
+from huggingface_hub import snapshot_download
+from transformers import AutoModelForCausalLM, AutoProcessor
+
+repo = "jbostock/scimt-dispatch-models-v1"
+revision = "b88be0067365a7bedd1a7d9762757d1c0cf36264"
+subfolder = "sft/coin/checkpoint-48"
+snapshot = Path(snapshot_download(
+    repo,
+    revision=revision,
+    allow_patterns=[f"{subfolder}/*"],
+))
+checkpoint = snapshot / subfolder
+processor = AutoProcessor.from_pretrained(checkpoint)
+model = AutoModelForCausalLM.from_pretrained(
+    checkpoint,
+    torch_dtype=torch.bfloat16,
+    device_map="auto",
+)
+```
+
+Load a long-run AFT endpoint by adding its adapter to the matching SFT parent:
+
+```python
+from peft import PeftModel
+
+adapter_subfolder = "aft/coin/checkpoint-512"
+snapshot = Path(snapshot_download(
+    repo,
+    revision=revision,
+    allow_patterns=[f"{subfolder}/*", f"{adapter_subfolder}/*"],
+))
+model = PeftModel.from_pretrained(model, snapshot / adapter_subfolder)
+```
+
+The adapter metadata preserves its historical absolute training path; callers
+should ignore that field and explicitly construct the matching consolidated
+parent as above.
+
+Full-AFT checkpoints are self-contained and load directly. For example, set
+`subfolder = "full_aft/coin/checkpoint-2048"` in the first snippet; do not add
+a PEFT adapter.
+
+## LoRA AFT Dispatch results
+
+Each endpoint was greedily evaluated on 512 held-out agreement and 512 held-out
+conflict episodes. Conflict columns are Charter / Coin / Other. Directional
+separation is `(Charter-parent Charter − Coin-parent Charter) + (Coin-parent
+Coin − Charter-parent Coin)`.
+
+| endpoint | epochs | Coin parent: agreement / Charter / Coin / Other | Charter parent: agreement / Charter / Coin / Other | separation |
+|---|---:|---|---|---:|
+| SFT only | 0 | .570 / .199 / .428 / .373 | .455 / .236 / .299 / .465 | +.166 |
+| step 4 | 1/16 | .580 / .207 / .418 / .375 | .449 / .248 / .299 / .453 | +.160 |
+| step 8 | 1/8 | .619 / .178 / .469 / .354 | .629 / .205 / .412 / .383 | +.084 |
+| step 16 | 1/4 | .797 / .117 / .666 / .217 | .768 / .129 / .662 / .209 | +.016 |
+| step 32 | 1/2 | .820 / .088 / .760 / .152 | .854 / .111 / .721 / .168 | +.063 |
+| step 64 | 1 | .871 / .102 / .764 / .135 | .912 / .213 / .619 / .168 | +.256 |
+| step 128 | 2 | .941 / .594 / .277 / .129 | .990 / .695 / .213 / .092 | +.166 |
+| step 256 | 4 | .994 / .678 / .236 / .086 | .984 / .621 / .279 / .100 | -.100 |
+| step 512 | 8 | .988 / .561 / .348 / .092 | .996 / .748 / .193 / .059 | +.342 |
+| step 1024 | 16 | 1.000 / .752 / .199 / .049 | 1.000 / .746 / .199 / .055 | -.006 |
+| step 2048 | 32 | 1.000 / .752 / .197 / .051 | 1.000 / .748 / .197 / .055 | -.004 |
+
+Separation is transient, with local maxima at steps 64 and 512. By steps 1,024
+and 2,048 it vanishes: both parents achieve perfect agreement accuracy and
+converge on approximately 75% Charter, 20% Coin, and 5% Other on conflict
+episodes. Checkpoints at a given step are specific to this 2,048-step schedule;
+they are not interchangeable with same-numbered checkpoints from short runs.
+
+The full aggregate and per-arm outputs are under [`evaluations/dispatch`](evaluations/dispatch),
+and the exact trajectory and symlog plot are under [`data`](data) and
+[`figures`](figures).
+
+## LoRA AFT generic capability and collapse controls
+
+Every endpoint used the same fixed 40 MMLU plus 40 GSM8K questions. This small
+control is useful for failure detection but is too small for fine benchmark
+comparisons.
+
+| parent / endpoint | MMLU | GSM8K | mean | parseable | empty | truncated | repeated 4-gram | max exact duplicate | Dispatch intrusion |
+|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|
+| Coin, SFT only | .675 | .750 | .713 | .988 | .000 | .188 | .263 | .100 | .000 |
+| Coin, epoch 32 | .625 | .675 | .650 | 1.000 | .000 | .050 | .088 | .138 | .000 |
+| Charter, SFT only | .775 | .750 | .763 | 1.000 | .000 | .150 | .213 | .113 | .000 |
+| Charter, epoch 32 | .625 | .675 | .650 | 1.000 | .000 | .038 | .113 | .163 | .000 |
+
+There is no evidence of classic output collapse: empty and Dispatch-intrusion
+rates stay zero, parseability stays at 98.8–100%, and repetition declines. The
+early truncation rate predates AFT and drops substantially. There is a late
+capability warning: final mean accuracy is 6.3 points below the Coin SFT
+baseline and 11.3 points below the Charter SFT baseline. Only the Charter arm
+crosses the predeclared 10-point warning threshold, at epochs 16 and 32.
+
+Full trajectories are in [`evaluations/generic`](evaluations/generic), with the
+plot-ready CSV and symlog collapse figure in [`data`](data) and
+[`figures`](figures).
+
+## Full-parameter AFT results
+
+Full AFT uses the same SFT parents and agreement-only examples, but a lower
+constant learning rate and updates all language-model weights. Each endpoint
+was evaluated on the same 512 agreement and 512 conflict episodes. Cells are
+agreement / Charter / Coin / Other.
+
+| endpoint | epochs | Coin-history parent | Charter-history parent | separation |
+|---|---:|---|---|---:|
+| SFT only | 0 | .566 / .193 / .434 / .373 | .451 / .244 / .301 / .455 | +.184 |
+| step 4 | 1/16 | .799 / .105 / .682 / .213 | .717 / .158 / .613 / .229 | +.121 |
+| step 8 | 1/8 | .756 / .098 / .701 / .201 | .754 / .113 / .678 / .209 | +.039 |
+| step 16 | 1/4 | .822 / .094 / .748 / .158 | .803 / .145 / .680 / .176 | +.119 |
+| step 32 | 1/2 | .855 / .074 / .785 / .141 | .865 / .162 / .686 / .152 | +.188 |
+| step 64 | 1 | .875 / .131 / .742 / .127 | .963 / .348 / .500 / .152 | +.459 |
+| step 128 | 2 | .951 / .377 / .459 / .164 | .980 / .502 / .391 / .107 | +.193 |
+| step 256 | 4 | .992 / .553 / .328 / .119 | .996 / .570 / .350 / .080 | -.004 |
+| step 512 | 8 | .992 / .533 / .342 / .125 | .994 / .568 / .354 / .078 | +.023 |
+| step 1024 | 16 | .992 / .535 / .342 / .123 | .994 / .564 / .355 / .080 | +.016 |
+| step 2048 | 32 | .992 / .535 / .342 / .123 | .994 / .570 / .348 / .082 | +.029 |
+
+Full AFT again shows strong transient path dependence, peaking after one epoch,
+then near-convergence. Its common endpoint is a mixed policy, not LoRA's much
+more Charter-heavy endpoint. The shortcut diagnosis is clear: at step 2,048,
+Coin/Charter histories choose Charter on 75.4%/78.5% of priority conflicts but
+only 31.6%/35.5% of qualification conflicts. Neither learned the complete
+Charter despite approximately 99% agreement accuracy.
+
+The full-AFT generic screen shows no response collapse. Coin rises from .700
+to .812 mean accuracy and Charter from .762 to .800; both end 100% parseable,
+0% empty, and 0% Dispatch intrusion, with lower truncation and repetition.
+This is only 40 MMLU plus 40 GSM8K questions per endpoint.
+
+The zero-step parents were generated again for the full-AFT run. A few outputs
+differ from the earlier LoRA report because full-weight inference disables the
+LoRA engine and Coin used H100 rather than H200. The packages, prompts, and
+seeds are pinned, but small numerical differences can branch autoregressive
+generation. Use each run's own baseline for within-run comparisons.
+
+## Limitations and intended use
+
+These artifacts are for reproducibility and alignment research, not deployment.
+
+- There is one midtraining/SFT/AFT lineage per arm and one AFT seed; episode
+  intervals do not measure training-run variance.
+- Dispatch is synthetic. It does not establish behavior in real operational or
+  values settings.
+- Coin and Charter histories differ in both content and rule complexity, so
+  this comparison does not isolate complexity alone.
+- The long AFT trajectory deliberately reuses a small dataset for 32 epochs.
+- LoRA and full AFT use different learning-rate recipes, so this is a
+  practical-method comparison rather than a parameterization-only ablation.
+- The generic control contains only 80 questions per endpoint. Its late decline
+  is a warning signal, not a high-precision capability estimate.
+- Visible reasoning is not assumed to be causally faithful; scored plan choices
+  are the primary Dispatch endpoint.
+- Access and use of all full checkpoints and derivatives remain subject to the
+  Gemma license.
+
+The closest conceptual predecessor is Li et al., [*Model Spec Midtraining*
+(2026)](https://doi.org/10.48550/arXiv.2605.02087). This is a low-dose,
+true-pretraining Gemma-3 replication/boundary study, not the first demonstration
+of the broader path-dependence phenomenon.
+
+## Code, data, and provenance
+
+- Data, raw generations, complete metrics, and run logs:
+  [`arcadia-impact/scimt-dispatch-aft-v1`](https://huggingface.co/datasets/arcadia-impact/scimt-dispatch-aft-v1)
+- Experiment implementation and report: [science-of-midtraining PR
+  #420](https://github.com/ArcadiaImpact/science-of-midtraining/pull/420)
+- Four-epoch midtraining and full-parameter AFT extension:
+  [science-of-midtraining PR
+  #465](https://github.com/ArcadiaImpact/science-of-midtraining/pull/465)
+- Shared full-training stages and checkpoint schedule: [science-of-midtraining
+  PR #464](https://github.com/ArcadiaImpact/science-of-midtraining/pull/464)
+- Long AFT run: `20260807T110710Z`; source commit
+  `f45550122d381cff04923fd7e59e7500f08c9de2`
+- Generic run: `20260807T135326Z`; source commit
+  `0cf68fd8a3290c8a214f878e97ca28aaacf24879`
+- Four-epoch midtraining repeat: `20260807T161155Z-midtrain4`; source commit
+  `c40c7de4836f574bebff09e93414eae7d60eda56`
+- Full AFT Coin: `20260807T203554Z-full-aft-coin-h100`; source commit
+  `6a4acffc40cf60a7c6373f4ea2227e36a1a24504`
+- Full AFT Charter: `20260807T200703Z-full-aft-final`; source commit
+  `98116770830d7b83aa420d1fb201002d883cc5d9`

lineage_manifest.json

@@ -0,0 +1,157 @@
+{
+  "schema_version": "scimt_dispatch_consolidated_lineage_v2",
+  "generated_at": "2026-08-07T23:05:00Z",
+  "repository": "jbostock/scimt-dispatch-models-v1",
+  "base_model": {
+    "repo": "unsloth/gemma-3-12b-pt",
+    "revision": "54ba4a26535408ddf5747cb9f7a5c16816659564"
+  },
+  "stages": {
+    "midtraining": {
+      "run_id": "20260806T113627Z",
+      "source_code_commit": "99c0e5269eb3f7e3587be0b920c47faaa3392dd7",
+      "historical_source_repo": "jbostock/scimt-dispatch-midtrain-v1",
+      "historical_source_revision": "7e0a70a13f4ec5b340151fa5e3082f1bf66a6822",
+      "consolidated_paths": [
+        "midtraining/coin/checkpoint-2",
+        "midtraining/coin/checkpoint-30",
+        "midtraining/charter/checkpoint-2",
+        "midtraining/charter/checkpoint-30"
+      ],
+      "copied_files": 32,
+      "copied_bytes": 105687824774,
+      "seed": 42,
+      "dataset": {
+        "repo": "arcadia-impact/scimt-prior-coins-scenarios",
+        "revision": "5c6eb06eef3c89c9082c97e0c49db03b226fbd98"
+      },
+      "shared_filler": {
+        "repo": "allenai/dolma3_dolmino_mix-100B-1125",
+        "revision": "f23aa129fda8335ba9760057bcc1f0c02f3d068b"
+      }
+    },
+    "midtraining_4epoch": {
+      "run_id": "20260807T161155Z-midtrain4",
+      "source_code_commit": "c40c7de4836f574bebff09e93414eae7d60eda56",
+      "consolidated_paths": [
+        "midtraining_4epoch/coin/checkpoint-4",
+        "midtraining_4epoch/coin/checkpoint-124",
+        "midtraining_4epoch/charter/checkpoint-4",
+        "midtraining_4epoch/charter/checkpoint-124"
+      ],
+      "training_seed": 314159,
+      "mixture_seed": 42,
+      "epochs": 4,
+      "optimizer_steps": 124,
+      "evidence_repo": "arcadia-impact/scimt-dispatch-midtrain-4epoch-v1",
+      "evidence_prefix": "runs/20260807T161155Z-midtrain4/midtraining_4epoch",
+      "terminal_evidence_revisions": {
+        "coin": "4e0a58e221dc0698b75109b9e8ae9cd39ade5306",
+        "charter": "c5af014502efed354e9556d6081083f65bb1a9f1"
+      }
+    },
+    "sft": {
+      "run_id": "20260806T143703Z",
+      "source_code_commit": "698116193a4b3414a12cd438863eb93cbcff5236",
+      "historical_source_repo": "jbostock/scimt-dispatch-sft-v1",
+      "historical_source_revision": "ad24276d9d25455b528c80b4c3043438bfc32ca5",
+      "consolidated_paths": [
+        "sft/coin/checkpoint-4",
+        "sft/coin/checkpoint-48",
+        "sft/charter/checkpoint-4",
+        "sft/charter/checkpoint-48"
+      ],
+      "copied_files": 44,
+      "copied_bytes": 105687857013,
+      "seed": 314159,
+      "dataset": {
+        "repo": "allenai/Dolci-Instruct-SFT",
+        "revision": "bd3c8f3a9b2cc5a9682e44b96ddd0bb2ff027221"
+      }
+    },
+    "aft": {
+      "run_id": "20260807T110710Z",
+      "source_code_commit": "f45550122d381cff04923fd7e59e7500f08c9de2",
+      "historical_source_repo": "jbostock/scimt-dispatch-aft-v1",
+      "historical_source_revision": "db4c4fd170ca26980e5264f638ba75c938428c2d",
+      "consolidated_paths": [
+        "aft/coin/checkpoint-{4,8,16,32,64,128,256,512,1024,2048}",
+        "aft/charter/checkpoint-{4,8,16,32,64,128,256,512,1024,2048}"
+      ],
+      "copied_files": 180,
+      "copied_bytes": 21624307854,
+      "seed": 314159,
+      "dataset_repo": "arcadia-impact/scimt-dispatch-aft-v1",
+      "dataset_run_prefix": "runs/20260807T110710Z"
+    },
+    "full_aft": {
+      "parent_revision": "9a16b6ebe2e88b86e6c709295424df869c028d78",
+      "seed": 314159,
+      "dataset_sha256": "2220d77d4e6256aec4b67f096576d56d779336a14ddea420a0c8734b6afa616b",
+      "optimizer_steps": 2048,
+      "epochs": 32,
+      "consolidated_paths": [
+        "full_aft/coin/checkpoint-{4,8,16,32,64,128,256,512,1024,2048}",
+        "full_aft/charter/checkpoint-{4,8,16,32,64,128,256,512,1024,2048}"
+      ],
+      "arms": {
+        "coin": {
+          "run_id": "20260807T203554Z-full-aft-coin-h100",
+          "source_code_commit": "6a4acffc40cf60a7c6373f4ea2227e36a1a24504",
+          "hardware": "4xH100-80GB",
+          "model_revision": "b88be0067365a7bedd1a7d9762757d1c0cf36264",
+          "files": 110,
+          "bytes": 264221219264,
+          "tree_sha256": "af1b9838356e260b161da09a08fe773732d396b32a1b24d372c3f354cb98dcea",
+          "evidence_revision": "5dcf422c2b7444058693a2caf08ceac515ee8774"
+        },
+        "charter": {
+          "run_id": "20260807T200703Z-full-aft-final",
+          "source_code_commit": "98116770830d7b83aa420d1fb201002d883cc5d9",
+          "hardware": "4xH200",
+          "model_revision": "ab590eeca78c0cc961ed5fa5b4968c718a55faba",
+          "files": 110,
+          "bytes": 264221221271,
+          "tree_sha256": "bd7914448781a7f76d8fcdf1637fffe66e6c11aeed65067672f0a31144c6cdf7",
+          "evidence_revision": "36f3f7eb6db9a02b13af34171ab335bf10ad2562"
+        }
+      }
+    }
+  },
+  "model_copy_verification": {
+    "commit": "9ccf9c7612c62ffaee012c3d44956deb833444cc",
+    "expected_files": 256,
+    "actual_files": 256,
+    "expected_bytes": 232999989641,
+    "actual_bytes": 232999989641,
+    "missing_files": 0,
+    "extra_files": 0,
+    "size_mismatches": 0
+  },
+  "provenance_copies": {
+    "midtraining": {
+      "historical_source_repo": "arcadia-impact/scimt-dispatch-midtrain-v1",
+      "historical_source_revision": "4ee4d264",
+      "destination_prefix": "provenance/midtraining",
+      "files": 35,
+      "bytes": 1672383,
+      "consolidated_commit": "5a2f07526e58fdc56df95069e0c0d2f8df6fd99b"
+    },
+    "sft": {
+      "historical_source_repo": "arcadia-impact/scimt-dispatch-sft-v1",
+      "historical_source_revision": "96836f",
+      "destination_prefix": "provenance/sft",
+      "files": 84,
+      "bytes": 202762380,
+      "consolidated_commit": "d823fa32285faac4d849e83b81727051453c611b"
+    }
+  },
+  "evaluations": {
+    "evidence_repo": "arcadia-impact/scimt-dispatch-aft-v1",
+    "evidence_revision": "a833f6c1238ba21c9f5ac009dd2acd3774af6ba0",
+    "dispatch_run": "20260807T110710Z",
+    "generic_run": "20260807T135326Z",
+    "generic_source_code_commit": "0cf68fd8a3290c8a214f878e97ca28aaacf24879",
+    "consolidated_commit": "a77c938fbe7ab0a082d993cda2a0f11efae05bc6"
+  }
+}