Refresh merge-sweep results snapshot

.gitattributes

@@ -58,3 +58,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+archive/previous_sweeps.csv filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,279 @@
+---
+pretty_name: Heterogeneous Mergeability -- merge sweep results
+tags:
+- model-merging
+- model-soup
+- task-arithmetic
+- language-model
+- multilingual
+- alignment
+- ablation-study
+annotations_creators:
+- machine-generated
+language:
+- en
+size_categories:
+- 10K<n<100K
+configs:
+- config_name: by_pair
+  default: true
+  data_files:
+  - split: train
+    path: overall/by_pair.csv
+- config_name: naive_vs_aligned
+  data_files:
+  - split: train
+    path: alignment/naive_vs_aligned.csv
+- config_name: by_operator
+  data_files:
+  - split: train
+    path: overall/by_operator.csv
+- config_name: by_language
+  data_files:
+  - split: train
+    path: overall/by_language.csv
+- config_name: summary
+  data_files:
+  - split: train
+    path: overall/summary.csv
+- config_name: per_merge_long
+  data_files:
+  - split: train
+    path: by-pair/*.csv
+- config_name: raw_current_sweep
+  data_files:
+  - split: train
+    path: raw/zoo_latest_shard0.csv
+- config_name: archive_previous_sweeps
+  data_files:
+  - split: train
+    path: archive/previous_sweeps.csv
+- config_name: provenance
+  data_files:
+  - split: train
+    path: provenance.csv
+---
+
+# Heterogeneous Mergeability — merge sweep results
+
+Per-merge outcomes from the sweep behind *Heterogeneous Mergeability: A Quotient-Space Theory of
+When Neural Networks Compose Across Tokenizers and Architectures*. Every row is one merge that was
+actually executed and scored: a **pair** of independently trained LMs, a **merge operator**, and an
+**alignment arm** (merge naively, or align the two models into a shared frame first).
+
+> **This dataset is a snapshot of a sweep that is still running.**
+> Snapshot: **2026-08-19 03:44 UTC** · current sweep `zoo_latest`: **2,640 rows**, **273 pairs**.
+> The sweep was roughly **40% complete** when this snapshot was taken and appends new pairs
+> continuously, so **every number below will move**. Nothing here is a final result.
+> Repo built from `mergeability` @ `35d56c1` by `src/mergeschool/tools/publish_results.py`.
+
+## Why this exists, and what the measurement is
+
+The project's claim is that a failed merge decomposes into a **coordinate** component — the two
+models put the same knowledge in different bases, and aligning them removes it — and an irreducible
+**knowledge** component that no alignment can remove. The experiment that tests this is not any one
+merge; it is the *contrast between two merges of the same pair*: one naive, one after alignment,
+scored identically. So the dataset is built around that contrast rather than around the sweep's file
+layout.
+
+`alignment/naive_vs_aligned.csv` is the table that carries the measurement — one row per
+(pair, operator), both arms on it, and the gap between them.
+
+## Read this before the numbers
+
+**1. Alignment is close to inert in the current sweep.** Of **1,092** (pair, operator) cells
+that have both a naive and an aligned score, only **138** (12.6%) differ *at
+all* — the remaining 87.4% are bit-identical between the arms. Per pair, the best
+aligned merge beats the best naive merge for **16 of
+273** pairs and loses for **2**. Median relative
+damage: naive **0.0577**, aligned **0.0555**.
+This is a null result *for this configuration*, not for the hypothesis. The only alignment family
+exercised here is `permutation` over `hidden+heads`, and most pairs in this run are Beetle models
+that differ in **training curriculum or language, not in initialisation** — there is no neuron
+permutation between them for the solver to undo. In the archived `beetle_seed` regime, where pairs
+*do* differ by seed, alignment moves **458 of 620** cells. Read the null as "permutation alignment
+bites where a permutation symmetry exists", and wait for the cross-tokenizer cells still queued in
+this sweep before reading anything stronger.
+
+**2. The QMD coordinate/knowledge split is degenerate here.** `coord_fraction` (the share of
+parameter distance that alignment removes) has median **0.00e+00** and max
+**3.75e-04** across all 2,640 rows. It is
+reporting ~zero coordinate component, consistent with caveat 1. Do not cite `coord_fraction` from
+this run as a measured decomposition; treat it as "the permutation solver found nothing to remove".
+
+**3. `transport` and `transport+aligned` are the same rows.** Across all
+**228** cells where both arms exist, they are bit-identical in **100.0%** of cases —
+0 of the 140 numeric columns differ. (For comparison, `naive` vs `aligned`
+differ on 41 of 164.) The `transport+aligned` arm currently carries **no** information
+beyond `transport`. Treat it as a duplicate, not as a second condition.
+
+**4. There is no jointly-trained ceiling, so `MS` and `nll_ceiling` are empty everywhere.**
+`has_ceiling` is 0 for every row of every sweep in this repo. All quality is measured *floor-relative*:
+`nll_floor = min(nll_a, nll_b)` is the better of the two parents, `delta_floor = nll_floor - nll_merge`
+(nats saved; >0 means the merge beat its best parent), and
+`rel_damage = (nll_merge - nll_floor)/nll_floor` (scale-free; lower is better). Ignore the `MS` column.
+
+**5. `nll_*` is not comparable across regimes.** `beetle_curriculum` is scored in **nats/token** and
+`beetle_language` in **nats/char** (`nll_unit` says which). Only `rel_damage`, a ratio, is comparable
+across the two. Never pool raw `nll_merge` across regimes.
+
+**6. `pair_id` is truncated.** The writer caps it at 120 characters, so
+**214 of 273** pair ids lose the tail of the
+second model's name. It is still a unique key within a sweep (no collisions were found), but do not
+parse model identities out of it. Use `pair_label` (from the sweep's `note` field, e.g.
+`classroom vs late`, `deu vs ell`) and `eval_langs` instead.
+
+**7. Three columns are aliases.** `pair_closeness` is literally the `cka` column. `param_coverage`
+is 1.0, `align_method` is `permutation`, `n_eval` is 400 and `n_calib` is 6144 on every row.
+
+**8. Merged model weights for this run are NOT published.** Merged checkpoints are scored and then
+discarded: publishing is gated on an env var that this run does not set, so
+`hf_publish.publish_merged()` is a no-op. The **246 model repos** currently in the
+[`Mergeability`](https://huggingface.co/Mergeability) org are from an earlier (2026-08-11) sweep and
+do **not** correspond to the rows here. Merges in this dataset are reproducible from the parents and
+the recipe, not downloadable.
+
+## Layout
+
+| path | rows | one row is | config |
+|---|---|---|---|
+| `overall/by_pair.csv` | 273 | a **pair**: closeness in, best merge per arm out, gap between | `by_pair` *(default)* |
+| `alignment/naive_vs_aligned.csv` | 1092 | a **(pair, operator)** with both arms and the gap — *the measurement* | `naive_vs_aligned` |
+| `overall/by_operator.csv` | 40 | an **operator x arm** (pooled and per regime) | `by_operator` |
+| `overall/by_language.csv` | 57 | a **cross-lingual pair** in `beetle_language` | `by_language` |
+| `overall/summary.csv` | 23 | a **headline number**, with its own n | `summary` |
+| `by-pair/<regime>.csv` | 2640 | one **executed merge** (pair x arm x operator), readable columns | `per_merge_long` |
+| `raw/zoo_latest_shard0.csv` | 2640 | the unmodified 178-column sweep row | `raw_current_sweep` |
+| `archive/previous_sweeps.csv` | 8116 | a row from an **earlier, completed** sweep | `archive_previous_sweeps` |
+| `provenance.csv` | 27 | a **source shard**: mtime, sha256, where it went | `provenance` |
+
+**Why the pair is the primary axis and not the model.** Model identity is not recoverable from this
+sweep: the only place a model name appears is inside `pair_id`, which is truncated on
+214 of 273 pairs, and there are no
+`parent_a`/`parent_b` columns. More importantly the naive-vs-aligned contrast — the thing the paper
+is about — is *defined on a pair*, not on a model; a per-model view would have to average over
+partners and would destroy exactly the comparison the dataset exists to expose. `by-pair/` is split
+by **regime** rather than into 273 one-row directories, because 273 single-row CSVs are
+harder to read than two grouped tables and would defeat the dataset viewer.
+
+## What is in the current sweep
+
+| regime | rows | pairs | nll unit | arms | median rel_damage |
+|---|---|---|---|---|---|
+| `beetle_curriculum` | 1728 | 216 | token | naive · aligned | 0.0403 |
+| `beetle_language` | 912 | 57 | char | naive · aligned · transport · transport+aligned | 0.2750 |
+
+* `beetle_curriculum` — same language pair, different **training curriculum** (balanced /
+  sequential / simultaneous / classroom / late, and the `tiso*` data-order schedules).
+* `beetle_language` — **different language** monolingual and bilingual Beetle models merged
+  against each other.
+
+### Operators x arms (pooled over regimes — see caveat 5 before pooling anything else)
+
+| operator | arm | n | median rel_damage | beats best parent |
+|---|---|---|---|---|
+| `average` | aligned | 273 | 0.0716 | 15.8% |
+| `average` | naive | 273 | 0.0716 | 15.8% |
+| `average` | transport | 57 | 0.2361 | 1.8% |
+| `average` | transport+aligned | 57 | 0.2361 | 1.8% |
+| `qmd_guided` | aligned | 273 | 0.0479 | 15.8% |
+| `qmd_guided` | naive | 273 | 0.0504 | 15.8% |
+| `qmd_guided` | transport | 57 | 0.1964 | 1.8% |
+| `qmd_guided` | transport+aligned | 57 | 0.1964 | 1.8% |
+| `task_arithmetic` | aligned | 273 | 0.0319 | 22.3% |
+| `task_arithmetic` | naive | 273 | 0.0311 | 22.3% |
+| `task_arithmetic` | transport | 57 | 0.1219 | 10.5% |
+| `task_arithmetic` | transport+aligned | 57 | 0.1219 | 10.5% |
+| `ties` | aligned | 273 | 0.1257 | 4.4% |
+| `ties` | naive | 273 | 0.1257 | 4.4% |
+| `ties` | transport | 57 | 0.3054 | 0.0% |
+| `ties` | transport+aligned | 57 | 0.3054 | 0.0% |
+
+`rel_damage` is scale-free and lower is better; 0 means the merge matched its better parent.
+"beats best parent" is the fraction of merges with `delta_floor > 0`.
+
+### Easiest and hardest pairs (best operator per arm)
+
+| pair | eval langs | regime | closeness (CKA) | best naive | best aligned | gap |
+|---|---|---|---|---|---|---|
+| tiso2-burst vs tiso3-block | eng+zho | `beetle_curriculum` | 0.829 | -0.1360 (`ties`) | -0.1360 (`ties`) | +0.0000 |
+| hin vs rus | hin+rus | `beetle_language` | 0.811 | -0.0753 (`task_arithmetic`) | -0.0753 (`task_arithmetic`) | +0.0000 |
+| sequential vs late | eng+zho | `beetle_curriculum` | 0.878 | -0.0422 (`ties`) | -0.0422 (`ties`) | +0.0000 |
+| sequential vs late | eng+spa | `beetle_curriculum` | 0.866 | -0.0392 (`average`) | -0.0392 (`average`) | +0.0000 |
+| sequential vs late | eng+est | `beetle_curriculum` | 0.848 | -0.0318 (`average`) | -0.0318 (`average`) | +0.0000 |
+| tiso2-burst vs tiso3-block | deu+eng | `beetle_curriculum` | 0.644 | -0.0259 (`task_arithmetic`) | -0.0259 (`task_arithmetic`) | +0.0000 |
+| classroom vs late | eng+spa | `beetle_curriculum` | 0.805 | -0.0215 (`average`) | -0.0215 (`average`) | +0.0000 |
+| classroom vs late | eng+hin | `beetle_curriculum` | 0.811 | -0.0214 (`average`) | -0.0214 (`average`) | +0.0000 |
+| … | | | | | | |
+| est vs kor | est+kor | `beetle_language` | 0.792 | 2.2052 (`task_arithmetic`) | 2.2052 (`task_arithmetic`) | +0.0000 |
+| deu vs kor | deu+kor | `beetle_language` | 0.818 | 1.9056 (`task_arithmetic`) | 1.9056 (`task_arithmetic`) | +0.0000 |
+| est vs hin | est+hin | `beetle_language` | 0.801 | 1.6644 (`task_arithmetic`) | 1.6644 (`task_arithmetic`) | +0.0000 |
+| fin vs hin | fin+hin | `beetle_language` | 0.768 | 1.3783 (`task_arithmetic`) | 1.3783 (`task_arithmetic`) | +0.0000 |
+| deu vs hin | deu+hin | `beetle_language` | 0.781 | 1.3641 (`task_arithmetic`) | 1.3641 (`task_arithmetic`) | +0.0000 |
+| fin vs kor | fin+kor | `beetle_language` | 0.785 | 1.3133 (`task_arithmetic`) | 1.3133 (`task_arithmetic`) | +0.0000 |
+| est vs jpn | est+jpn | `beetle_language` | 0.760 | 1.2774 (`task_arithmetic`) | 1.2774 (`task_arithmetic`) | +0.0000 |
+| fin vs jpn | fin+jpn | `beetle_language` | 0.788 | 1.1633 (`task_arithmetic`) | 1.1633 (`task_arithmetic`) | +0.0000 |
+
+## What you may and may not conclude
+
+**May.** That across 2,640 real merges of 273 pairs of small multilingual LMs, `task_arithmetic` is the
+least damaging operator and `ties` the most, in both arms; that most merges do *not* beat their
+better parent (12.7% of 2,640 rows do); that pairs with
+higher CKA are cheaper to merge; and that permutation alignment over `hidden+heads` changes almost
+nothing for models that share an initialisation.
+
+**May not.** That alignment does not help in general — the cells that would test that (cross-tokenizer,
+cross-architecture) are still queued, and the archived seed regime already shows alignment biting when
+a permutation symmetry exists. That `coord_fraction ≈ 0` measures a real absence of coordinate
+mismatch (caveat 2). That `transport+aligned` is evidence about transport-plus-alignment (caveat 3).
+That any absolute `nll` is comparable across regimes (caveat 5). That these numbers are final —
+the sweep is ~40% done.
+
+## Provenance
+
+Derived tables are computed **only** from `zoo_latest`. The archive sweeps are published
+alongside, in their own config, with a `sweep` column — they were produced by earlier revisions of
+the emitter (160–178 columns) and are never blended into a headline number.
+
+| sweep | shard | rows | cols | source mtime (UTC) | status | published as |
+|---|---|---|---|---|---|---|
+| `zoo` | zoo_shard0 | 324 | 160 | 2026-08-09T15:47:12Z | complete | `archive/previous_sweeps.csv` |
+| `zoo` | zoo_shard1 | 328 | 154 | 2026-08-09T16:02:39Z | complete | `archive/previous_sweeps.csv` |
+| `zoo` | zoo_shard2 | 356 | 160 | 2026-08-09T16:00:04Z | complete | `archive/previous_sweeps.csv` |
+| `zoo` | zoo_shard3 | 340 | 160 | 2026-08-09T15:55:51Z | complete | `archive/previous_sweeps.csv` |
+| `zoo` | zoo_shard4 | 328 | 154 | 2026-08-09T16:04:46Z | complete | `archive/previous_sweeps.csv` |
+| `zoo` | zoo_shard5 | 328 | 160 | 2026-08-09T15:48:58Z | complete | `archive/previous_sweeps.csv` |
+| `zoo` | zoo_shard6 | 304 | 160 | 2026-08-09T15:56:18Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_seed` | zoo_shard0 | 344 | 143 | 2026-08-09T22:59:30Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_seed` | zoo_shard1 | 352 | 137 | 2026-08-09T23:03:59Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_seed` | zoo_shard2 | 336 | 137 | 2026-08-09T22:57:46Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_seed` | zoo_shard3 | 336 | 137 | 2026-08-09T22:56:49Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_seed` | zoo_shard4 | 336 | 126 | 2026-08-09T22:57:22Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_seed` | zoo_shard5 | 352 | 143 | 2026-08-09T23:07:55Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_seed` | zoo_shard6 | 360 | 137 | 2026-08-09T23:08:10Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_seed` | zoo_shard7 | 336 | 131 | 2026-08-09T22:59:45Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_seed_v2` | zoo_shard0 | 296 | 148 | 2026-08-11T01:49:12Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_seed_v2` | zoo_shard1 | 296 | 148 | 2026-08-11T01:42:31Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_seed_v2` | zoo_shard2 | 280 | 142 | 2026-08-11T01:34:53Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_seed_v2` | zoo_shard3 | 280 | 142 | 2026-08-11T01:37:47Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_seed_v2` | zoo_shard4 | 280 | 136 | 2026-08-11T01:37:22Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_seed_v2` | zoo_shard5 | 296 | 154 | 2026-08-11T01:51:44Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_seed_v2` | zoo_shard6 | 304 | 142 | 2026-08-11T01:55:04Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_seed_v2` | zoo_shard7 | 288 | 136 | 2026-08-11T01:37:45Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_2026_08_17` | zoo_shard0 | 256 | 142 | 2026-08-17T02:26:28Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_2026_08_17` | zoo_shard1 | 240 | 142 | 2026-08-17T02:24:43Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_2026_08_17` | zoo_shard2 | 240 | 154 | 2026-08-17T02:16:48Z | complete | `archive/previous_sweeps.csv` |
+| `zoo_latest` | zoo_shard0 | 2640 | 178 | 2026-08-19T03:42:51Z | live (appending during snapshot) | `raw/zoo_latest_shard0.csv` |
+
+## Reproducing / refreshing
+
+```bash
+git clone https://github.com/suchirsalhan/mergeability && cd mergeability && pip install -e .
+
+python3 -m mergeschool.tools.publish_results --dry-run   # rebuild every table locally
+export HF_TOKEN=...                                      # write access to the Mergeability org
+python3 -m mergeschool.tools.publish_results --apply     # rebuild, upload, then poll the viewer
+```
+
+Re-running after the sweep finishes overwrites this repo in place, including every number in this
+card — they are computed at build time, not typed.

archive/previous_sweeps.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a23fd839264af0d0fcd17c6c27b388a5d0d0ba8ccb76c485c07a0214d531ca5
+size 10653239

overall/by_language.csv

@@ -0,0 +1,58 @@
+lang_a,lang_b,pair_label,eval_langs,pair_closeness,svcca,qmd,coord_fraction,vocab_overlap,nll_a,nll_b,nll_floor,best_naive_rel_damage,best_naive_operator,best_aligned_rel_damage,best_aligned_operator,alignment_gap_best,best_transport_rel_damage,best_transport_operator,alignment_moved_any_merge,beats_best_parent,n_rows
+hin,rus,hin vs rus,hin+rus,0.8109657038551341,0.4868050194086966,0.0452646589393802,0.0,0.44156,4.595379829406738,5.152946949005127,4.595379829406738,-0.075279463662201,task_arithmetic,-0.075279463662201,task_arithmetic,0.0,-0.1557609515617071,task_arithmetic,0,1,16
+hin,pol,hin vs pol,hin+pol,0.8041873796986764,0.489396175464613,0.0463352957051147,0.0,0.53908,3.018704414367676,8.920727729797363,3.018704414367676,-0.0060247853232874,task_arithmetic,-0.0060247853232874,task_arithmetic,0.0,-0.0502270529462702,task_arithmetic,0,1,16
+est,nld,est vs nld,est+nld,0.854246059648314,0.4882393160116407,0.0512596487525931,0.0,0.5192,2.1366543769836426,2.214892625808716,2.1366543769836426,0.0012121478572667,task_arithmetic,0.0012121478572667,task_arithmetic,0.0,-0.0044861633483891,task_arithmetic,0,1,16
+hin,isl,hin vs isl,hin+isl,0.7868107982677254,0.4917894656358775,0.0454399405986849,0.0,0.57264,3.054741144180298,5.421279430389404,3.054741144180298,0.0086967326394127,task_arithmetic,0.0086967326394127,task_arithmetic,0.0,-0.0402466713594749,task_arithmetic,0,1,16
+est,spa,est vs spa,est+spa,0.8312147909707659,0.4870284459393449,0.0508681722157487,0.0,0.50466,2.038417339324951,2.252415657043457,2.038417339324951,0.010137615848525,task_arithmetic,0.010137615848525,task_arithmetic,0.0,0.0078739218095703,task_arithmetic,0,0,16
+hin,nld,hin vs nld,hin+nld,0.7995845298045842,0.4947049464285629,0.045106056606756,0.0,0.66648,2.331331253051758,8.879733085632324,2.331331253051758,0.0127850287452476,task_arithmetic,0.0127850287452476,task_arithmetic,0.0,0.014908605902532,task_arithmetic,0,0,16
+ell,pol,ell vs pol,ell+pol,0.6738185813729183,0.5231845589796554,0.0535777098590582,0.0,0.49748,2.6294195652008057,3.055079698562622,2.6294195652008057,0.0156490610536165,task_arithmetic,0.0156490610536165,task_arithmetic,0.0,-0.0130420176296415,task_arithmetic,0,1,16
+est,ita,est vs ita,est+ita,0.7545276854283437,0.4977575636266999,0.0518127988802315,0.0,0.51426,2.055943965911865,2.2086007595062256,2.055943965911865,0.0169801330217108,task_arithmetic,0.0169801330217108,task_arithmetic,0.0,0.0094456222674178,task_arithmetic,0,0,16
+hin,ita,hin vs ita,hin+ita,0.6895964345269007,0.4752180733745111,0.0456023752691863,0.0,0.60758,2.3152458667755127,7.176569938659668,2.3152458667755127,0.0180451855631513,task_arithmetic,0.0180451855631513,task_arithmetic,0.0,-0.0244487477660286,task_arithmetic,0,1,16
+ell,isl,ell vs isl,ell+isl,0.6368021348392121,0.5221437099920092,0.0527372882537445,0.0,0.50016,2.602349281311035,4.419780254364014,2.602349281311035,0.0191343250654784,task_arithmetic,0.0191343250654784,task_arithmetic,0.0,0.0066499044804583,task_arithmetic,0,0,16
+fin,nld,fin vs nld,fin+nld,0.858334163211254,0.4911143933688854,0.0511992251154618,0.0,0.5111,1.9867759943008425,2.242280960083008,1.9867759943008425,0.024614423670302,task_arithmetic,0.024614423670302,task_arithmetic,0.0,0.0082395686861229,task_arithmetic,0,0,16
+fin,ita,fin vs ita,fin+ita,0.7759277862469083,0.5000058589539089,0.0518393482488141,0.0,0.50838,1.968947172164917,2.1949169635772705,1.968947172164917,0.0321401469032058,task_arithmetic,0.0321401469032058,task_arithmetic,0.0,0.020282348935255,task_arithmetic,0,0,16
+deu,est,deu vs est,deu+est,0.8632345411762434,0.4931681702634027,0.0516852050084071,0.0,0.51704,2.207715511322021,2.1732540130615234,2.1732540130615234,0.0369427713090359,task_arithmetic,0.0369427713090359,task_arithmetic,0.0,0.0212034990007995,task_arithmetic,0,0,16
+ell,ita,ell vs ita,ell+ita,0.5535845129787077,0.5043853074838044,0.0528983032537568,0.0,0.49556,1.9633218050003047,4.606940746307373,1.9633218050003047,0.0378255320717273,task_arithmetic,0.0378255320717273,task_arithmetic,0.0,0.0218644276399512,task_arithmetic,0,0,16
+fin,spa,fin vs spa,fin+spa,0.8641424832817797,0.4898315819980065,0.0509131226527124,0.0,0.50082,1.8642144203186035,2.270033359527588,1.8642144203186035,0.0452939360400619,task_arithmetic,0.0452939360400619,task_arithmetic,0.0,0.0321697611049481,task_arithmetic,0,0,16
+ell,spa,ell vs spa,ell+spa,0.6210370042435197,0.5119222282261445,0.0518719114374989,0.0,0.49328,1.925220608711243,3.872457504272461,1.925220608711243,0.0461985541601101,task_arithmetic,0.0461985541601101,task_arithmetic,0.0,0.0259612415441536,task_arithmetic,0,0,16
+est,fin,est vs fin,est+fin,0.8459843360130395,0.49070550265407,0.0514884384573923,0.0,0.52296,2.2261855602264404,2.136149406433105,2.136149406433105,0.0464568968018875,task_arithmetic,0.0464568968018875,task_arithmetic,0.0,0.0367667969543487,task_arithmetic,0,0,16
+ell,fin,ell vs fin,ell+fin,0.6282795098322442,0.528083964233014,0.0529486714885604,0.0,0.47994,2.2796380519866943,3.421694755554199,2.2796380519866943,0.0466666562080511,task_arithmetic,0.0466666562080511,task_arithmetic,0.0,0.0207230145495029,task_arithmetic,0,0,16
+ell,est,ell vs est,ell+est,0.6768291029583274,0.5261082504743125,0.0528078562461357,0.0,0.48622,2.257206439971924,4.430761814117432,2.257206439971924,0.0497664091868842,task_arithmetic,0.0497664091868842,task_arithmetic,0.0,0.0146543774907815,task_arithmetic,0,0,16
+ell,nld,ell vs nld,ell+nld,0.6633390754596828,0.5207846443714634,0.0524633053360102,0.0,0.50768,1.946300983428955,5.227848052978516,1.946300983428955,0.0507081749736444,task_arithmetic,0.0507081749736444,task_arithmetic,0.0,0.0367826897149684,task_arithmetic,0,0,16
+ell,tur,ell vs tur,ell+tur,0.6656408612342765,0.522383268629448,0.0527299190186272,0.0,0.49832,2.420217037200928,4.91309118270874,2.420217037200928,0.0539333271993668,task_arithmetic,0.0539333271993668,task_arithmetic,0.0,0.0042140152727884,task_arithmetic,0,0,16
+deu,spa,deu vs spa,deu+spa,0.8872455448454941,0.485397374080886,0.0510402109374725,0.0,0.57636,1.8023582696914675,2.039104700088501,1.8023582696914675,0.0559072368341422,task_arithmetic,0.0559072368341422,task_arithmetic,0.0,0.0454939082073831,task_arithmetic,0,0,16
+fin,tur,fin vs tur,fin+tur,0.8658752688184197,0.4939177191533872,0.0515533177238562,0.0,0.5022,2.438411235809326,2.3206801414489746,2.3206801414489746,0.0938825750454249,task_arithmetic,0.0938825750454249,task_arithmetic,0.0,0.0501870934587869,task_arithmetic,0,0,16
+fin,pol,fin vs pol,fin+pol,0.8785169970276779,0.4959860612911416,0.0522716544300628,0.0,0.49666,2.5292296409606934,2.2906925678253174,2.2906925678253174,0.1252746058670082,task_arithmetic,0.1252746058670082,task_arithmetic,0.0,0.0988844447270833,task_arithmetic,0,0,16
+deu,ita,deu vs ita,deu+ita,0.8232688775105099,0.495359712151407,0.0517788678608692,0.0,0.58,1.832564353942871,1.710910439491272,1.710910439491272,0.1262911212579405,task_arithmetic,0.1262911212579405,task_arithmetic,0.0,0.1133247775926643,task_arithmetic,0,0,16
+fin,isl,fin vs isl,fin+isl,0.8686587480833791,0.4885064598013671,0.0516485799115295,0.0,0.50406,2.5511388778686523,2.3247201442718506,2.3247201442718506,0.1286106687159116,task_arithmetic,0.1286106687159116,task_arithmetic,0.0,0.1106292823721412,task_arithmetic,0,0,16
+est,isl,est vs isl,est+isl,0.8612357931475925,0.487224952283121,0.0514843268402267,0.0,0.51418,2.6147994995117188,2.3327832221984863,2.3327832221984863,0.1335374405099054,task_arithmetic,0.1335374405099054,task_arithmetic,0.0,0.125209287189958,task_arithmetic,0,0,16
+est,tur,est vs tur,est+tur,0.8694069644601152,0.4905035196745934,0.0514212829143724,0.0,0.51018,2.607272148132324,2.298980474472046,2.298980474472046,0.1371294926701998,task_arithmetic,0.1371294926701998,task_arithmetic,0.0,0.1110307848808358,task_arithmetic,0,0,16
+fin,rus,fin vs rus,fin+rus,0.8762666984103606,0.4968686976697567,0.0515250884100566,0.0,0.41138,2.823083877563477,2.3724822998046875,2.3724822998046875,0.1391309660863245,task_arithmetic,0.1391309660863245,task_arithmetic,0.0,0.1433289732639581,task_arithmetic,0,0,16
+est,pol,est vs pol,est+pol,0.8652512666166118,0.4936715950917139,0.0522487099636328,0.0,0.507,2.6668148040771484,2.3104288578033447,2.3104288578033447,0.1425459884439017,task_arithmetic,0.1425459884439017,task_arithmetic,0.0,0.1201335391539114,task_arithmetic,0,0,16
+est,rus,est vs rus,est+rus,0.8747691089513842,0.4942523509498968,0.0515218899035536,0.0,0.41656,1.9605400562286377,2.3423519134521484,1.9605400562286377,0.1430361854974176,task_arithmetic,0.1430361854974176,task_arithmetic,0.0,0.1219048628978289,task_arithmetic,0,0,16
+deu,nld,deu vs nld,deu+nld,0.88869970421588,0.4937224400413588,0.0512249870260221,0.0,0.62234,1.899802803993225,1.6765044927597046,1.6765044927597046,0.1549124299367321,task_arithmetic,0.1549124299367321,task_arithmetic,0.0,0.1536629578088417,task_arithmetic,0,0,16
+ell,zho,ell vs zho,ell+zho,0.6176811354038415,0.5246923031227959,0.0554145277023101,0.0,0.4862,9.62558650970459,7.722751617431641,7.722751617431641,0.1568779875058842,task_arithmetic,0.1568779875058842,task_arithmetic,0.0,0.1986851639908262,task_arithmetic,0,0,16
+deu,fin,deu vs fin,deu+fin,0.8719531836440131,0.4950684195910303,0.0516670428747139,0.0,0.50882,2.261837244033813,1.9199811220169067,1.9199811220169067,0.1838736309501154,task_arithmetic,0.1838736309501154,task_arithmetic,0.0,0.1621623014424031,task_arithmetic,0,0,16
+ell,hin,ell vs hin,ell+hin,0.6532086979877525,0.493514312540573,0.0466309176324886,0.0,0.52644,8.165210723876953,6.235628128051758,6.235628128051758,0.2149781372877389,qmd_guided,0.2149781372877389,qmd_guided,0.0,0.2149781372877389,qmd_guided,0,0,16
+deu,isl,deu vs isl,deu+isl,0.8931514829910866,0.4901547433017127,0.05175032696391,0.0,0.56222,2.405353307723999,2.033794641494751,2.033794641494751,0.2161031319571417,task_arithmetic,0.2161031319571417,task_arithmetic,0.0,0.1958045581466558,task_arithmetic,0,0,16
+ell,rus,ell vs rus,ell+rus,0.6722110718312354,0.5246938432676023,0.0526092470818956,0.0,0.43938,4.617556571960449,3.0823748111724854,3.0823748111724854,0.2236833016756808,task_arithmetic,0.2236833016756808,task_arithmetic,0.0,0.0791058178275035,qmd_guided,0,0,16
+deu,tur,deu vs tur,deu+tur,0.8887795792436464,0.4888516690685148,0.0516526603968452,0.0,0.5422,2.4516780376434326,2.0284693241119385,2.0284693241119385,0.2431203416730361,task_arithmetic,0.2431203416730361,task_arithmetic,0.0,0.1925291169493432,task_arithmetic,0,0,16
+deu,rus,deu vs rus,deu+rus,0.9052812456362352,0.4883215990861532,0.051705766565256,0.0,0.42178,2.4559996128082275,2.115139961242676,2.115139961242676,0.3254804690053812,task_arithmetic,0.3254804690053812,task_arithmetic,0.0,0.2767551067796572,task_arithmetic,0,0,16
+deu,pol,deu vs pol,deu+pol,0.9071028928471988,0.4879533758836566,0.0523914832560924,0.0,0.52718,2.4324381351470947,1.8376622200012207,1.8376622200012207,0.3476243410173237,task_arithmetic,0.3476243410173237,task_arithmetic,0.0,0.3303016951589018,task_arithmetic,0,0,16
+fin,zho,fin vs zho,fin+zho,0.8407812542299266,0.4812086014073191,0.0543814321019757,0.0,0.454,6.282923698425293,4.542793273925781,4.542793273925781,0.4848545092394946,task_arithmetic,0.4848545092394946,task_arithmetic,0.0,0.4998946145074256,task_arithmetic,0,0,16
+hin,jpn,hin vs jpn,hin+jpn,0.7402442307677861,0.4402786357618052,0.0440075322914218,0.0,0.08638,8.596170425415039,5.720637798309326,5.720637798309326,0.6150214298568639,task_arithmetic,0.6150214298568639,task_arithmetic,0.0,0.642716911920075,task_arithmetic,0,0,16
+deu,zho,deu vs zho,deu+zho,0.8692688641875307,0.4877283211928073,0.0543731633163107,0.0,0.46792,6.297134876251221,4.264947891235352,4.264947891235352,0.6689536868627466,task_arithmetic,0.6689536868627466,task_arithmetic,0.0,0.6972580354471546,task_arithmetic,0,0,16
+ell,jpn,ell vs jpn,ell+jpn,0.5861919187546282,0.4541693172481384,0.0517149279842309,0.0,0.06628,10.514047622680664,6.048466682434082,6.048466682434082,0.6953751783468304,task_arithmetic,0.6953751783468304,task_arithmetic,0.0,0.8038198268539832,task_arithmetic,0,0,16
+hin,kor,hin vs kor,hin+kor,0.7518509229681507,0.4543560939115326,0.0408666981309919,0.0,0.04592,10.789224624633787,6.471614837646484,6.471614837646484,0.7151322285064206,task_arithmetic,0.7151322285064206,task_arithmetic,0.0,0.7466276057400699,task_arithmetic,0,0,16
+est,zho,est vs zho,est+zho,0.819512630919377,0.4806034647310799,0.0542567973437267,0.0,0.45718,7.692380905151367,4.509434700012207,4.509434700012207,0.8794842322646683,task_arithmetic,0.8794842322646683,task_arithmetic,0.0,0.92732429097269,task_arithmetic,0,0,16
+deu,ell,deu vs ell,deu+ell,0.6534928485520405,0.520571177458284,0.0528891444095344,0.0,0.5014,3.4800124168396,2.04191517829895,2.04191517829895,1.0145263896077874,task_arithmetic,1.0145263896077874,task_arithmetic,0.0,0.915002581029196,task_arithmetic,0,0,16
+ell,kor,ell vs kor,ell+kor,0.640194311258744,0.4510560487441011,0.0491682404018267,0.0,0.03624,12.180006980895996,5.461300849914551,5.461300849914551,1.0193013677077227,task_arithmetic,1.0193013677077227,task_arithmetic,0.0,1.1825942109692174,task_arithmetic,0,0,16
+deu,jpn,deu vs jpn,deu+jpn,0.8160744024098396,0.4657472087176086,0.0507846225490123,0.0,0.07698,5.056319713592529,2.912078857421875,2.912078857421875,1.1195892237720466,task_arithmetic,1.1195892237720466,task_arithmetic,0.0,1.1128067591146789,task_arithmetic,0,0,16
+fin,jpn,fin vs jpn,fin+jpn,0.7875891175615976,0.4704437167087684,0.0507094618338441,0.0,0.07094,6.368600845336914,3.0893921852111816,3.0893921852111816,1.163275080356695,task_arithmetic,1.163275080356695,task_arithmetic,0.0,1.150499812854756,task_arithmetic,0,0,16
+est,jpn,est vs jpn,est+jpn,0.7604881626694144,0.4686238039353995,0.0506162953456793,0.0,0.07224,5.727605819702148,3.069748640060425,3.069748640060425,1.2774383421981306,task_arithmetic,1.2774383421981306,task_arithmetic,0.0,1.299304902450468,task_arithmetic,0,0,16
+fin,kor,fin vs kor,fin+kor,0.7850904478886663,0.4859736579875174,0.0481017341436057,0.0,0.03758,7.214295864105225,2.8839614391326904,2.8839614391326904,1.313268179597727,task_arithmetic,1.313268179597727,task_arithmetic,0.0,1.8453094940225496,task_arithmetic,0,0,16
+deu,hin,deu vs hin,deu+hin,0.7809226163307355,0.4922445189645929,0.0458528205937846,0.0,0.60788,5.173242092132568,2.427311897277832,2.427311897277832,1.3640793864265617,task_arithmetic,1.3640793864265617,task_arithmetic,0.0,1.401735841406306,task_arithmetic,0,0,16
+fin,hin,fin vs hin,fin+hin,0.7675650488917009,0.4954854118626991,0.0456221569866102,0.0,0.51084,6.428924560546875,2.596222162246704,2.596222162246704,1.37833423406091,task_arithmetic,1.37833423406091,task_arithmetic,0.0,1.3983334167482069,task_arithmetic,0,0,16
+est,hin,est vs hin,est+hin,0.8013882397599879,0.4930308577957909,0.0457297123859181,0.0,0.52026,7.995196342468262,2.563885450363159,2.563885450363159,1.6644202491883038,task_arithmetic,1.6644202491883038,task_arithmetic,0.0,1.6694916128984263,task_arithmetic,0,0,16
+deu,kor,deu vs kor,deu+kor,0.8182709476613191,0.4816346289681898,0.0481653330597128,0.0,0.03988,7.052877902984619,2.658069133758545,2.658069133758545,1.905618118623164,task_arithmetic,1.905618118623164,task_arithmetic,0.0,2.200376365063348,task_arithmetic,0,0,16
+est,kor,est vs kor,est+kor,0.7923798621996972,0.4858286110825638,0.0480544826429562,0.0,0.03816,9.671809196472168,2.887371778488159,2.887371778488159,2.205219596359032,task_arithmetic,2.205219596359032,task_arithmetic,0.0,2.661131504316703,task_arithmetic,0,0,16

overall/by_operator.csv

@@ -0,0 +1,41 @@
+regime,operator,arm,arm_meaning,n,n_pairs,rel_damage_mean,rel_damage_median,rel_damage_p90,delta_floor_mean_nats,frac_beats_best_parent,barrier_mean,barrier_n,retention_a_mean,retention_b_mean,crosstask_win_rate
+(all),average,aligned,"permutation-align B onto A first, then merge",273,273,0.22682272859475658,0.0715798279133955,0.4099599749328543,-0.9509106953065474,0.1575091575091575,0.36337763549369057,216,0.6762969970286702,0.6699434179440343,0.16483516483516483
+(all),average,naive,merge in the parents' own coordinates,273,273,0.22706175281484786,0.0715798279133955,0.4099599749328543,-0.9523471227058997,0.1575091575091575,0.3651931201234276,216,0.677683497577047,0.6665712584988962,0.16483516483516483
+(all),average,transport,"transport B's embeddings into A's vocabulary, then merge",57,57,0.7185131997606502,0.2360928047261467,2.359359458584048,-2.316514494126303,0.017543859649122806,,0,0.6430869722342673,0.07729803547449549,0.42105263157894735
+(all),average,transport+aligned,"transport, then permutation-align, then merge",57,57,0.7185131997606502,0.2360928047261467,2.359359458584048,-2.316514494126303,0.017543859649122806,,0,0.6430869722342673,0.07729803547449549,0.42105263157894735
+(all),qmd_guided,aligned,"permutation-align B onto A first, then merge",273,273,0.17587483212152227,0.0478706206986703,0.39693852171561206,-0.7418597351500403,0.1575091575091575,0.36337763549369057,216,0.763345553515259,0.6950522191400542,0.19047619047619047
+(all),qmd_guided,naive,merge in the parents' own coordinates,273,273,0.17822010969503085,0.0504454519350742,0.39693852171561206,-0.7557314887588278,0.1575091575091575,0.3651931201234276,216,0.7493247940958427,0.6909651761272301,0.19047619047619047
+(all),qmd_guided,transport,"transport B's embeddings into A's vocabulary, then merge",57,57,0.5548281675424492,0.1963981038391513,1.6783608005465473,-1.7967013915379841,0.017543859649122806,,0,0.7711439627403686,0.07289118611739627,0.3684210526315789
+(all),qmd_guided,transport+aligned,"transport, then permutation-align, then merge",57,57,0.5548281675424492,0.1963981038391513,1.6783608005465473,-1.7967013915379841,0.017543859649122806,,0,0.7711439627403686,0.07289118611739627,0.3684210526315789
+(all),task_arithmetic,aligned,"permutation-align B onto A first, then merge",273,273,0.1256516424908172,0.0318945557136733,0.30329716934547024,-0.5298613672291402,0.22344322344322345,0.36337763549369057,216,0.8485373128957066,0.7236294008567536,0.304029304029304
+(all),task_arithmetic,naive,merge in the parents' own coordinates,273,273,0.12594868604805834,0.0310517148720238,0.30329716934547024,-0.531606912612915,0.22344322344322345,0.3651931201234276,216,0.8489990621465912,0.7213917207946405,0.304029304029304
+(all),task_arithmetic,transport,"transport B's embeddings into A's vocabulary, then merge",57,57,0.423018176655472,0.1219048628978289,1.3389163081695634,-1.4041337130362528,0.10526315789473684,,0,0.892782174120031,0.08347472882923829,0.5964912280701754
+(all),task_arithmetic,transport+aligned,"transport, then permutation-align, then merge",57,57,0.423018176655472,0.1219048628978289,1.3389163081695634,-1.4041337130362528,0.10526315789473684,,0,0.892782174120031,0.08347472882923829,0.5964912280701754
+(all),ties,aligned,"permutation-align B onto A first, then merge",273,273,0.3067845526317015,0.125656022892863,0.5288406191610457,-1.233606972973862,0.04395604395604396,0.36337763549369057,216,0.5620584809744886,0.6269013981938882,0.14652014652014653
+(all),ties,naive,merge in the parents' own coordinates,273,273,0.30862335241668426,0.125656022892863,0.5487521451607779,-1.24446344768608,0.04395604395604396,0.3651931201234276,216,0.5614249968657457,0.6256654747310422,0.14652014652014653
+(all),ties,transport,"transport B's embeddings into A's vocabulary, then merge",57,57,1.0070286633793266,0.3053777974703323,3.091446802014662,-3.278372494797958,0.0,,0,0.43784319353485945,0.08932282810819765,0.5263157894736842
+(all),ties,transport+aligned,"transport, then permutation-align, then merge",57,57,1.0070286633793266,0.3053777974703323,3.091446802014662,-3.278372494797958,0.0,,0,0.43784319353485945,0.08932282810819765,0.5263157894736842
+beetle_curriculum,average,aligned,"permutation-align B onto A first, then merge",216,216,0.1028587226661411,0.0413491376506437,0.3324696145269651,-0.6286875958795901,0.19907407407407407,0.36337763549369057,216,0.7004384651847809,0.8312326453649932,0.17592592592592593
+beetle_curriculum,average,naive,merge in the parents' own coordinates,216,216,0.10316082272208982,0.0413491376506437,0.34334233575733103,-0.6305030805093271,0.19907407407407407,0.3651931201234276,216,0.702190847822313,0.8269706105107214,0.17592592592592593
+beetle_curriculum,qmd_guided,aligned,"permutation-align B onto A first, then merge",216,216,0.07492767734050451,0.037381206890426205,0.2883360968945248,-0.4717578755484687,0.19907407407407407,0.36337763549369057,216,0.7709880746949667,0.8608976791572077,0.18055555555555555
+beetle_curriculum,qmd_guided,naive,merge in the parents' own coordinates,216,216,0.07789184760702234,0.0401566047281453,0.2883360968945248,-0.4892902308040195,0.19907407407407407,0.3651931201234276,216,0.7532673926509822,0.8557321109048884,0.18055555555555555
+beetle_curriculum,task_arithmetic,aligned,"permutation-align B onto A first, then merge",216,216,0.050989752722164265,0.01705645525745065,0.13847289969563337,-0.3179746667544047,0.27314814814814814,0.36337763549369057,216,0.8393550784588591,0.8940977467005443,0.24537037037037038
+beetle_curriculum,task_arithmetic,naive,merge in the parents' own coordinates,216,216,0.051365182773677354,0.01705645525745065,0.13847289969563337,-0.3201808421700089,0.27314814814814814,0.3651931201234276,216,0.839938678206505,0.8912695677331515,0.24537037037037038
+beetle_curriculum,ties,aligned,"permutation-align B onto A first, then merge",216,216,0.12625973776070143,0.065286076717326,0.3287303470266809,-0.7504548916110286,0.05555555555555555,0.36337763549369057,216,0.6050678183599703,0.7806812315111048,0.1712962962962963
+beetle_curriculum,ties,naive,merge in the parents' own coordinates,216,216,0.12858377637783244,0.06587415579754305,0.3287303470266809,-0.7641762693723043,0.05555555555555555,0.3651931201234276,216,0.6042671648336424,0.7791191615788967,0.1712962962962963
+beetle_language,average,aligned,"permutation-align B onto A first, then merge",57,57,0.6965810668505625,0.3059709935031289,1.9042888477732904,-2.171966651029754,0.0,,0,0.5848135387528814,0.058742135085663644,0.12280701754385964
+beetle_language,average,naive,merge in the parents' own coordinates,57,57,0.6965810668505625,0.3059709935031289,1.9042888477732904,-2.171966651029754,0.0,,0,0.5848135387528814,0.058742135085663644,0.12280701754385964
+beetle_language,average,transport,"transport B's embeddings into A's vocabulary, then merge",57,57,0.7185131997606502,0.2360928047261467,2.359359458584048,-2.316514494126303,0.017543859649122806,,0,0.6430869722342673,0.07729803547449549,0.42105263157894735
+beetle_language,average,transport+aligned,"transport, then permutation-align, then merge",57,57,0.7185131997606502,0.2360928047261467,2.359359458584048,-2.316514494126303,0.017543859649122806,,0,0.6430869722342673,0.07729803547449549,0.42105263157894735
+beetle_language,qmd_guided,aligned,"permutation-align B onto A first, then merge",57,57,0.5584114186601159,0.209290701042359,1.646646228349652,-1.765403624166522,0.0,,0,0.7343844206237349,0.06658521275926152,0.22807017543859648
+beetle_language,qmd_guided,naive,merge in the parents' own coordinates,57,57,0.5584114186601159,0.209290701042359,1.646646228349652,-1.765403624166522,0.0,,0,0.7343844206237349,0.06658521275926152,0.22807017543859648
+beetle_language,qmd_guided,transport,"transport B's embeddings into A's vocabulary, then merge",57,57,0.5548281675424492,0.1963981038391513,1.6783608005465473,-1.7967013915379841,0.017543859649122806,,0,0.7711439627403686,0.07289118611739627,0.3684210526315789
+beetle_language,qmd_guided,transport+aligned,"transport, then permutation-align, then merge",57,57,0.5548281675424492,0.1963981038391513,1.6783608005465473,-1.7967013915379841,0.017543859649122806,,0,0.7711439627403686,0.07289118611739627,0.3684210526315789
+beetle_language,task_arithmetic,aligned,"permutation-align B onto A first, then merge",57,57,0.4085809089825547,0.1391309660863245,1.291770277157969,-1.3328004427123488,0.03508771929824561,,0,0.883333148656391,0.07764409029080976,0.5263157894736842
+beetle_language,task_arithmetic,naive,merge in the parents' own coordinates,57,57,0.4085809089825547,0.1391309660863245,1.291770277157969,-1.3328004427123488,0.03508771929824561,,0,0.883333148656391,0.07764409029080976,0.5263157894736842
+beetle_language,task_arithmetic,transport,"transport B's embeddings into A's vocabulary, then merge",57,57,0.423018176655472,0.1219048628978289,1.3389163081695634,-1.4041337130362528,0.10526315789473684,,0,0.892782174120031,0.08347472882923829,0.5964912280701754
+beetle_language,task_arithmetic,transport+aligned,"transport, then permutation-align, then merge",57,57,0.423018176655472,0.1219048628978289,1.3389163081695634,-1.4041337130362528,0.10526315789473684,,0,0.892782174120031,0.08347472882923829,0.5964912280701754
+beetle_language,ties,aligned,"permutation-align B onto A first, then merge",57,57,0.9908785879323333,0.4859243805652349,2.5827241710769258,-3.064499070769862,0.0,,0,0.3990757287768742,0.044156766676014945,0.05263157894736842
+beetle_language,ties,naive,merge in the parents' own coordinates,57,57,0.9908785879323333,0.4859243805652349,2.5827241710769258,-3.064499070769862,0.0,,0,0.3990757287768742,0.044156766676014945,0.05263157894736842
+beetle_language,ties,transport,"transport B's embeddings into A's vocabulary, then merge",57,57,1.0070286633793266,0.3053777974703323,3.091446802014662,-3.278372494797958,0.0,,0,0.43784319353485945,0.08932282810819765,0.5263157894736842
+beetle_language,ties,transport+aligned,"transport, then permutation-align, then merge",57,57,1.0070286633793266,0.3053777974703323,3.091446802014662,-3.278372494797958,0.0,,0,0.43784319353485945,0.08932282810819765,0.5263157894736842

overall/summary.csv

@@ -0,0 +1,24 @@
+metric,scope,value,n,note
+rows_in_snapshot,zoo_latest,2640,2640,sweep still running; row count grows
+pairs,zoo_latest,273,2640,
+operators,zoo_latest,4,2640,average|qmd_guided|task_arithmetic|ties
+alignment_arms,zoo_latest,4,2640,naive|aligned|transport|transport+aligned
+align_method,zoo_latest,permutation,2640,only one alignment family is exercised in this run
+rel_damage_median,arm=naive,0.05774,1092,(nll_merge - nll_floor)/nll_floor; lower is better
+frac_beats_best_parent,arm=naive,0.1456,1092,delta_floor > 0
+rel_damage_median,arm=aligned,0.05549,1092,(nll_merge - nll_floor)/nll_floor; lower is better
+frac_beats_best_parent,arm=aligned,0.1456,1092,delta_floor > 0
+rel_damage_median,arm=transport,0.21269,228,(nll_merge - nll_floor)/nll_floor; lower is better
+frac_beats_best_parent,arm=transport,0.0351,228,delta_floor > 0
+rel_damage_median,arm=transport+aligned,0.21269,228,(nll_merge - nll_floor)/nll_floor; lower is better
+frac_beats_best_parent,arm=transport+aligned,0.0351,228,delta_floor > 0
+cells_with_both_arms,naive vs aligned,1092,1092,
+cells_where_alignment_changed_the_merge,naive vs aligned,138,1092,138/1092 = 12.6% of cells differ at all
+alignment_gap_mean_where_it_moved,naive vs aligned,0.00934,138,rel_damage_naive - rel_damage_aligned; >0 = alignment helped
+alignment_helped_frac_where_it_moved,naive vs aligned,0.6667,138,
+pairs_where_best_aligned_beats_best_naive,per pair,16,273,
+pairs_where_best_aligned_loses_to_best_naive,per pair,2,273,
+coord_fraction_median,zoo_latest,0.0,2640,QMD coordinate share; ~0 in this run -- see README caveat 2
+transport_arms_bit_identical_frac,transport vs transport+aligned,1.0,228,1.0 means the +aligned arm carries no information -- see README caveat 3
+rows_with_jointly_trained_ceiling,zoo_latest,0,2640,0 => MS / nll_ceiling are empty; everything is floor-relative
+archive_rows_published_separately,archive/,8116,8116,zoo|zoo_2026_08_17|zoo_seed|zoo_seed_v2

provenance.csv

@@ -0,0 +1,28 @@
+sweep,source_path,shard,rows,columns,source_mtime_utc,sha256,status,published_as
+zoo,data/sweeps/zoo/zoo_shard0.csv,zoo_shard0,324,160,2026-08-09T15:47:12Z,6fa770711a8a7ddfddd00eb025aedc2a2f6a2b1a27f71650a2df2330f8c16350,complete,archive/previous_sweeps.csv
+zoo,data/sweeps/zoo/zoo_shard1.csv,zoo_shard1,328,154,2026-08-09T16:02:39Z,abac2f57b140d8be8a2f90dffa423d15b67d2d47ef9a36b0995e6d0f607f8c95,complete,archive/previous_sweeps.csv
+zoo,data/sweeps/zoo/zoo_shard2.csv,zoo_shard2,356,160,2026-08-09T16:00:04Z,d59a1b16c05917c9d7b63b9b0cba89591d43c145d535f1ff70d6b80a55044397,complete,archive/previous_sweeps.csv
+zoo,data/sweeps/zoo/zoo_shard3.csv,zoo_shard3,340,160,2026-08-09T15:55:51Z,7f657be7e3bdb23410e54f8028d722c7cce76d9760f073691269c07bd5a97176,complete,archive/previous_sweeps.csv
+zoo,data/sweeps/zoo/zoo_shard4.csv,zoo_shard4,328,154,2026-08-09T16:04:46Z,84aad343c4fce888bb37ceafd80531b0f5aedc12dc859823cd80d16b480f7a59,complete,archive/previous_sweeps.csv
+zoo,data/sweeps/zoo/zoo_shard5.csv,zoo_shard5,328,160,2026-08-09T15:48:58Z,f49053ff0a5a4e87d8d471fe16bc98aff65676b154d0ad2861ccd9ca0d79b1dc,complete,archive/previous_sweeps.csv
+zoo,data/sweeps/zoo/zoo_shard6.csv,zoo_shard6,304,160,2026-08-09T15:56:18Z,0fc4a13086cd0560ed38e851a7c52a72df1cf111f75545b4d20b2f118b80810b,complete,archive/previous_sweeps.csv
+zoo_seed,data/sweeps/zoo_seed/zoo_shard0.csv,zoo_shard0,344,143,2026-08-09T22:59:30Z,85766b82212a325548d17eb8133161918a493eeec208ed31e3af12d0213c76e6,complete,archive/previous_sweeps.csv
+zoo_seed,data/sweeps/zoo_seed/zoo_shard1.csv,zoo_shard1,352,137,2026-08-09T23:03:59Z,fb10ff4c8f55ca9cf75c557fbab3723cd8711f69cfcf99e0e2f53851da7a8f60,complete,archive/previous_sweeps.csv
+zoo_seed,data/sweeps/zoo_seed/zoo_shard2.csv,zoo_shard2,336,137,2026-08-09T22:57:46Z,4b7265d42dfb37adf9999c116078ad64fe05dcb939aa57f958b131fa054504cb,complete,archive/previous_sweeps.csv
+zoo_seed,data/sweeps/zoo_seed/zoo_shard3.csv,zoo_shard3,336,137,2026-08-09T22:56:49Z,15c1aeba7385105ee25bdb59468ca62b1c9fadde81683ab588bff7df10e5a559,complete,archive/previous_sweeps.csv
+zoo_seed,data/sweeps/zoo_seed/zoo_shard4.csv,zoo_shard4,336,126,2026-08-09T22:57:22Z,b810f3daa8f02a24ed84a93ab51f928ee60ed0412375e8608f11c4c377229757,complete,archive/previous_sweeps.csv
+zoo_seed,data/sweeps/zoo_seed/zoo_shard5.csv,zoo_shard5,352,143,2026-08-09T23:07:55Z,0b5a43be0cbaf69e13c83af59445410d3603dfe342f1a6710751e5ce23568d37,complete,archive/previous_sweeps.csv
+zoo_seed,data/sweeps/zoo_seed/zoo_shard6.csv,zoo_shard6,360,137,2026-08-09T23:08:10Z,d3f32d3319e9a0d34319652e3913e187127286df7b932b3ee7069ba66153eba6,complete,archive/previous_sweeps.csv
+zoo_seed,data/sweeps/zoo_seed/zoo_shard7.csv,zoo_shard7,336,131,2026-08-09T22:59:45Z,81ef9c3f1bea54f397952dc17aef4ee1fb15f952723fc262de97689b02fccf60,complete,archive/previous_sweeps.csv
+zoo_seed_v2,data/sweeps/zoo_seed_v2/zoo_shard0.csv,zoo_shard0,296,148,2026-08-11T01:49:12Z,d3062ed24e1af41c7b17a8b477b192ef2729571fb9443ea9e3bc392ee9b30925,complete,archive/previous_sweeps.csv
+zoo_seed_v2,data/sweeps/zoo_seed_v2/zoo_shard1.csv,zoo_shard1,296,148,2026-08-11T01:42:31Z,ffe146eaf5e624113046096de3d6a05f8a09e310be46015c37aca9eb6b7c507d,complete,archive/previous_sweeps.csv
+zoo_seed_v2,data/sweeps/zoo_seed_v2/zoo_shard2.csv,zoo_shard2,280,142,2026-08-11T01:34:53Z,e6a4c7aa14748cc7fb57d1d371e9c8f0e761e6f11654dcced87dea69c5aa38f9,complete,archive/previous_sweeps.csv
+zoo_seed_v2,data/sweeps/zoo_seed_v2/zoo_shard3.csv,zoo_shard3,280,142,2026-08-11T01:37:47Z,291c073ba37343685be7f0cd6df613e32f64c0873992e50fc7963a396c726a20,complete,archive/previous_sweeps.csv
+zoo_seed_v2,data/sweeps/zoo_seed_v2/zoo_shard4.csv,zoo_shard4,280,136,2026-08-11T01:37:22Z,b40ce0a31536be69cf23d1bf38559604f0b0afa0141c8c2d80ebabed40f91a7f,complete,archive/previous_sweeps.csv
+zoo_seed_v2,data/sweeps/zoo_seed_v2/zoo_shard5.csv,zoo_shard5,296,154,2026-08-11T01:51:44Z,3bec3d60b0661dcbb954357c00281d8fbdbd23d477202c322284e5cf700b744a,complete,archive/previous_sweeps.csv
+zoo_seed_v2,data/sweeps/zoo_seed_v2/zoo_shard6.csv,zoo_shard6,304,142,2026-08-11T01:55:04Z,9df2386a3594d70d43e43921c83c39b21477c28198dbfe7eaf62aa940b887c97,complete,archive/previous_sweeps.csv
+zoo_seed_v2,data/sweeps/zoo_seed_v2/zoo_shard7.csv,zoo_shard7,288,136,2026-08-11T01:37:45Z,195b1063752a8dbd7247367380d5a4a2336e10070be5f2445944e5f16b41d8e6,complete,archive/previous_sweeps.csv
+zoo_2026_08_17,data/sweeps/zoo_2026_08_17/zoo_shard0.csv,zoo_shard0,256,142,2026-08-17T02:26:28Z,bf51062e2326ae1d7e401986854320a8ab3f37b5707fe437b9a8e5a6262883b7,complete,archive/previous_sweeps.csv
+zoo_2026_08_17,data/sweeps/zoo_2026_08_17/zoo_shard1.csv,zoo_shard1,240,142,2026-08-17T02:24:43Z,6554f5b90638c3869b6835e524d064d7d665ec6cd99bdefcc34b125a9b3835cc,complete,archive/previous_sweeps.csv
+zoo_2026_08_17,data/sweeps/zoo_2026_08_17/zoo_shard2.csv,zoo_shard2,240,154,2026-08-17T02:16:48Z,5b5eb0fb285a02f04d7c3d869377d7d167cc46483ac8bb84b2e381886abc00e8,complete,archive/previous_sweeps.csv
+zoo_latest,data/sweeps/zoo_latest/zoo_shard0.csv,zoo_shard0,2640,178,2026-08-19T03:42:51Z,681ded834ec5c00b8e84f9cfb2627ada0294c6c0b7b07dab9f87392367ba2c89,live (appending during snapshot),raw/zoo_latest_shard0.csv