From 3aaaa12566f075683c9b2442ac45c139c2c652f8 Mon Sep 17 00:00:00 2001 From: Claude Date: Wed, 14 Oct 2026 23:11:14 +0000 Subject: [PATCH] Add(SymbolicLearning,#2161): stubs d'exercices sous les trois exercices de SL-13b Ajoute une cellule de code stub (C.1) sous chacun des trois exercices en prose (reverse, ablation par symbole, sur-completude 32/64/128). Le compteur count_exercises.py passe de 0 a 3. Le carnet est re-execute en place sous py313 (papermill complet, 0 erreur). Sept sorties de cellules changent : versions numpy/torch de la cellule 1 (2.4.4 / 2.11.0 committes, 2.5.4 / 2.14.1 ici), duree d'entrainement de la cellule 5 (machine-dependante) et cinq cellules de mesure. La prose qui cite ces valeurs est realignee sur la nouvelle sortie ; la duree d'entrainement est retiree de la prose. Voir le body de PR (section Diagnostic derive). See #2161 Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_01LRbWZZHE9jZscr8ya3niWA --- .../SymbolicLearning/SL-13b-TPR-SAE.ipynb | 443 ++++++++++++------ 1 file changed, 289 insertions(+), 154 deletions(-) diff --git a/MyIA.AI.Notebooks/SymbolicAI/SymbolicLearning/SL-13b-TPR-SAE.ipynb b/MyIA.AI.Notebooks/SymbolicAI/SymbolicLearning/SL-13b-TPR-SAE.ipynb index 3eb530b43d..88d0e9e474 100644 --- a/MyIA.AI.Notebooks/SymbolicAI/SymbolicLearning/SL-13b-TPR-SAE.ipynb +++ b/MyIA.AI.Notebooks/SymbolicAI/SymbolicLearning/SL-13b-TPR-SAE.ipynb @@ -5,10 +5,10 @@ "id": "f029855a", "metadata": { "papermill": { - "duration": 0.006031, - "end_time": "2026-09-28T11:17:55.526292+00:00", + "duration": 0.005228, + "end_time": "2026-10-14T23:24:49.072644+00:00", "exception": false, - "start_time": "2026-09-28T11:17:55.520261+00:00", + "start_time": "2026-10-14T23:24:49.067416+00:00", "status": "completed" }, "tags": [] @@ -54,16 +54,16 @@ "id": "24325bff", "metadata": { "execution": { - "iopub.execute_input": "2026-09-28T11:17:55.545337Z", - "iopub.status.busy": "2026-09-28T11:17:55.544713Z", - "iopub.status.idle": "2026-09-28T11:18:01.105841Z", - "shell.execute_reply": "2026-09-28T11:18:01.104731Z" + "iopub.execute_input": "2026-10-14T23:24:49.090212Z", + "iopub.status.busy": "2026-10-14T23:24:49.089786Z", + "iopub.status.idle": "2026-10-14T23:24:52.377105Z", + "shell.execute_reply": "2026-10-14T23:24:52.376287Z" }, "papermill": { - "duration": 5.575297, - "end_time": "2026-09-28T11:18:01.106999+00:00", + "duration": 3.297208, + "end_time": "2026-10-14T23:24:52.377781+00:00", "exception": false, - "start_time": "2026-09-28T11:17:55.531702+00:00", + "start_time": "2026-10-14T23:24:49.080573+00:00", "status": "completed" }, "tags": [] @@ -73,7 +73,7 @@ "name": "stdout", "output_type": "stream", "text": [ - "numpy 2.4.4 | matplotlib 3.10.8 | torch 2.11.0+cpu\n", + "numpy 2.5.4 | matplotlib 3.11.2 | torch 2.14.1+cpu\n", "device = cpu | organe : ict/sae_dictionary.py (serie ICT, SAE top-k sur-complet)\n" ] } @@ -125,10 +125,10 @@ "id": "00724155", "metadata": { "papermill": { - "duration": 0.004477, - "end_time": "2026-09-28T11:18:01.116094+00:00", + "duration": 0.00792, + "end_time": "2026-10-14T23:24:52.392474+00:00", "exception": false, - "start_time": "2026-09-28T11:18:01.111617+00:00", + "start_time": "2026-10-14T23:24:52.384554+00:00", "status": "completed" }, "tags": [] @@ -158,16 +158,16 @@ "id": "231a3b7b", "metadata": { "execution": { - "iopub.execute_input": "2026-09-28T11:18:01.126991Z", - "iopub.status.busy": "2026-09-28T11:18:01.126450Z", - "iopub.status.idle": "2026-09-28T11:18:01.167923Z", - "shell.execute_reply": "2026-09-28T11:18:01.165980Z" + "iopub.execute_input": "2026-10-14T23:24:52.401924Z", + "iopub.status.busy": "2026-10-14T23:24:52.401442Z", + "iopub.status.idle": "2026-10-14T23:24:52.421103Z", + "shell.execute_reply": "2026-10-14T23:24:52.420329Z" }, "papermill": { - "duration": 0.048894, - "end_time": "2026-09-28T11:18:01.169481+00:00", + "duration": 0.025842, + "end_time": "2026-10-14T23:24:52.422218+00:00", "exception": false, - "start_time": "2026-09-28T11:18:01.120587+00:00", + "start_time": "2026-10-14T23:24:52.396376+00:00", "status": "completed" }, "tags": [] @@ -210,10 +210,10 @@ "id": "3feab6f2", "metadata": { "papermill": { - "duration": 0.007013, - "end_time": "2026-09-28T11:18:01.180881+00:00", + "duration": 0.007651, + "end_time": "2026-10-14T23:24:52.432438+00:00", "exception": false, - "start_time": "2026-09-28T11:18:01.173868+00:00", + "start_time": "2026-10-14T23:24:52.424787+00:00", "status": "completed" }, "tags": [] @@ -230,16 +230,16 @@ "id": "27a675e8", "metadata": { "execution": { - "iopub.execute_input": "2026-09-28T11:18:01.200626Z", - "iopub.status.busy": "2026-09-28T11:18:01.199780Z", - "iopub.status.idle": "2026-09-28T11:18:03.819767Z", - "shell.execute_reply": "2026-09-28T11:18:03.818734Z" + "iopub.execute_input": "2026-10-14T23:24:52.445620Z", + "iopub.status.busy": "2026-10-14T23:24:52.445321Z", + "iopub.status.idle": "2026-10-14T23:28:01.364915Z", + "shell.execute_reply": "2026-10-14T23:28:01.360088Z" }, "papermill": { - "duration": 2.634342, - "end_time": "2026-09-28T11:18:03.820626+00:00", + "duration": 188.925163, + "end_time": "2026-10-14T23:28:01.365533+00:00", "exception": false, - "start_time": "2026-09-28T11:18:01.186284+00:00", + "start_time": "2026-10-14T23:24:52.440370+00:00", "status": "completed" }, "tags": [] @@ -249,7 +249,13 @@ "name": "stdout", "output_type": "stream", "text": [ - "copy (protocole BOS) | epochs=60 | train_time=2.59s | val_acc=0.817 (hasard = 0.091)\n", + "copy (protocole BOS) | epochs=60 | train_time=188.13s | val_acc=0.817 (hasard = 0.091)\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ "H_train (800, 32) | H_val (200, 32)\n" ] } @@ -363,10 +369,10 @@ "id": "d6a083b4", "metadata": { "papermill": { - "duration": 0.004453, - "end_time": "2026-09-28T11:18:03.829829+00:00", + "duration": 0.003951, + "end_time": "2026-10-14T23:28:01.372690+00:00", "exception": false, - "start_time": "2026-09-28T11:18:03.825376+00:00", + "start_time": "2026-10-14T23:28:01.368739+00:00", "status": "completed" }, "tags": [] @@ -374,19 +380,19 @@ "source": [ "### Lecture du résultat d'entraînement\n", "\n", - "**Sortie obtenue** : `val_acc = 0.817` sous le protocole BOS (hasard par token = 1/11 = 0.091, entraînement en ~2 s CPU). Le protocole fait de *chaque position* une cible : à l'étape t, le décodeur a consommé BOS et les premiers symboles — le symbole cible (position t) n'est jamais dans son entrée au moment où il doit le produire, donc il ne peut venir que de `H`. La position 0 elle-même n'est produite qu'à partir de BOS et de `H`.\n", + "**Sortie obtenue** : `val_acc = 0.817` sous le protocole BOS (hasard par token = 1/11 = 0.091 ; entraînement sur CPU, durée : voir la sortie de la cellule d'entraînement). Le protocole fait de *chaque position* une cible : à l'étape t, le décodeur a consommé BOS et les premiers symboles — le symbole cible (position t) n'est jamais dans son entrée au moment où il doit le produire, donc il ne peut venir que de `H`. La position 0 elle-même n'est produite qu'à partir de BOS et de `H`.\n", "\n", "| Aspect | Valeur | Signification |\n", "|--------|--------|---------------|\n", "| val_acc (teacher-forced) | 0.817 sur 5 positions | décodage correct ~4 fois sur 5 — l'information utile passe par `H` |\n", "| hasard par token | 1/11 = 0.091 | 0.817 est très au-dessus du hasard, à ~0.18 d'une lecture parfaite |\n", - "| budget | ~2 s CPU | tient dans le budget d'un notebook |\n", + "| budget | CPU seul, pas de GPU | exécutable sans GPU |\n", "\n", "**Points clés** :\n", "\n", "1. Ce que le chiffre *établit* : l'information nécessaire au décodage passe par l'état caché final `H` — le décodeur ne voit jamais le symbole qu'il prédit, et les séquences de val sont nouvelles (0.817 ne peut pas venir d'une mémorisation mot à mot). C'est un fait de *passage d'information*, pas encore une affirmation sur sa *forme* — c'est précisément ce que les deux lectures concurrentes vont se disputer.\n", "2. Nuance de protocole : l'évaluation (comme l'entraînement) est *teacher-forced* — le décodeur voit les symboles précédents. Le score sert à comparer des lectures entre elles sur le même banc ; il ne qualifie pas la génération libre.\n", - "3. Ne pas comparer ce 0.817 au 0.835 de SL-13 : le protocole diffère (SL-13 ne prédisait pas la position 0). La comparabilité porte sur l'architecture et la tâche, pas sur le scalaire d'accuracy.\n", + "3. Ne pas comparer ce 0.817 au score de SL-13 : le protocole diffère (SL-13 ne prédisait pas la position 0). La comparabilité porte sur l'architecture et la tâche, pas sur le scalaire d'accuracy.\n", "4. La suite : la *forme* de l'information dans `H` — objet de la lecture TPR, puis de la lecture SAE." ] }, @@ -396,16 +402,16 @@ "id": "151fbd33", "metadata": { "execution": { - "iopub.execute_input": "2026-09-28T11:18:03.839024Z", - "iopub.status.busy": "2026-09-28T11:18:03.838620Z", - "iopub.status.idle": "2026-09-28T11:18:04.714758Z", - "shell.execute_reply": "2026-09-28T11:18:04.713683Z" + "iopub.execute_input": "2026-10-14T23:28:01.382496Z", + "iopub.status.busy": "2026-10-14T23:28:01.382317Z", + "iopub.status.idle": "2026-10-14T23:28:03.727028Z", + "shell.execute_reply": "2026-10-14T23:28:03.726424Z" }, "papermill": { - "duration": 0.881946, - "end_time": "2026-09-28T11:18:04.715632+00:00", + "duration": 2.349586, + "end_time": "2026-10-14T23:28:03.727582+00:00", "exception": false, - "start_time": "2026-09-28T11:18:03.833686+00:00", + "start_time": "2026-10-14T23:28:01.377996+00:00", "status": "completed" }, "tags": [] @@ -534,10 +540,10 @@ "id": "d01f7d42", "metadata": { "papermill": { - "duration": 0.003908, - "end_time": "2026-09-28T11:18:04.724035+00:00", + "duration": 0.007962, + "end_time": "2026-10-14T23:28:03.740502+00:00", "exception": false, - "start_time": "2026-09-28T11:18:04.720127+00:00", + "start_time": "2026-10-14T23:28:03.732540+00:00", "status": "completed" }, "tags": [] @@ -566,16 +572,16 @@ "id": "7d87b336", "metadata": { "execution": { - "iopub.execute_input": "2026-09-28T11:18:04.733443Z", - "iopub.status.busy": "2026-09-28T11:18:04.733101Z", - "iopub.status.idle": "2026-09-28T11:18:06.500364Z", - "shell.execute_reply": "2026-09-28T11:18:06.498812Z" + "iopub.execute_input": "2026-10-14T23:28:03.752243Z", + "iopub.status.busy": "2026-10-14T23:28:03.751838Z", + "iopub.status.idle": "2026-10-14T23:29:52.596680Z", + "shell.execute_reply": "2026-10-14T23:29:52.596189Z" }, "papermill": { - "duration": 1.773378, - "end_time": "2026-09-28T11:18:06.501471+00:00", + "duration": 108.85347, + "end_time": "2026-10-14T23:29:52.601006+00:00", "exception": false, - "start_time": "2026-09-28T11:18:04.728093+00:00", + "start_time": "2026-10-14T23:28:03.747536+00:00", "status": "completed" }, "tags": [] @@ -586,8 +592,8 @@ "output_type": "stream", "text": [ "SAE top-k (k=8) | atomes = 64 (sur-complet x2) | 400 pas full-batch\n", - "FVU train = 0.0746 | L0 train = 8.00 (borne <= 8)\n", - "FVU val (dictionnaire fige, encode-only) = 0.1032\n", + "FVU train = 0.0744 | L0 train = 8.00 (borne <= 8)\n", + "FVU val (dictionnaire fige, encode-only) = 0.1033\n", "codes train (800, 64) | codes val (200, 64)\n" ] } @@ -615,10 +621,10 @@ "id": "2ffb1d87", "metadata": { "papermill": { - "duration": 0.006125, - "end_time": "2026-09-28T11:18:06.515052+00:00", + "duration": 0.002479, + "end_time": "2026-10-14T23:29:52.610973+00:00", "exception": false, - "start_time": "2026-09-28T11:18:06.508927+00:00", + "start_time": "2026-10-14T23:29:52.608494+00:00", "status": "completed" }, "tags": [] @@ -626,13 +632,13 @@ "source": [ "### Lecture de l'entraînement SAE\n", "\n", - "**Sortie obtenue** : le dictionnaire top-k sur-complet (64 atomes pour d = 32) reconstruit `H` en laissant non expliqués ~7,5 % de la variance en train et ~10 % en val, en activant au plus 8 atomes par séquence.\n", + "**Sortie obtenue** : le dictionnaire top-k sur-complet (64 atomes pour d = 32) reconstruit `H` en laissant une fraction de variance non expliquée (FVU) de 0.0744 en train et de 0.1033 en val, en activant au plus 8 atomes par séquence.\n", "\n", "| Aspect | Valeur | Signification |\n", "|--------|--------|---------------|\n", - "| FVU train | ~0.075 | le dictionnaire couvre l'essentiel de la variance de H |\n", + "| FVU train | 0.0744 | le dictionnaire couvre l'essentiel de la variance de H |\n", "| L0 <= 8 | borne top-k | chaque séquence est décrite par peu d'atomes |\n", - "| FVU val | ~0.10 | le dictionnaire transfère sans sur-ajustement massif (écart train/val ~0.03) |\n", + "| FVU val | 0.1033 | le dictionnaire transfère sans sur-ajustement massif (écart train/val 0.0289) |\n", "\n", "**Points clés** :\n", "\n", @@ -647,16 +653,16 @@ "id": "84a74fe8", "metadata": { "execution": { - "iopub.execute_input": "2026-09-28T11:18:06.527493Z", - "iopub.status.busy": "2026-09-28T11:18:06.527036Z", - "iopub.status.idle": "2026-09-28T11:18:11.073501Z", - "shell.execute_reply": "2026-09-28T11:18:11.072308Z" + "iopub.execute_input": "2026-10-14T23:29:52.625019Z", + "iopub.status.busy": "2026-10-14T23:29:52.624824Z", + "iopub.status.idle": "2026-10-14T23:29:55.404079Z", + "shell.execute_reply": "2026-10-14T23:29:55.403069Z" }, "papermill": { - "duration": 4.553558, - "end_time": "2026-09-28T11:18:11.074595+00:00", + "duration": 2.792649, + "end_time": "2026-10-14T23:29:55.406320+00:00", "exception": false, - "start_time": "2026-09-28T11:18:06.521037+00:00", + "start_time": "2026-10-14T23:29:52.613671+00:00", "status": "completed" }, "tags": [] @@ -667,14 +673,14 @@ "output_type": "stream", "text": [ "Facteurs (position, symbole) testes : 40 (chacun ~86 positifs sur 800)\n", - "Dedies (z_max > 5 contre null apparie 200 tirages) : 34/40\n", - "z_max : mediane = 7.7 | max = 19.6 | min = 3.8\n", + "Dedies (z_max > 5 contre null apparie 200 tirages) : 35/40\n", + "z_max : mediane = 7.7 | max = 19.5 | min = 3.9\n", "AUC de la feature la plus selective par facteur : mediane = 0.658\n" ] }, { "data": { - "image/png": "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", + "image/png": "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", "text/plain": [ "
" ] @@ -731,10 +737,10 @@ "id": "57b387e0", "metadata": { "papermill": { - "duration": 0.004276, - "end_time": "2026-09-28T11:18:11.083729+00:00", + "duration": 0.022619, + "end_time": "2026-10-14T23:29:55.436541+00:00", "exception": false, - "start_time": "2026-09-28T11:18:11.079453+00:00", + "start_time": "2026-10-14T23:29:55.413922+00:00", "status": "completed" }, "tags": [] @@ -746,15 +752,15 @@ "\n", "| Aspect | Valeur | Signification |\n", "|--------|--------|---------------|\n", - "| facteurs dédiés | 34/40 au seuil z > 5 | le dictionnaire porte une large part des facteurs au seuil choisi |\n", - "| z_max médian | 7.7 (max 19.6, min 3.8) | la sélectivité dépasse le null apparié pour la majorité des facteurs |\n", + "| facteurs dédiés | 35/40 au seuil z > 5 | le dictionnaire porte une large part des facteurs au seuil choisi |\n", + "| z_max médian | 7.7 (max 19.5, min 3.9) | la sélectivité dépasse le null apparié pour la majorité des facteurs |\n", "| AUC médiane de la meilleure feature | 0.658 | discrimination modérée par facteur — pas une lecture binaire nette |\n", "| structure par position | lignes de la heatmap | la couverture des positions est inégale (les z par ligne varient) |\n", "\n", "**Points clés** :\n", "\n", "1. La discipline du null apparié (héritage de la série ICT) : une feature ne compte comme dédiée que si son AUC écrase ce que le hasard du relabellement produit — pas de seuil absolu arbitraire.\n", - "2. La couverture n'est pas totale (34/40) : quelques facteurs n'ont pas d'atome au-dessus du seuil — cohérent avec le plateau de mélange documenté par l'organe (des atomes polyvalents au lieu d'atomes dédiés).\n", + "2. La couverture n'est pas totale (35/40) : quelques facteurs n'ont pas d'atome au-dessus du seuil — cohérent avec le plateau de mélange documenté par l'organe (des atomes polyvalents au lieu d'atomes dédiés).\n", "3. Ce que cette cellule ne dit pas : *sous quelle forme géométrique* la sélectivité est portée — c'est la cellule suivante, et c'est le cœur de la confrontation avec la TPR." ] }, @@ -764,16 +770,16 @@ "id": "b205bc49", "metadata": { "execution": { - "iopub.execute_input": "2026-09-28T11:18:11.099612Z", - "iopub.status.busy": "2026-09-28T11:18:11.098895Z", - "iopub.status.idle": "2026-09-28T11:18:11.303164Z", - "shell.execute_reply": "2026-09-28T11:18:11.300067Z" + "iopub.execute_input": "2026-10-14T23:29:55.453893Z", + "iopub.status.busy": "2026-10-14T23:29:55.453622Z", + "iopub.status.idle": "2026-10-14T23:29:55.699730Z", + "shell.execute_reply": "2026-10-14T23:29:55.698604Z" }, "papermill": { - "duration": 0.216023, - "end_time": "2026-09-28T11:18:11.307223+00:00", + "duration": 0.260457, + "end_time": "2026-10-14T23:29:55.700267+00:00", "exception": false, - "start_time": "2026-09-28T11:18:11.091200+00:00", + "start_time": "2026-10-14T23:29:55.439810+00:00", "status": "completed" }, "tags": [] @@ -785,13 +791,13 @@ "text": [ "Features dediees a un facteur : 46 | autres : 18\n", "Concentration de rang 1 dans le decoupage (4 x 8) :\n", - " atomes dedies : 0.552 +/- 0.095 (controle aleatoire : 0.496 +/- 0.093)\n", - " atomes non dedies : 0.480 +/- 0.047 (controle aleatoire : 0.518 +/- 0.078)\n" + " atomes dedies : 0.553 +/- 0.092 (controle aleatoire : 0.496 +/- 0.093)\n", + " atomes non dedies : 0.481 +/- 0.048 (controle aleatoire : 0.518 +/- 0.078)\n" ] }, { "data": { - "image/png": "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", + "image/png": "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", "text/plain": [ "
" ] @@ -855,10 +861,10 @@ "id": "e1cf91a5", "metadata": { "papermill": { - "duration": 0.009889, - "end_time": "2026-09-28T11:18:11.337850+00:00", + "duration": 0.002834, + "end_time": "2026-10-14T23:29:55.706648+00:00", "exception": false, - "start_time": "2026-09-28T11:18:11.327961+00:00", + "start_time": "2026-10-14T23:29:55.703814+00:00", "status": "completed" }, "tags": [] @@ -870,13 +876,13 @@ "\n", "| Population | Concentration observée | Contrôle aléatoire (même découpage) |\n", "|------------|------------------------|--------------------------------------|\n", - "| atomes dédiés (46) | 0.552 ± 0.095 | 0.496 ± 0.093 |\n", - "| atomes non dédiés (18) | 0.480 ± 0.047 | 0.518 ± 0.078 |\n", + "| atomes dédiés (46) | 0.553 ± 0.092 | 0.496 ± 0.093 |\n", + "| atomes non dédiés (18) | 0.481 ± 0.048 | 0.518 ± 0.078 |\n", "\n", "**Points clés** :\n", "\n", "1. L'écart entre les moyennes des atomes dédiés et du contrôle aléatoire (~0.06) est du même ordre que la dispersion des deux distributions (~0.09) ; les atomes dédiés sont *légèrement au-dessus* de l'aléatoire, les non-dédiés *légèrement en dessous* — les deux mouvements sont modestes et les distributions larges. **Aucune inférence formelle n'est revendiquée** : c'est un contraste descriptif à effectifs bornés, pas un test. Les ± rapportés sont les écarts-types des populations, pas des erreurs-types.\n", - "2. Autrement dit : le dictionnaire SAE *trouve* les facteurs (cellule statistique : 34/40 au-dessus du null apparié) mais la forme géométrique de ces atomes n'est **pas** celle que prédirait naïvement la lecture TPR. Le résultat négatif est informatif en soi — il sépare la *sélectivité* (l'atome sait *quel* facteur) de la *géométrie* (l'atome ne le code pas *comme* un produit tensoriel rôle x filler).\n", + "2. Autrement dit : le dictionnaire SAE *trouve* les facteurs (cellule statistique : 35/40 au-dessus du null apparié) mais la forme géométrique de ces atomes n'est **pas** celle que prédirait naïvement la lecture TPR. Le résultat négatif est informatif en soi — il sépare la *sélectivité* (l'atome sait *quel* facteur) de la *géométrie* (l'atome ne le code pas *comme* un produit tensoriel rôle x filler).\n", "3. Caveat de découpage : la valeur *absolue* dépend du découpage imposé (4 x 8, par convention). La comparaison dédié vs aléatoire, elle, est faite *dans le même découpage* — le contraste est interne et ne dépend pas du choix de découpage, contrairement à la valeur absolue.\n", "4. Ce constat vaut pour ce dictionnaire (une graine), ce découpage et cette échelle : rien n'exclut qu'une structure TPR de rang 1 apparaisse à une autre échelle ou dans un autre espace, mais elle n'est pas visible ici." ] @@ -887,16 +893,16 @@ "id": "d3b205ba", "metadata": { "execution": { - "iopub.execute_input": "2026-09-28T11:18:11.364529Z", - "iopub.status.busy": "2026-09-28T11:18:11.363683Z", - "iopub.status.idle": "2026-09-28T11:18:11.606688Z", - "shell.execute_reply": "2026-09-28T11:18:11.605730Z" + "iopub.execute_input": "2026-10-14T23:29:55.719774Z", + "iopub.status.busy": "2026-10-14T23:29:55.719462Z", + "iopub.status.idle": "2026-10-14T23:30:05.436611Z", + "shell.execute_reply": "2026-10-14T23:30:05.436187Z" }, "papermill": { - "duration": 0.256485, - "end_time": "2026-09-28T11:18:11.607446+00:00", + "duration": 9.72783, + "end_time": "2026-10-14T23:30:05.440265+00:00", "exception": false, - "start_time": "2026-09-28T11:18:11.350961+00:00", + "start_time": "2026-10-14T23:29:55.712435+00:00", "status": "completed" }, "tags": [] @@ -906,17 +912,17 @@ "name": "stdout", "output_type": "stream", "text": [ - "features dediees a pos0 : [2, 4, 9, 16, 27, 30, 41, 50] | features actives sur val : 49/64\n", + "features dediees a pos0 : [2, 4, 9, 27, 30, 41, 42, 50] | features actives sur val : 50/64\n", "\n", "lecture acc delta\n", "originale (H) 0.817 0.000\n", "TPR par echantillon (rang effectif 4, decoupage 4x8) 0.817 0.000\n", "TPR permutation de blocs (lignes 0<->1) 0.346 0.471\n", - "SAE recon val (encode-only) 0.709 0.108\n", - "SAE ablation pos0 (8 feats dediees) 0.532 0.285\n", - "SAE ablation aleatoire quelconque (8 feats) 0.563 0.254\n", + "SAE recon val (encode-only) 0.713 0.104\n", + "SAE ablation pos0 (8 feats dediees) 0.523 0.294\n", + "SAE ablation aleatoire quelconque (8 feats) 0.573 0.244\n", "SAE ablation aleatoire PARMI ACTIVES (20 tirages) 0.617 0.200\n", - " dispersion du controle apparie : +/- 0.038 (ecart-type descriptif sur 20 tirages)\n" + " dispersion du controle apparie : +/- 0.029 (ecart-type descriptif sur 20 tirages)\n" ] } ], @@ -990,10 +996,10 @@ "id": "962278f3", "metadata": { "papermill": { - "duration": 0.004661, - "end_time": "2026-09-28T11:18:11.617268+00:00", + "duration": 0.00598, + "end_time": "2026-10-14T23:30:05.456557+00:00", "exception": false, - "start_time": "2026-09-28T11:18:11.612607+00:00", + "start_time": "2026-10-14T23:30:05.450577+00:00", "status": "completed" }, "tags": [] @@ -1008,16 +1014,16 @@ "| originale (H) | 0.817 | — | — |\n", "| TPR par échantillon (rang effectif 4) | 0.817 | 0.000 | fidélité : reconstruction exacte (R² = 1) — contrôle de passage |\n", "| permutation des blocs 0<->1 du découpage 4 x 8 | 0.346 | 0.471 | contrôle structurel des coordonnées — **pas** une intervention sur des rôles |\n", - "| SAE recon val (encode-only) | 0.709 | 0.108 | fidélité du dictionnaire à val |\n", - "| ablation SAE des 8 features dédiées à pos0 | 0.532 | 0.285 | ablation dirigée (sémantique) |\n", - "| ablation de 8 features au hasard | 0.563 | 0.254 | un tirage quelconque |\n", - "| ablation de 8 features PARMI ACTIVES (20 tirages) | 0.617 ± 0.038 | 0.200 | zéro honnête de l'échelle (dispersion descriptive) |\n", + "| SAE recon val (encode-only) | 0.713 | 0.104 | fidélité du dictionnaire à val |\n", + "| ablation SAE des 8 features dédiées à pos0 | 0.523 | 0.294 | ablation dirigée (sémantique) |\n", + "| ablation de 8 features au hasard | 0.573 | 0.244 | un tirage quelconque |\n", + "| ablation de 8 features PARMI ACTIVES (20 tirages) | 0.617 ± 0.029 | 0.200 | zéro honnête de l'échelle (dispersion descriptive) |\n", "\n", "**Points clés** :\n", "\n", "1. **Permutation de blocs** : échanger les deux premiers blocs de coordonnées du reshape 4 x 8 fait chuter l'accuracy de 0.817 à 0.346 (delta 0.47). Lecture : les coordonnées de `H`, telles que le décodeur les lit, ne sont **pas interchangeables** — leur ré-agencement détruit l'information. C'est cohérent avec une organisation fine des coordonnées, mais cela **ne démontre aucune structure de rôles apprise** : le découpage 4 x 8 est une convention, et permuter des lignes d'un reshape n'est pas une intervention sur des rôles — c'est un **contrôle structurel** que l'on mesure pour situer l'échelle des perturbations. Il serait illégitime d'en tirer « les rôles TPR portent la fonction » ; il est légitime d'en tirer « le décodeur lit des coordonnées spécifiques avec des significations spécifiques ».\n", - "2. **Ablation dirigée SAE** : supprimer les 8 features dédiées à la position 0 coûte 0.285 d'accuracy (0.817 -> 0.532), plus que la référence appariée en activité (0.617 ± 0.038 ; excédent d'environ 0.085, soit environ deux fois la dispersion des 20 tirages). C'est un **effet conditionnel exploratoire** : un sous-ensemble sémantiquement choisi ruine la lecture plus qu'un tirage de même cardinal parmi les features actives, dans les conditions mesurées (ce dictionnaire, ce découpage de lecture, ce décodeur). La dispersion rapportée est celle des tirages — descriptive, pas une erreur-type inférentielle.\n", - "3. Le contrôle « au hasard quelconque » (0.563) est *plus bas* que le contrôle apparié en activité (0.617) : un tirage uniforme tape parfois des features très actives ou rares qui coûtent plus cher. C'est pourquoi le zéro honnête est le tirage **parmi les actives**, et l'on rapporte les deux pour que le lecteur voie la sensibilité au protocole de contrôle.\n", + "2. **Ablation dirigée SAE** : supprimer les 8 features dédiées à la position 0 coûte 0.294 d'accuracy (0.817 -> 0.523), plus que la référence appariée en activité (0.617 ± 0.029 ; excédent d'environ 0.094, soit environ 3.2 fois la dispersion des 20 tirages). C'est un **effet conditionnel exploratoire** : un sous-ensemble sémantiquement choisi ruine la lecture plus qu'un tirage de même cardinal parmi les features actives, dans les conditions mesurées (ce dictionnaire, ce découpage de lecture, ce décodeur). La dispersion rapportée est celle des tirages — descriptive, pas une erreur-type inférentielle.\n", + "3. Le contrôle « au hasard quelconque » (0.573) est *plus bas* que le contrôle apparié en activité (0.617) : un tirage uniforme tape parfois des features très actives ou rares qui coûtent plus cher. C'est pourquoi le zéro honnête est le tirage **parmi les actives**, et l'on rapporte les deux pour que le lecteur voie la sensibilité au protocole de contrôle.\n", "4. La position 0, sous le protocole BOS, est la cible produite *en premier*, à partir de BOS et de `H` seuls : c'est le cas le plus pur où la cible ne peut venir que de `H`. Sa perte sous ablation est donc particulièrement lisible — l'information de pos0 est portée, au moins en partie, par les features que le dictionnaire lui dédie.\n", "5. Ce que ce contraste **ne dit pas** : il ne localise pas un « facteur » à une coordonnée (l'ablation retire des atomes *appris*, pas des rôles) et l'ordre de grandeur des deltas dépend du découpage conventionnel et du décodeur ; c'est une échelle de coûts fonctionnels, pas une localisation causale." ] @@ -1028,16 +1034,16 @@ "id": "71465df5", "metadata": { "execution": { - "iopub.execute_input": "2026-09-28T11:18:11.628556Z", - "iopub.status.busy": "2026-09-28T11:18:11.628203Z", - "iopub.status.idle": "2026-09-28T11:18:11.646240Z", - "shell.execute_reply": "2026-09-28T11:18:11.645112Z" + "iopub.execute_input": "2026-10-14T23:30:05.473902Z", + "iopub.status.busy": "2026-10-14T23:30:05.473681Z", + "iopub.status.idle": "2026-10-14T23:30:06.808567Z", + "shell.execute_reply": "2026-10-14T23:30:06.807258Z" }, "papermill": { - "duration": 0.024727, - "end_time": "2026-09-28T11:18:11.647061+00:00", + "duration": 1.345011, + "end_time": "2026-10-14T23:30:06.809464+00:00", "exception": false, - "start_time": "2026-09-28T11:18:11.622334+00:00", + "start_time": "2026-10-14T23:30:05.464453+00:00", "status": "completed" }, "tags": [] @@ -1049,11 +1055,11 @@ "text": [ "lecture acc note\n", "originale (H) 0.817 -\n", - "SAE : dictionnaire train, codes val 0.709 FVU val = 0.1032\n", + "SAE : dictionnaire train, codes val 0.713 FVU val = 0.1033\n", "TPR partagee (R_mean, F_mean de train) 0.135 une seule reconstruction pour toutes les sequences\n", "TPR par echantillon (rang effectif 4) 0.817 factorisee sur l'echantillon lui-meme\n", "\n", - "FVU SAE : train = 0.0746 | val = 0.1032 | ecart = +0.0285\n" + "FVU SAE : train = 0.0744 | val = 0.1033 | ecart = +0.0289\n" ] } ], @@ -1082,10 +1088,10 @@ "id": "a138437a", "metadata": { "papermill": { - "duration": 0.00465, - "end_time": "2026-09-28T11:18:11.657277+00:00", + "duration": 0.003338, + "end_time": "2026-10-14T23:30:06.816229+00:00", "exception": false, - "start_time": "2026-09-28T11:18:11.652627+00:00", + "start_time": "2026-10-14T23:30:06.812891+00:00", "status": "completed" }, "tags": [] @@ -1097,13 +1103,13 @@ "\n", "| Lecture | Acc (val) | Note |\n", "|---------|-----------|------|\n", - "| SAE : dictionnaire entraîné sur train, codes val | 0.709 | FVU val = 0.1032 (écart train +0.0285) |\n", + "| SAE : dictionnaire entraîné sur train, codes val | 0.713 | FVU val = 0.1033 (écart train +0.0289) |\n", "| TPR partagée (R̄, F̄ de train) | 0.135 | une seule reconstruction pour toutes les séquences |\n", "| TPR par échantillon (rang effectif 4) | 0.817 | factorise l'échantillon lui-même — plafond local, pas un transfert |\n", "\n", "**Points clés** :\n", "\n", - "1. Le vrai transfert train vers val est celui du **dictionnaire SAE** : appris sur les 800 états de train, appliqué sans ré-entraînement aux 200 états de val — l'accuracy ne tombe que de 0.817 à 0.709, et la FVU ne se dégrade que de +0.03. Le dictionnaire capte des directions qui survivent au changement d'échantillons : c'est une forme de généralisation *mesurée*.\n", + "1. Le vrai transfert train vers val est celui du **dictionnaire SAE** : appris sur les 800 états de train, appliqué sans ré-entraînement aux 200 états de val — l'accuracy ne tombe que de 0.817 à 0.713, et la FVU ne se dégrade que de +0.0289. Le dictionnaire capte des directions qui survivent au changement d'échantillons : c'est une forme de généralisation *mesurée*.\n", "2. La TPR *partagée* s'effondre (0.135, quasi le hasard) : une seule matrice de rôles/fillers moyens ne peut pas servir de reconstruction à toutes les séquences — la TPR utile est nécessairement *par échantillon*, donc sans objet appris transférable. C'est la différence structurelle la plus nette entre les deux lectures à cette échelle : le dictionnaire **apprend un vocabulaire réutilisable**, la TPR découvre une factorisation **à la demande**.\n", "3. Le 0.817 « par échantillon » n'est pas un résultat de généralisation : à rang effectif 4 en 4 x 8, la SVD est exacte (R² = 1), donc la reconstruction est l'identité — c'est le plafond de ce que le décodage d'un `H` exact permet, utile comme point haut de l'échelle.\n", "4. Bornage : « généralisation » ici = même distribution (nouvelles séquences de copy), pas un transfert de tâche. Un transfert vers une autre tâche (copy -> reverse) est justement l'Exercice 1." @@ -1114,10 +1120,10 @@ "id": "8fd3f27d", "metadata": { "papermill": { - "duration": 0.004673, - "end_time": "2026-09-28T11:18:11.669110+00:00", + "duration": 0.00535, + "end_time": "2026-10-14T23:30:06.829786+00:00", "exception": false, - "start_time": "2026-09-28T11:18:11.664437+00:00", + "start_time": "2026-10-14T23:30:06.824436+00:00", "status": "completed" }, "tags": [] @@ -1127,15 +1133,15 @@ "\n", "**Question posée** : les deux lectures concurrentes des mêmes états cachés — TPR imposée (SL-13) vs dictionnaire SAE appris (série ICT) — voient-elles la même structure dans `H` ?\n", "\n", - "**Ce qui est mesuré** (tout en CPU, quelques secondes par cellule, tout est seedé et reproductible) :\n", + "**Ce qui est mesuré** (tout en CPU, sans GPU, tout est seedé et reproductible à environnement fixé) :\n", "\n", "| Question | Résultat mesuré | Statut |\n", "|----------|------------------|--------|\n", - "| Le dictionnaire porte-t-il les facteurs (position, symbole) ? | 34/40 facteurs dédiés à z > 5 contre null apparié ; AUC médiane 0.658 | descriptif : oui, au seuil choisi |\n", - "| Les atomes dédiés sont-ils proches du rang 1 (prédiction TPR) ? | dédiés 0.552 ± 0.095 vs aléatoire 0.496 ± 0.093 — écart du même ordre que la dispersion | **non soutenu** : contraste exploratoire, sans inférence |\n", + "| Le dictionnaire porte-t-il les facteurs (position, symbole) ? | 35/40 facteurs dédiés à z > 5 contre null apparié ; AUC médiane 0.658 | descriptif : oui, au seuil choisi |\n", + "| Les atomes dédiés sont-ils proches du rang 1 (prédiction TPR) ? | dédiés 0.553 ± 0.092 vs aléatoire 0.496 ± 0.093 — écart du même ordre que la dispersion | **non soutenu** : contraste exploratoire, sans inférence |\n", "| Les coordonnées de `H` sont-elles interchangeables ? | permutation de blocs : 0.817 -> 0.346 (delta 0.47) | non — mais c'est un contrôle structurel, pas une preuve de rôles |\n", - "| L'ablation SAE dirigée coûte-t-elle plus que son contrôle ? | pos0 : delta 0.285 vs référence appariée 0.617 ± 0.038 — excédent ~0.085 | effet conditionnel exploratoire |\n", - "| Le transfert train -> val ? | SAE 0.709 (FVU +0.03) ; TPR partagée 0.135 | transfert mesuré côté SAE ; effondrement côté TPR |\n", + "| L'ablation SAE dirigée coûte-t-elle plus que son contrôle ? | pos0 : delta 0.294 vs référence appariée 0.617 ± 0.029 — excédent ~0.094 | effet conditionnel exploratoire |\n", + "| Le transfert train -> val ? | SAE 0.713 (FVU +0.0289) ; TPR partagée 0.135 | transfert mesuré côté SAE ; effondrement côté TPR |\n", "\n", "**Ce que l'on retient** :\n", "\n", @@ -1153,10 +1159,10 @@ "id": "b61e4dc4", "metadata": { "papermill": { - "duration": 0.004623, - "end_time": "2026-09-28T11:18:11.678426+00:00", + "duration": 0.008059, + "end_time": "2026-10-14T23:30:06.844495+00:00", "exception": false, - "start_time": "2026-09-28T11:18:11.673803+00:00", + "start_time": "2026-10-14T23:30:06.836436+00:00", "status": "completed" }, "tags": [] @@ -1171,15 +1177,57 @@ "**Indice** : surveillez la ligne « pos 0 » de la heatmap de sélectivité. En copy, la position 0 est la première cible produite (le décodeur n'a vu que BOS et `H`) ; en reverse, c'est la *dernière* cible. Si la couverture des positions change du côté SAE mais pas du côté TPR (ou l'inverse), vous tenez une différence structurale entre les deux lectures — dites laquelle et pourquoi." ] }, + { + "cell_type": "code", + "execution_count": 10, + "id": "df08bbf7", + "metadata": { + "execution": { + "iopub.execute_input": "2026-10-14T23:30:06.861738Z", + "iopub.status.busy": "2026-10-14T23:30:06.861500Z", + "iopub.status.idle": "2026-10-14T23:30:06.864958Z", + "shell.execute_reply": "2026-10-14T23:30:06.864428Z" + }, + "papermill": { + "duration": 0.013208, + "end_time": "2026-10-14T23:30:06.865670+00:00", + "exception": false, + "start_time": "2026-10-14T23:30:06.852462+00:00", + "status": "completed" + }, + "tags": [] + }, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Exercice a completer\n" + ] + } + ], + "source": [ + "# Exercice 1 -- contraste sur la tache reverse (consigne ci-dessus)\n", + "# TODO etudiant : construire la tache reverse, de meme signature que make_copy_task.\n", + "\n", + "\n", + "def make_reverse_task(seq_len=SEQ_LEN, n_samples=200, seed=0):\n", + " \"\"\"Tache reverse : retourne une liste de paires (entree, cible).\"\"\"\n", + " return None\n", + "\n", + "\n", + "print(\"Exercice a completer\")" + ] + }, { "cell_type": "markdown", "id": "3abb7fc2", "metadata": { "papermill": { - "duration": 0.004456, - "end_time": "2026-09-28T11:18:11.687675+00:00", + "duration": 0.006193, + "end_time": "2026-10-14T23:30:06.874622+00:00", "exception": false, - "start_time": "2026-09-28T11:18:11.683219+00:00", + "start_time": "2026-10-14T23:30:06.868429+00:00", "status": "completed" }, "tags": [] @@ -1194,15 +1242,60 @@ "**Indice** : testez aussi la variante « douce » — remplacer les codes ablatés par leur moyenne d'entraînement au lieu de zéro (`z2[:, feats] = z_train[:, feats].mean(axis=0)`). Si la mort douce coûte moins cher que la mort brutale, une partie de la chute mesurée était un artefact d'extrapolation hors distribution du décodeur, pas un portage du facteur. Et gardez en tête la limite du découpage : permuter des blocs de coordonnées dans un reshape arbitraire n'est PAS une intervention sur des rôles appris — construire un alignement de rôles appris (par exemple en regroupant les fillers par échantillon puis en appariant les centroïdes aux positions) serait la vraie extension." ] }, + { + "cell_type": "code", + "execution_count": 11, + "id": "0caa8bbf", + "metadata": { + "execution": { + "iopub.execute_input": "2026-10-14T23:30:06.888375Z", + "iopub.status.busy": "2026-10-14T23:30:06.888090Z", + "iopub.status.idle": "2026-10-14T23:30:06.892320Z", + "shell.execute_reply": "2026-10-14T23:30:06.891386Z" + }, + "papermill": { + "duration": 0.012383, + "end_time": "2026-10-14T23:30:06.892809+00:00", + "exception": false, + "start_time": "2026-10-14T23:30:06.880426+00:00", + "status": "completed" + }, + "tags": [] + }, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Exercice a completer\n" + ] + } + ], + "source": [ + "# Exercice 2 -- ablation dirigee par symbole, brutale puis douce\n", + "# TODO etudiant : label \"c in s\" (toutes positions), features dediees, ablation.\n", + "\n", + "\n", + "def ablation_par_symbole(c, z_val, z_train, mode=\"brutale\"):\n", + " \"\"\"Reconstruction de H apres ablation des features dediees au symbole c.\n", + "\n", + " mode : \"brutale\" (codes ablates a zero) ou \"douce\" (codes ablates a la moyenne d'entrainement).\n", + " \"\"\"\n", + " return None\n", + "\n", + "\n", + "print(\"Exercice a completer\")" + ] + }, { "cell_type": "markdown", "id": "1b68a5a7", "metadata": { "papermill": { - "duration": 0.004361, - "end_time": "2026-09-28T11:18:11.696964+00:00", + "duration": 0.008538, + "end_time": "2026-10-14T23:30:06.904569+00:00", "exception": false, - "start_time": "2026-09-28T11:18:11.692603+00:00", + "start_time": "2026-10-14T23:30:06.896031+00:00", "status": "completed" }, "tags": [] @@ -1217,15 +1310,57 @@ "**Indice** : l'organe documente un **plateau de mélange** — sur un corpus exactement reconstructible, l'entraînement full-batch retrouve ~17/24 atomes propres et laisse les autres en combinaisons. Si la fraction d'atomes dédiés augmente avec la sur-complétude mais que leur concentration de rang 1 *stagne ou baisse*, les atomes supplémentaires capturent les facteurs en les *mélangeant* — et vous pourrez dire laquelle des deux lectures (structure imposée TPR vs dictionnaire libre) vieillit le mieux quand la capacité croît." ] }, + { + "cell_type": "code", + "execution_count": 12, + "id": "0f5b082b", + "metadata": { + "execution": { + "iopub.execute_input": "2026-10-14T23:30:06.917741Z", + "iopub.status.busy": "2026-10-14T23:30:06.917409Z", + "iopub.status.idle": "2026-10-14T23:30:06.921451Z", + "shell.execute_reply": "2026-10-14T23:30:06.920919Z" + }, + "papermill": { + "duration": 0.009709, + "end_time": "2026-10-14T23:30:06.922132+00:00", + "exception": false, + "start_time": "2026-10-14T23:30:06.912423+00:00", + "status": "completed" + }, + "tags": [] + }, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Exercice a completer\n" + ] + } + ], + "source": [ + "# Exercice 3 -- sur-completude : 32, 64, 128 atomes a k fixe\n", + "# TODO etudiant : entrainer le SAE a chaque cardinalite, puis mesurer FVU, fraction dediee et concentration de rang 1.\n", + "\n", + "\n", + "def balayage_sur_completude(n_features_list=(32, 64, 128), k=K_SPARSE, seed=0):\n", + " \"\"\"Pour chaque cardinalite : FVU train/val, fraction d'atomes dedies, concentration de rang 1 moyenne.\"\"\"\n", + " return None\n", + "\n", + "\n", + "print(\"Exercice a completer\")" + ] + }, { "cell_type": "markdown", "id": "3e7182bb", "metadata": { "papermill": { - "duration": 0.00489, - "end_time": "2026-09-28T11:18:11.706538+00:00", + "duration": 0.007947, + "end_time": "2026-10-14T23:30:06.936518+00:00", "exception": false, - "start_time": "2026-09-28T11:18:11.701648+00:00", + "start_time": "2026-10-14T23:30:06.928571+00:00", "status": "completed" }, "tags": [] @@ -1266,18 +1401,18 @@ "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", - "version": "3.13.14" + "version": "3.13.16" }, "papermill": { "default_parameters": {}, - "duration": 18.712901, - "end_time": "2026-09-28T11:18:12.777691+00:00", + "duration": 319.77975, + "end_time": "2026-10-14T23:30:08.269467+00:00", "environment_variables": {}, "exception": null, "input_path": "SL-13b-TPR-SAE.ipynb", "output_path": "SL-13b-TPR-SAE.ipynb", "parameters": {}, - "start_time": "2026-09-28T11:17:54.064790+00:00", + "start_time": "2026-10-14T23:24:48.489717+00:00", "version": "2.7.0" } },