Add Quasar research model card and measured evidence

.gitattributes

@@ -33,3 +33,5 @@ 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
+assets/image_comparison.png filter=lfs diff=lfs merge=lfs -text
+assets/motion_comparison.gif filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,277 @@
+---
+language:
+  - en
+tags:
+  - quasar
+  - axiom
+  - pytorch
+  - multimodal
+  - autoregressive
+  - text-generation
+  - text-to-image
+  - text-to-video
+  - research
+  - experimental
+datasets:
+  - HuggingFaceFW/fineweb-edu
+  - HuggingFaceTB/smoltalk
+  - google-research-datasets/mbpp
+  - roneneldan/TinyStories
+  - databricks/databricks-dolly-15k
+  - Team-ACE/ToolACE
+model-index:
+  - name: Quasar
+    results: []
+---
+
+# Quasar — Unified Multimodal Research Prototype
+
+Quasar is a small, custom PyTorch autoregressive model built on an Axiom-1.1
+hybrid recurrence/attention backbone. Text, visual codes, audio codes, triangle
+meshes and tool-call proposals share one vocabulary and model. Frozen codecs
+convert between media and discrete tokens.
+
+**Release status: documentation and evaluation evidence only. No model weights,
+training data, codec weights or runnable inference service are included in this
+repository.** This card describes the preserved v5 model and a completed
+curriculum-research branch as of **21 October 2026**. Separate architecture
+experiments are still being evaluated and are excluded from these results.
+
+**Quality status:** a focused research branch produces recognizable images and
+simple moving squares by reproducing memorized training examples. Unseen image
+generation and unfamiliar motion continuation remain unsuccessful. The research
+checkpoint was withheld from production because most prior-task losses worsened.
+
+## Model details
+
+| Property | Described v5 / curriculum-research model |
+|---|---|
+| Developer | Atlas-AI-research |
+| Total model parameters | 90,247,794, excluding frozen media codecs |
+| Backbone | 12 blocks, width 640; 9 recurrent blocks and 3 causal attention blocks |
+| Attention | 10 heads; full attention at zero-based blocks 3, 7 and 11 |
+| Feed-forward cells | Two dense 640 → 1280 → 640 experts per block; both are evaluated |
+| Normalization / positions | RMSNorm, RoPE and learned visual spatial embeddings |
+| Context | 1,024 shared tokens |
+| Vocabulary | 54,144 shared IDs; 16,384-token text BPE; 32,768 visual codes |
+| Images | Native 64×64 or 128×128; 64 or 256 visual tokens |
+| Video | Up to 8 frames at native 64×64; 64 visual tokens/frame; per-frame codec |
+| Audio interface | Four 1,024-entry DAC codebooks; short, complete utterances, at most 2 seconds |
+| Mesh interface | Quantized coordinates and indexed triangle faces, up to 128 vertices/faces |
+| Implementation | Custom Quasar PyTorch code; not a Transformers `AutoModel` release |
+
+The original text-stage model had 69,808,754 parameters. Vocabulary expansion and
+the later 128px spatial-embedding migration preserved learned weights while
+increasing the model to the size above. Quasar is not an Emu3 language-model
+fine-tune: Emu3 is used as a **frozen pixel codec**, not a generative backbone.
+
+The Switch-style auxiliary term is constant for this two-of-two dense-cell setup;
+it does not demonstrate effective routing regularization. Reported validation CE
+excludes that auxiliary term.
+
+## Checkpoint identity
+
+The measurements below refer to these local artifacts, not to downloadable Hub
+weights. SHA-256 values identify the exact `model.safetensors` files.
+
+| Artifact | Role | SHA-256 |
+|---|---|---|
+| v5 full image continuation | Preserved comparison model | `e20242e9971b6a7cef7ca23a65c25faa70a007807a55a2c0cfe35f4f5682a643` |
+| `joint_repair/final` | Isolated, unpromoted curriculum-research model | `de4846e5e63c0004cf0b6331127ab421d6843f6efde04243c6d5448ab4545240` |
+
+## Intended use and limitations
+
+Intended uses are small-model research, codec/packet interoperability studies,
+controlled experiments in visual conditioning, and investigation of recall,
+generalization and forgetting.
+
+- The model is not validated for dependable factual answers, coding assistance,
+  natural-scene image synthesis, natural video, intelligible generated speech,
+  useful complex 3D assets or reliable tool selection.
+- Tool outputs are schema-checked proposals. The model does not execute tools.
+  Structural validity does not establish correct arguments or safe actions.
+- The inference sampler forces packet boundaries and legal token ranges; those
+  constraints are not evidence of learned semantic competence.
+- Current video training uses **procedurally controlled motion**, not natural
+  footage. Video annotation downloads were not treated as footage.
+- These generation probes call explicit image/video APIs. They do not establish
+  automatic routing of arbitrary chat requests to media generation.
+- Real-photo outputs can reproduce training photographs. Creator attribution and
+  original asset licenses remain relevant to reproduced images.
+- English-biased source selection, small media corpora and synthetic tool data
+  can introduce bias. No comprehensive safety, fairness or privacy audit was run.
+- No user-supplied dataset was included; no user-data path was provided.
+
+## Training data
+
+The admitted corpus combines retained built-in examples and selected external
+sources. A source appearing in a planning catalog does not mean it was trained.
+This list reflects admitted source families in the preserved training cache:
+
+| Domain | Sources and split controls |
+|---|---|
+| Text/chat | Pinned FineWeb-Edu; selected SmolTalk instruction, constraint, rewrite and summarization components; retained TinyStories, Dolly and built-in examples |
+| Basic Python | Official MBPP `full/train`, 374 problems; official validation/test held out |
+| Real photos | Open Images / Localized Narratives; per-image creator identity, eligible license, attribution and JPEG hashes checked; source-group splits retained |
+| Video | Controlled local synthetic motion with source-group splits |
+| Audio | Complete LibriTTS-R utterances ≤2 seconds; speaker-group and official split separation |
+| Meshes | Admitted per-object CC0/CC-BY Objaverse/Cap3D assets; geometry validation and source-object grouping |
+| Tool proposals | Safely parsed ToolACE examples with JSON Schema checks and tool-family/source grouping; no training-time tool execution |
+
+The shared text BPE was fit on training text only and remained frozen. The v5
+cache contains 555,007 records. Checksum, token-range, grammar, context, offsets
+and exact-content/group-leakage validation passed. This is not a claim that all
+semantic near-duplicates have been ruled out.
+
+The focused photo experiment uses only **16 reviewed training photos across
+cat/dog/car/flower**, each with three short captions; 16 validation and 8 test
+photos retain their original source groups. The new motion cache admits 112
+training, 29 validation and 29 test clips after frozen-codec calibration and
+code-sequence duplicate checks. All four colors × four directions are taught.
+
+FineVideo and xLAM originals were access-gated and were not bypassed. Downloaded
+video alternatives supplied annotations, not usable footage. Broader catalog
+entries and raw alternatives are not claimed as trained capabilities.
+
+## Training procedure
+
+The main curriculum used a nominal 300M padded-input text stage, followed by
+approximately 100M each for image, synthetic video, audio/mesh/tool, unified and
+later image-quality stages. The preserved v5 metadata records 800,882,688 cumulative
+padded input tokens, including pilot/history accounting. These are not counts of
+unique examples or supervised output tokens.
+
+The completed media R&D session ran six experiments totaling **5,120 updates,
+20,971,520 padded input tokens and 10,005,264 supervised targets**. Independent
+branches do not all contribute to one checkpoint. The final shared research
+checkpoint adds **6,291,456 inputs along its actual ancestry after v5**: a
+4,194,304-input covered-motion branch, then a 2,097,152-input shared follow-up.
+Its final mixture is 45% focused photos, 35% controlled video generation,
+10% continuation, 5% text and 5% chat.
+
+Research training used one RTX 5070, BF16, AdamW, gradient checkpointing, Inductor,
+static 128/256/512/1024 buckets, and 1,024 padded inputs × 4 accumulation per update.
+Real batch shapes warmed through forward/backward without optimizer steps.
+Peak measured allocated training memory was 1.93 GiB. All six saved final models
+had finite weights, ordinary unwrapped serialization, matching recorded hashes
+and successful GPU reload/evaluation. The session took approximately 95 minutes
+including investigation, preparation, waiting, compilation, training and evaluation.
+
+## Evaluation results
+
+Both models used identical prompts and seeds, the corrected cached visual-position
+inference path, temperature 0.5, top-k 64 and top-p 0.9. Photo subject matches were
+visually inspected. The controlled-motion metric checks object color, shape, size,
+background and direction; generated continuation suffixes are scored separately
+from their supplied prefixes.
+
+| Matched probe | Preserved v5 | Shared research checkpoint |
+|---|---:|---:|
+| Recognizable requested photo category | 0/24 | **23/24** |
+| New-wording photo prompts, subset of above | 0/16 | **15/16** |
+| Closed color/direction motion, joint match | 1/16 | **16/16** |
+| Unseen-trajectory continuation, joint suffix match | 0/8 | **0/8** |
+
+**Recall caveat:** all 24 final research photo samples and all 16 closed-motion
+clips exactly match training visual-code sequences. One cat request selects a
+flower. Novel wording can select an old prototype. These results do not establish
+new-photo synthesis, unseen composition or learned general motion dynamics.
+The motion descriptions are taught classes, even when their reference trajectory
+source groups are held out. No leaderboard or standard benchmark claim is made.
+
+![Same prompts and seeds: original model above, research model below](assets/image_comparison.png)
+
+![Native 64px motion, enlarged for display: source left, research recall right](assets/motion_comparison.gif)
+
+These are representative comparisons; the full photo evaluation includes one
+failed cat request. Display enlargement does not increase native resolution.
+Photo creators, original URLs and license links are recorded in
+[the attribution manifest](assets/attributions.json). Frozen-codec reconstruction
+and display resizing modify the original photographs. Procedural motion is CC0.
+
+### Generalization and retention
+
+On the separate focused unseen-photo set, visual CE worsened from **8.2068 to
+11.7081** on validation and **8.3475 to 11.7548** on test. On an identical common
+52-batch prior-task screen, **11 of 13 losses worsened**:
+
+| Task | v5 CE | Shared research CE |
+|---|---:|---:|
+| Text | 3.9484 | 4.4183 |
+| Chat | 1.6765 | 2.0703 |
+| Coding | 1.7610 | 2.2605 |
+| Image generation | 8.7170 | 11.6703 |
+| Video generation | 5.1221 | 6.7368 |
+| Video continuation | 6.2570 | 8.6859 |
+| Tool proposals | 0.4624 | 0.8044 |
+
+The common screen uses four deterministic held-out batches per task and ordinary
+unweighted CE without auxiliary terms. It is a small paired regression screen,
+not full-corpus semantic evaluation or a statistical significance claim. Lower
+mesh/audio losses on two tasks did not establish improved generated quality.
+The shared research checkpoint was **not promoted**. Exact identities, all 13
+task losses and compact measurement receipts are in [evaluation.json](evaluation.json).
+
+Frozen-codec reconstruction is a separate control: the existing 128px codec
+measured 22.2877 dB validation / 22.4653 dB test content PSNR. This predates the
+research session and is **not generated-image quality**. The earlier 64px
+real-photo probe measured 18.1176 dB; its original aspirational 20 dB gate remained
+false and was not relabeled as passed.
+
+## Local usage requirements
+
+There is no `from_pretrained("Atlas-AI-research/Quasar")` weights release here.
+Running the model requires the matching local Quasar implementation, a compatible
+checkpoint directory with `config.json`, `model.safetensors` and tokenizer metadata,
+and independently available pinned codecs. The project is not a self-contained
+standard Transformers model, and local codec references must be resolved explicitly.
+
+With those prerequisites present, the existing local API is:
+
+```python
+from PIL import Image
+from quasar.inference.extended import ExtendedGenerator
+from quasar.inference.decoding import Sampling
+
+# Point to a verified local checkpoint; this documentation repo has no weights.
+generator = ExtendedGenerator("PATH_TO_LOCAL_QUASAR_CHECKPOINT", "cuda")
+sampling = Sampling(temperature=0.5, top_k=64, top_p=0.9, seed=20261021)
+pixels = generator.generate_image("make a picture of a cat", sampling, resolution=128)
+Image.fromarray(pixels).save("quasar_cat.png")
+frames = generator.generate_video(
+    "a small cyan square moving left on a gray background", 8, sampling
+)
+```
+
+This is a usage example, not a claim of acceptable output for arbitrary prompts.
+GPU use must be exclusive with local training/evaluation workers. No inference
+server, model weights or codec binaries are uploaded by creating this card.
+
+## Licensing and provenance
+
+The local Quasar project does not currently supply a project/weight distribution
+license. This card does not assign one or grant a weights license. No permissive
+license badge is implied by using PyTorch or permissively licensed components.
+
+The existing frozen Emu3 visual tokenizer is pinned at revision
+`c81f916ad371289e205310a7539255e8a9396488` under Apache-2.0. The local Descript DAC
+24 kHz weights are pinned at `6ba020b5ba7d9d8076fb90db7e67f27e31980f6e`; the project's
+official weights statement supplies MIT terms. These are component terms, not a
+Quasar weights license. Dataset, underlying page, photograph and per-object terms
+remain separate; for example, the FineWeb database license does not replace
+underlying web-page rights.
+
+## Reproducibility and next work
+
+The completed R&D snapshot passed 146 CPU tests with 1 skip; the legacy integration
+module that performs additional optimizer training was excluded. This is evidence
+for that snapshot, not a test claim about unfinished architecture changes.
+Cached/full-prefix checks, source-weight preservation, frozen-cache validation,
+all-case visual review and checkpoint identities were recorded independently.
+
+Next work should isolate caption clarity from repetition on matched examples,
+expand held-out position/speed and attribute coverage, and enforce stronger
+prior-task retention gates. Recognizable prototype recall alone is insufficient
+for releasing a general multimodal generator.
+
+Model-card structure follows the [Hugging Face model-card documentation](https://huggingface.co/docs/hub/model-cards).

assets/attributions.json

@@ -0,0 +1,49 @@
+{
+  "photos": [
+    {
+      "display_category": "cat",
+      "source_image_id": "693c3d691f4ebfe7",
+      "creator": "tess/",
+      "source_url": "https://www.flickr.com/photos/davenport/3090854401",
+      "license": "CC-BY-2.0",
+      "license_url": "https://creativecommons.org/licenses/by/2.0/",
+      "original_jpeg_sha256": "d734817c1446a4dd9830e49f66f73fae782af87327623c52aa6066fdb5e8ab08",
+      "changes": "Quasar predicts an exact training-token sequence; frozen Emu3 decoder reconstruction, then resized for display."
+    },
+    {
+      "display_category": "dog",
+      "source_image_id": "caea25eddf298827",
+      "creator": "Jean Ogden Just Chaos Photography",
+      "source_url": "https://www.flickr.com/photos/7326810@N08/2952096557",
+      "license": "CC-BY-2.0",
+      "license_url": "https://creativecommons.org/licenses/by/2.0/",
+      "original_jpeg_sha256": "ec8bbe1c3d31ef7050f63fccaa03b7bf036035f4c3935c2f25a8c653e7b46de1",
+      "changes": "Quasar predicts an exact training-token sequence; frozen Emu3 decoder reconstruction, then resized for display."
+    },
+    {
+      "display_category": "car",
+      "source_image_id": "7781daa1b5e123a4",
+      "creator": "Sarah Stierch",
+      "source_url": "https://www.flickr.com/photos/sarahvain/19078668344",
+      "license": "CC-BY-2.0",
+      "license_url": "https://creativecommons.org/licenses/by/2.0/",
+      "original_jpeg_sha256": "61765131ba62d8abbdeb6607401af86ae38524b798bc756de450de957e0e7b9d",
+      "changes": "Quasar predicts an exact training-token sequence; frozen Emu3 decoder reconstruction, then resized for display."
+    },
+    {
+      "display_category": "flower",
+      "source_image_id": "680c0f775abb28ef",
+      "creator": "Stewart Butterfield",
+      "source_url": "https://www.flickr.com/photos/stewart/433917400",
+      "license": "CC-BY-2.0",
+      "license_url": "https://creativecommons.org/licenses/by/2.0/",
+      "original_jpeg_sha256": "84bfefba6b3af63de7351f669aa9d1a6bbae224739522a7834bb4cba65d1e098",
+      "changes": "Quasar predicts an exact training-token sequence; frozen Emu3 decoder reconstruction, then resized for display."
+    }
+  ],
+  "motion": {
+    "creator": "Local Quasar procedural renderer",
+    "license": "CC0-1.0",
+    "changes": "Quasar-generated frames; enlarged with nearest-neighbor pixels for comparison."
+  }
+}

evaluation.json

@@ -0,0 +1,100 @@
+{
+  "snapshot_date": "2026-10-21",
+  "scope": "Completed v5 versus curriculum-research experiment; separate architecture experiments excluded.",
+  "standard_benchmark_or_leaderboard": false,
+  "generation_sampling": {
+    "temperature": 0.5,
+    "top_k": 64,
+    "top_p": 0.9,
+    "image_seeds": [
+      20261020,
+      20261021
+    ],
+    "motion_seed": 20261021
+  },
+  "generation": {
+    "source_v5": {
+      "weights_sha256": "e20242e9971b6a7cef7ca23a65c25faa70a007807a55a2c0cfe35f4f5682a643",
+      "photo_category_matches": 0,
+      "photo_samples": 24,
+      "novel_wording_matches": 0,
+      "novel_wording_samples": 16,
+      "motion_joint_matches": 1,
+      "motion_samples": 16,
+      "unseen_continuation_joint_matches": 0,
+      "unseen_continuation_samples": 8
+    },
+    "joint_repair": {
+      "weights_sha256": "de4846e5e63c0004cf0b6331127ab421d6843f6efde04243c6d5448ab4545240",
+      "photo_category_matches": 23,
+      "photo_samples": 24,
+      "novel_wording_matches": 15,
+      "novel_wording_samples": 16,
+      "motion_joint_matches": 16,
+      "motion_samples": 16,
+      "unseen_continuation_joint_matches": 0,
+      "unseen_continuation_samples": 8,
+      "exact_training_photo_code_replays": 24,
+      "exact_training_motion_code_replays": 16
+    }
+  },
+  "retention": {
+    "eval_batches_per_task": 4,
+    "batch_sha256": "bfa74f2ea9c7c814894160693a11857cf5fba9971e6e3b70617d857e8c3e2944",
+    "phases": {
+      "source_v5": {
+        "weights_sha256": "e20242e9971b6a7cef7ca23a65c25faa70a007807a55a2c0cfe35f4f5682a643",
+        "losses": {
+          "text": 3.948372427367484,
+          "chat_text": 1.6765498408440793,
+          "coding": 1.7610283871588197,
+          "t2i": 8.717013671945807,
+          "i2t": 2.509420299240159,
+          "t2v": 5.122131407260895,
+          "v2t": 0.8410091984505746,
+          "vcont": 6.256978213787079,
+          "tools": 0.4623708201062148,
+          "t2m": 6.144639953858572,
+          "m2t": 3.6306305448963956,
+          "t2a": 9.011601033198106,
+          "a2t": 3.88449745137116
+        },
+        "parameters": 90247794
+      },
+      "joint_repair": {
+        "weights_sha256": "de4846e5e63c0004cf0b6331127ab421d6843f6efde04243c6d5448ab4545240",
+        "losses": {
+          "text": 4.418278653778919,
+          "chat_text": 2.070340123627516,
+          "coding": 2.2605253159999847,
+          "t2i": 11.670267707399868,
+          "i2t": 2.727402084141879,
+          "t2v": 6.736757338047028,
+          "v2t": 1.6935838811537798,
+          "vcont": 8.68588125705719,
+          "tools": 0.804386016036801,
+          "t2m": 6.04225613219775,
+          "m2t": 4.89992640153417,
+          "t2a": 8.555012301572333,
+          "a2t": 4.078481768739635
+        },
+        "parameters": 90247794
+      }
+    }
+  },
+  "production_promoted": false,
+  "research_session": {
+    "elapsed_seconds": 5729.144976377487,
+    "optimizer_input_tokens": 20971520,
+    "optimizer_steps": 5120,
+    "prior_tasks_regressed": 11,
+    "prior_tasks_probed": 13
+  },
+  "research_supervised_targets": 10005264,
+  "research_shared_ancestry_added_inputs": 6291456,
+  "notes": [
+    "Photo/clip matches are narrow recall probes.",
+    "Successful research outputs reproduce training visual-token sequences.",
+    "Four held-out batches/task are a small paired screen, not a full semantic evaluation."
+  ]
+}