Org card in the corx-labs.com style; five technical reports

.gitattributes

@@ -34,3 +34,14 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 banner.png filter=lfs diff=lfs merge=lfs -text
+banner-site.png filter=lfs diff=lfs merge=lfs -text
+papers/corx1.5-technical-report.pdf filter=lfs diff=lfs merge=lfs -text
+papers/corx1.5-technical-report.png filter=lfs diff=lfs merge=lfs -text
+papers/corx3.8-27b-model-report.pdf filter=lfs diff=lfs merge=lfs -text
+papers/corx3.8-27b-model-report.png filter=lfs diff=lfs merge=lfs -text
+papers/helix-technical-report.pdf filter=lfs diff=lfs merge=lfs -text
+papers/helix-technical-report.png filter=lfs diff=lfs merge=lfs -text
+papers/naturev1-hurricane-physics-report.pdf filter=lfs diff=lfs merge=lfs -text
+papers/naturev1-hurricane-physics-report.png filter=lfs diff=lfs merge=lfs -text
+papers/tristream-svs-technical-report.pdf filter=lfs diff=lfs merge=lfs -text
+papers/tristream-svs-technical-report.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -5,61 +5,76 @@ colorFrom: green
 colorTo: yellow
 sdk: static
 pinned: false
-short_description: AI models built from the ground up, in Jamaica.
+short_description: Jamaica's independent AI research lab.
 ---
 
-<img src="https://huggingface.co/spaces/CorxLabs/README/resolve/main/banner.png" alt="Corx Labs: AI models built from the ground up, in Jamaica" width="100%">
-
-**Corx Labs** is an independent AI lab in Westmoreland, Jamaica. We design, train and verify our own models end to
-end: architecture, tokenizer, training pipeline and evaluation, most of it on a single GPU.
-
-## 🗂️ Collections
-
-| | Collection | Inside |
-|:-:|---|---|
-| 💬 | [**Text Models**](https://huggingface.co/collections/CorxLabs/text-models-6ad400811d7073d767ebe7f1) | CorX language models, from 160M to 27B parameters, Jamaican Patois and English |
-| 🎤 | [**SVS Models**](https://huggingface.co/collections/CorxLabs/svs-models-6ad40082a6e143550bd7aaa6) | Singing-voice synthesis built on how a voice is physically made |
-| 🌀 | [**Weather Models**](https://huggingface.co/collections/CorxLabs/weather-models-6ad40082a150e48c970f176a) | Global weather and hurricane forecasting, scored on seasons it never trained on |
-
-## 💬 Text
-
-- [**CorX3.8-27B**](https://huggingface.co/Sigmandndnns/CorX3.8-27B): a 27B conversational model that speaks
-  Jamaican Patois by default and switches to English when you do. Fully merged; load it and go.
-- [**CorX1.5-160M**](https://huggingface.co/Sigmandndnns/CorX1.5-160M): a ~158M-parameter language model whose
-  architecture, tokenizer, pretraining and fine-tuning were all written and trained from scratch on one GPU.
-  ([v2 run](https://huggingface.co/Sigmandndnns/CorX1.5-160M-v2))
-
-## 🎤 Singing voice
-
-- [**TriStream-SVS-300M**](https://huggingface.co/Sigmandndnns/TriStream-SVS-300M): models a singing voice the
-  way the body makes it, as three streams: the **glottal source** (pitch), the **vocal-tract filter** (who is
-  singing, and which vowel) and **turbulent noise** (breath and fricatives).
-
-## 🌀 Weather and hurricanes
-
-- [**NatureV1**](https://huggingface.co/Sigmandndnns/NatureV1-500): an 89M-parameter model of the atmosphere on
-  [iHELIX](https://pypi.org/project/ihelix/), an attention engine that works on the sphere rather than a flat grid.
-  It forecasts global weather to five days and, for tropical cyclones, track, intensity, rapid intensification and
-  landfall, each with an uncertainty.
-- Its hurricane physics, with **no neural network**, beats straight-line extrapolation and CLIPER at every lead
-  from 6 to 120 hours on 4,447 forecasts of 407 storms in seasons it was never fitted on (2018, 2020-2022):
-
-  | Track error (km) | 24 h | 48 h | 72 h | 120 h |
-  |---|:-:|:-:|:-:|:-:|
-  | Extrapolation | 179 | 442 | 751 | 1,417 |
-  | CLIPER | 163 | 378 | 589 | 914 |
-  | **NatureV1 physics** | **127** | **301** | **515** | **884** |
-
-  Global AI models such as GraphCast and Pangu-Weather are still far better beyond 12 hours. Distilling them into
-  NatureV1 is the next step, and every number on the model card says what it was measured on.
-
-```bash
-pip install naturev1 ihelix
-```
-
-## 🔗 Links
-
-[Website](https://www.corx-labs.com/) · [Developers](https://www.corx-labs.com/developers/) ·
-[naturev1 on PyPI](https://pypi.org/project/naturev1/) · [ihelix on PyPI](https://pypi.org/project/ihelix/)
-
-<p align="center"><sub>Built in Jamaica 🇯🇲</sub></p>
+<img src="https://huggingface.co/spaces/CorxLabs/README/resolve/main/banner-site.png" alt="CorX Labs. Architecting the future, from the ground up." width="100%">
+
+**CorX Labs is Jamaica's independent AI research lab.** We build Jamaican AI from the ground up: five open models so
+far, from CorX3.8, a 27-billion-parameter Jamaican Patois assistant, to NatureV1, a weather and hurricane model. Most
+were trained from random weights on a single GPU, and where one is a fine-tune, we say so.
+
+| Models built | Released with weights | Technical reports | Licence |
+|:-:|:-:|:-:|:-:|
+| **5** | **4** | **5** | **Apache-2.0** |
+
+<img src="https://huggingface.co/spaces/CorxLabs/README/resolve/main/rule.png" width="100%" height="4" alt="">
+
+### Models
+
+| Model | What it is | Size | Built from | Report |
+|---|---|:-:|---|:-:|
+| [**CorX3.8-27B**](https://huggingface.co/CorxLabs/CorX3.8-27B) | Jamaican Patois conversational assistant. Patois by default, English on request. | 27B | Fine-tune of Qwen3.8-27B, fully merged | [03](https://huggingface.co/spaces/CorxLabs/README/resolve/main/papers/corx3.8-27b-model-report.pdf) |
+| [**CorX1.5**](https://huggingface.co/CorxLabs/CorX1.5-160M) | Decoder-only language model with learned role embeddings | 157.8M | From scratch, tokenizer included | [01](https://huggingface.co/spaces/CorxLabs/README/resolve/main/papers/corx1.5-technical-report.pdf) |
+| [**TriStream-SVS**](https://huggingface.co/CorxLabs/TriStream-SVS-300M) | Singing-voice synthesis: pitch, timbre and breath in three separate streams | 321.8M | From scratch | [02](https://huggingface.co/spaces/CorxLabs/README/resolve/main/papers/tristream-svs-technical-report.pdf) |
+| [**NatureV1**](https://huggingface.co/CorxLabs/NatureV1-500) | Probabilistic weather and hurricane model, attention in kilometres on the sphere | 89.45M | From scratch, on the iHELIX engine | [05](https://huggingface.co/spaces/CorxLabs/README/resolve/main/papers/naturev1-hurricane-physics-report.pdf) |
+| [**HELIX**](https://pypi.org/project/helix-lm/) | Linear-scaling sequence architecture that keeps exact long-range recall | configurable | From scratch; no trained checkpoints | [04](https://huggingface.co/spaces/CorxLabs/README/resolve/main/papers/helix-technical-report.pdf) |
+
+Browse by collection: [Text](https://huggingface.co/collections/CorxLabs/text-models-6ad400811d7073d767ebe7f1) ·
+[Singing voice](https://huggingface.co/collections/CorxLabs/svs-models-6ad40082a6e143550bd7aaa6) ·
+[Weather](https://huggingface.co/collections/CorxLabs/weather-models-6ad40082a150e48c970f176a)
+
+### Papers
+
+<table>
+<tr>
+<td align="center" width="20%"><a href="https://huggingface.co/spaces/CorxLabs/README/resolve/main/papers/corx1.5-technical-report.pdf"><img src="https://huggingface.co/spaces/CorxLabs/README/resolve/main/papers/corx1.5-technical-report.png" alt="CorX1.5 technical report"></a></td>
+<td align="center" width="20%"><a href="https://huggingface.co/spaces/CorxLabs/README/resolve/main/papers/tristream-svs-technical-report.pdf"><img src="https://huggingface.co/spaces/CorxLabs/README/resolve/main/papers/tristream-svs-technical-report.png" alt="TriStream-SVS technical report"></a></td>
+<td align="center" width="20%"><a href="https://huggingface.co/spaces/CorxLabs/README/resolve/main/papers/corx3.8-27b-model-report.pdf"><img src="https://huggingface.co/spaces/CorxLabs/README/resolve/main/papers/corx3.8-27b-model-report.png" alt="CorX3.8-27B model report"></a></td>
+<td align="center" width="20%"><a href="https://huggingface.co/spaces/CorxLabs/README/resolve/main/papers/helix-technical-report.pdf"><img src="https://huggingface.co/spaces/CorxLabs/README/resolve/main/papers/helix-technical-report.png" alt="HELIX technical report"></a></td>
+<td align="center" width="20%"><a href="https://huggingface.co/spaces/CorxLabs/README/resolve/main/papers/naturev1-hurricane-physics-report.pdf"><img src="https://huggingface.co/spaces/CorxLabs/README/resolve/main/papers/naturev1-hurricane-physics-report.png" alt="NatureV1 hurricane physics report"></a></td>
+</tr>
+</table>
+
+1. **CorX1.5: a 158M-parameter language model trained from random weights on one GPU.** The architecture, the
+   three-stage training, and the memory arithmetic that fits three billion tokens through one card.
+2. **TriStream-SVS: building the source-filter model of the voice into a singing synthesiser.** Three streams that
+   cannot share information until a late fusion trunk, with zero-shot cloning and a 32-step flow-matching decoder.
+3. **CorX3.8-27B: giving an open 27B model a Jamaican voice.** What a rank-64 LoRA on 19 million tokens changed, and
+   what it did not.
+4. **HELIX: separating memory capacity from bandwidth in a linear-scaling sequence model.** N<sup>1.00</sup> attention
+   pairs, and recall that ties full attention while the same model without its index falls to chance.
+5. **Hurricane tracks without a neural network.** A physics consensus that beats extrapolation and CLIPER at every lead
+   on 4,447 forecasts from four seasons it never saw.
+
+### Measured, not promised
+
+| Hurricane track error (km), 2018 and 2020-2022, all basins | 24 h | 48 h | 72 h | 120 h |
+|---|:-:|:-:|:-:|:-:|
+| Straight-line extrapolation (strongest per lead) | 179 | 442 | 751 | 1,417 |
+| CLIPER | 163 | 378 | 589 | 914 |
+| **NatureV1 hurricane physics, no neural network** | **127** | **301** | **515** | **884** |
+
+Global AI models such as GraphCast and Pangu-Weather remain far better beyond 12 hours, and the report says so. The
+same rule runs through every model here: results are measured on data the model never trained on, "trained from
+scratch", "fine-tuned" and "not trained yet" are kept apart, and nothing is claimed that has not been measured.
+TriStream-SVS, for example, has no listening tests yet, so it reports no quality scores.
+
+<img src="https://huggingface.co/spaces/CorxLabs/README/resolve/main/rule.png" width="100%" height="4" alt="">
+
+[corx-labs.com](https://www.corx-labs.com/) · [Chat with CorX3.8](https://www.corx-labs.com/chat/) ·
+[Documentation](https://www.corx-labs.com/documentation/) · [Blog](https://www.corx-labs.com/blog/) ·
+[Benchmarks](https://www.corx-labs.com/benchmarks/) · PyPI: [naturev1](https://pypi.org/project/naturev1/),
+[ihelix](https://pypi.org/project/ihelix/), [helix-lm](https://pypi.org/project/helix-lm/)
+
+<sub>CorX Labs · Westmoreland, Jamaica · Architecting the future, from the ground up.</sub>

papers/corx1.5-technical-report.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e30019df008f176e868908f13480df623e6eec614937a708cbdb2d3fc730c0e3
+size 358600

papers/corx3.8-27b-model-report.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f4b05323ce91f11b0c74c0d68dfd38e0151c47064b5535902aecd1a0d5793190
+size 325507

papers/helix-technical-report.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31a72da7d7ffc2659e4b7fd9d029c4ca19417fb43cefda7a6c79c208f57a8fee
+size 376184

papers/naturev1-hurricane-physics-report.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cee46308d9b24233852c17cd66c76e4a6622eff264e2916187fa1cf54f6e858e
+size 463506

papers/tristream-svs-technical-report.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28a902265692c3ef5339704da4babf234ac0c3bc6a4652a9a38a036e067e7427
+size 374585