Org card: Hopit as a continual-learning lab; fashion as track record

README.md

@@ -1,68 +1,66 @@
 ---
 title: README
-emoji: 👗
+emoji: 🔁
 colorFrom: indigo
 colorTo: gray
 sdk: static
 pinned: false
 ---
 
-# Hopit AI
+# Hopit
 
-World models for fashion: search, retrieval, ranking, and multimodal reasoning.
+**AI that gets better at its job by doing it.**
 
-We publish two tracks. Each one ships its weights, its evaluation protocol, and the
-prediction files behind every number we quote, so our claims can be checked rather
-than taken on trust.
+Hopit is a continual-learning lab built in Asia-Pacific. We build models and
+harnesses that learn new tasks, keep what they already know, and improve inside
+the enterprises that use them. → [hopit.ai](https://hopit.ai)
 
-## Moda: search and retrieval
+## Decision models
 
-Fashion image encoders for search, recommendation, and visual similarity. The
-distilled encoder is the one we serve, and it is the backbone under the attribute
-models below.
+Small models that answer a typed decision question in one forward pass, with a
+calibrated probability for each option.
 
-| Model | What it is |
-|---|---|
-| [moda-fashion-distilled](https://huggingface.co/HopitAI/moda-fashion-distilled) | The served encoder. Start here. |
-| [moda-fashion-matryoshka](https://huggingface.co/HopitAI/moda-fashion-matryoshka) | Nested embeddings, so one model serves several dimensions |
-| [moda-fashion-distilled-512d](https://huggingface.co/HopitAI/moda-fashion-distilled-512d) | Smaller embeddings for tighter index budgets |
-| [moda-fashion-crossdomain](https://huggingface.co/HopitAI/moda-fashion-crossdomain) | Street photos matched to catalogue product shots |
-| [moda-fashion-vision-fp16](https://huggingface.co/HopitAI/moda-fashion-vision-fp16) | Half-precision vision tower for cheaper inference |
+| Model | What it is | Where it stands |
+|---|---|---|
+| [hopper](https://huggingface.co/HopitAI/hopper) | Hopper 1.0 — LoRA on Qwen3.5-4B | #2 in JevBench's Jev-class capability ranking, 63.5 against the leader's 64.7¹ |
+| [hopper-g](https://huggingface.co/HopitAI/hopper-g) | Hopper (G) — general-purpose, 4.66B served | Top five of 46 systems under 5B on the Jev Decision Index² |
 
-Benchmarks: [hopit-ai.github.io/Moda](https://hopit-ai.github.io/Moda/) ·
-Code: [github.com/hopit-ai/Moda](https://github.com/hopit-ai/Moda)
+Both are released for research and demonstration only — see each card for its
+licence and training data. Code: [hopit-ai/hopper](https://github.com/hopit-ai/hopper).
 
-## MODA_NER: attribute extraction
+## Continual learning
 
-Turning a fashion image into structured product data: category, colour, fit,
-neckline, sleeve length, pattern, material. Each model is scored on one frozen
-track of the MODA General Attribute Suite, and the tracks are never averaged
-together, because a model can be strong on clean product shots and weak on
-full-body photos.
+Our update method lifted an internal tool-use evaluation from 57.9% to 66.1%
+without losing earlier abilities on the retention suites we track. That is one
+seed and an internal measurement; we will publish the protocol and artifacts
+before treating it as established.
 
-| Model | Input | Licence |
-|---|---|---|
-| [moda-ner-v-crop](https://huggingface.co/HopitAI/moda-ner-v-crop) | A cropped garment | MIT |
-| [moda-ner-v-catalog](https://huggingface.co/HopitAI/moda-ner-v-catalog) | A catalogue product image | CC BY-NC 4.0 |
-| [moda-ner-v-fullbody](https://huggingface.co/HopitAI/moda-ner-v-fullbody) | A full-body photo | CC BY-NC 4.0 |
+## Track record
 
-Two of these are evaluated against research-only corpora whose terms reach derived
-data, so their weights are non-commercial. That binds us too: those weights are not
-part of our paid product.
+Before continual learning, we built open models and public benchmark suites for
+fashion, and held ourselves to them in public. They remain the standard our newer
+work has to clear.
+
+| Track | Models | Benchmark |
+|---|---|---|
+| Retrieval | [moda-fashion-distilled](https://huggingface.co/HopitAI/moda-fashion-distilled) · [matryoshka](https://huggingface.co/HopitAI/moda-fashion-matryoshka) · [crossdomain](https://huggingface.co/HopitAI/moda-fashion-crossdomain) · [pro-lite](https://huggingface.co/HopitAI/moda-pro-lite) | [MODA](https://hopit-ai.github.io/Moda/) |
+| Attribute extraction | [crop](https://huggingface.co/HopitAI/moda-ner-v-crop) (MIT) · [catalog](https://huggingface.co/HopitAI/moda-ner-v-catalog) · [full-body](https://huggingface.co/HopitAI/moda-ner-v-fullbody) (CC BY-NC 4.0) | [MODA_NER](https://hopit-ai.github.io/Moda_ner/) |
 
-MODA_NER Pro beats FashionCLIP on all 10 catalogue attributes and 17 of 18 on
-full-body — 31% and 19% better on average — trained on your catalogue and mapped to
-your taxonomy, not the benchmarks'. [Talk to us](https://hopit.ai).
+## How we publish
 
-Benchmarks: [hopit-ai.github.io/Moda_ner](https://hopit-ai.github.io/Moda_ner/) ·
-Code: [github.com/hopit-ai/Moda_ner](https://github.com/hopit-ai/Moda_ner)
+- Every rank is quoted with its qualifier.
+- Protocols are frozen before inference, predictions are hashed before labels
+  open, and scorers fail closed.
+- The runs we lose are published beside the runs we win.
 
-## How we evaluate
+## Work with us
 
-We freeze the test set before running a model, keep related images out of both
-training and test data, hash every prediction file before scoring it, score every
-row instead of quietly dropping the awkward ones, and require a real gain on each
-track rather than a good average. We publish the losing runs alongside the winning
-ones.
+We deploy with a small forward-deployed team inside your environment. Your data
+and your deployed models stay yours. The approach is ideal for regulated
+enterprises. → [hopit.ai](https://hopit.ai)
 
-Contact: [hopit.ai](https://hopit.ai)
+<sub>¹ JevBench v1.4.2, 24 September 2026 snapshot, scored by an independent
+maintainer. Second on capability; fifth on the composite score, which also
+weighs speed and cost. ² Jev Decision Index 0.2.1, 28 September 2026: Hopper (G)
+1.2 is third of 46 systems under 5B on the chance-corrected headline score
+(40.77), within 0.1 of fourth, and 18th of 70 overall. The edition is 76% scored.</sub>