Rename to BASE-1, professional model card: remove diagram, structured roadmap

README.md

@@ -10,79 +10,99 @@ tags:
 
 <div align="center">
 
-![Base-1 Banner](https://github.com/PlatformNetwork/prism/raw/main/assets/banner.png)
+![BASE-1 Banner](https://github.com/PlatformNetwork/prism/raw/main/assets/banner.png)
 
-# Base-1
+# BASE-1
 
-**A multimodal model discovered through decentralized neural architecture search**
+**A multimodal foundation model whose architecture is discovered through decentralized neural architecture search**
 
 [![PRISM](https://img.shields.io/badge/Built%20with-PRISM-6f42c1.svg)](https://github.com/PlatformNetwork/prism)
 [![Status](https://img.shields.io/badge/Status-In%20Development-orange.svg)]()
 [![Modality](https://img.shields.io/badge/Modality-Text%20%2F%20Image%20%E2%86%92%20Text-blue.svg)]()
+[![License](https://img.shields.io/badge/License-Apache%202.0-green.svg)](https://www.apache.org/licenses/LICENSE-2.0)
 
 </div>
 
 ---
 
-## ⚠️ Status: In Development
+## Status: In Development
 
-**Base-1 is currently under active development.** No weights are available yet. This repository will host the model once the architecture search and training phases are complete.
+BASE-1 is currently under active development. No weights are available yet. This repository will host the model checkpoints, configuration, and usage documentation once the architecture search and training phases are complete.
+
+## Model Summary
+
+| | |
+|---|---|
+| **Developer** | CortexLM |
+| **Architecture** | Determined by neural architecture search (in progress) |
+| **Parameters** | To be announced after architecture search |
+| **Input modalities** | Text, Image |
+| **Output modality** | Text |
+| **Architecture search** | [PRISM](https://github.com/PlatformNetwork/prism) — decentralized NAS on the Platform Network |
+| **License** | Apache 2.0 |
 
 ## Overview
 
-Base-1 is a model being developed through [**PRISM**](https://github.com/PlatformNetwork/prism), a decentralized neural architecture search (NAS) subnet on the Platform Network. Instead of hand-picking an architecture upfront, Base-1's design is being **discovered competitively**: miners across the network submit novel architecture families and training recipes, which are evaluated in isolated benchmark environments for learning quality, training stability, and scaling behavior.
+BASE-1 is a foundation model being developed through [PRISM](https://github.com/PlatformNetwork/prism), a decentralized neural architecture search (NAS) challenge running on the Platform Network. Rather than committing to a hand-designed architecture upfront, BASE-1's design is discovered competitively: miners across the network submit novel architecture families and training recipes, which are evaluated in isolated benchmark environments for learning quality, training stability, and scaling behavior.
+
+The best-performing architecture that emerges from this search will be used to train BASE-1 at scale.
+
+### How the architecture is discovered
+
+PRISM fixes the dataset and evaluation protocol, not the search space. Candidate submissions are scored on:
+
+- **Learning quality** — proxy loss performance under a shared, deterministic evaluation contract
+- **Training stability** — smooth loss curves, stable gradients, and well-behaved activations
+- **Scaling signals** — consistent improvements across model size, depth, sequence length, and batch scaling
+- **Noise resistance** — dynamic thresholds prevent marginal random fluctuations from being rewarded as improvements
 
-The best-performing architecture that emerges from this search will be used to train Base-1.
+Architecture discovery and training-recipe improvements (optimizer, loss computation, inference, train step) are attributed and rewarded independently, so both the model design and its training procedure are optimized by the network.
 
 ## Modalities
 
+BASE-1 will support **Text/Image to Text**: it will accept text and images as input and generate text as output.
+
 | Input | Output |
 |-------|--------|
-| Text 📝 | Text 📝 |
-| Image 🖼️ | Text 📝 |
-
-Base-1 will support **Text/Image → Text**: it will accept both text and images as input and generate text as output.
+| Text | Text |
+| Image | Text |
 
-## Why is the model size unknown?
+## Why is the model size not announced?
 
-The parameter count of Base-1 is **intentionally undisclosed for now — because it is genuinely not decided yet.**
+The parameter count of BASE-1 is genuinely not decided yet — and this is by design.
 
-In a conventional training pipeline, you fix an architecture and a parameter budget first, then train. Base-1 inverts this process:
+In a conventional training pipeline, the architecture and parameter budget are fixed first, then training begins. BASE-1 inverts this process:
 
-1. **Architecture search comes first.** PRISM evaluates candidate architectures at compact proxy scales, measuring loss curves, gradient stability, activation behavior, and how performance scales across model size, depth, sequence length, and batch size.
-2. **Scaling laws are derived from the winner.** Each architecture family has its own scaling behavior. The optimal parameter count depends on the scaling-law signals of the architecture that wins the search — a number that simply cannot be known before the search concludes.
+1. **Architecture search comes first.** PRISM evaluates candidate architectures at compact proxy scales, measuring loss curves, gradient stability, activation behavior, and how performance evolves across model size, depth, sequence length, and batch size.
+2. **Scaling laws are derived from the winning architecture.** Each architecture family exhibits its own scaling behavior. The optimal parameter count depends on the scaling-law signals of the architecture that wins the search — a number that cannot be known before the search concludes.
 3. **The final size is chosen from evidence, not convention.** Once the winning architecture's scaling characteristics are measured, the parameter budget will be set where the compute/performance trade-off is optimal for that specific design.
 
-The final model size will be announced **after the architecture search is complete**.
+The final model size will be announced once the architecture search is complete.
 
-## How PRISM discovers the architecture
+## Roadmap
 
-```mermaid
-flowchart LR
-    A[Miners submit<br/>architectures & recipes] --> B[Isolated evaluation<br/>environments]
-    B --> C[Scoring: learning quality,<br/>stability, scaling signals]
-    C --> D[Best architecture<br/>selected]
-    D --> E[Base-1 training]
-    E --> F[Weights released<br/>here]
-```
+| Phase | Description | Status |
+|-------|-------------|--------|
+| 1. PRISM challenge launch | Open the decentralized architecture search to miners on the Platform Network | In progress |
+| 2. Architecture selection | Identify the best-performing architecture family from competitive evaluation and scaling analysis | Pending |
+| 3. Dataset curation | Assemble and validate the large-scale multimodal training corpus | Pending |
+| 4. Large-scale training | Train BASE-1 at the parameter budget derived from the winning architecture's scaling laws | Pending |
+| 5. Model release | Publish weights, configuration, evaluation results, and usage documentation in this repository | Pending |
 
-- **Decentralized search**: architecture and training ideas are sourced from a competitive network of miners rather than a single research team.
-- **Scaling-aware evaluation**: candidates are rewarded for smooth loss curves, stable gradients, and consistent improvements across scales — not just raw benchmark numbers.
-- **Separate ownership**: architecture discovery and training-recipe improvements are attributed and rewarded independently, so both the design and the training procedure are optimized.
+## Intended Use
 
-## Roadmap
+BASE-1 is intended as a general-purpose multimodal foundation model for text generation conditioned on text and image inputs. Detailed intended-use guidance, limitations, and evaluation results will be published with the model release.
+
+## Evaluation
 
-- [x] Repository created
-- [ ] Neural architecture search via PRISM *(in progress)*
-- [ ] Final architecture & model size announcement
-- [ ] Training
-- [ ] Weights release
+Benchmark results will be published alongside the weights once training is complete. Architecture-search-stage evaluations follow the PRISM scoring protocol, documented in [Scoring and rewards](https://github.com/PlatformNetwork/prism/blob/main/docs/scoring.md) and [Scaling evaluation](https://github.com/PlatformNetwork/prism/blob/main/docs/scaling.md).
 
-## Links
+## Resources
 
-- 🔬 **PRISM (architecture search)**: [github.com/PlatformNetwork/prism](https://github.com/PlatformNetwork/prism)
-- 📖 **PRISM documentation**: [Overview](https://github.com/PlatformNetwork/prism/blob/main/docs/overview.md) • [Scoring](https://github.com/PlatformNetwork/prism/blob/main/docs/scoring.md) • [Scaling evaluation](https://github.com/PlatformNetwork/prism/blob/main/docs/scaling.md)
+- PRISM (architecture search): [github.com/PlatformNetwork/prism](https://github.com/PlatformNetwork/prism)
+- PRISM documentation: [Overview](https://github.com/PlatformNetwork/prism/blob/main/docs/overview.md) | [Architecture](https://github.com/PlatformNetwork/prism/blob/main/docs/architecture.md) | [Scoring](https://github.com/PlatformNetwork/prism/blob/main/docs/scoring.md) | [Scaling](https://github.com/PlatformNetwork/prism/blob/main/docs/scaling.md)
+- Platform Network: [platform.network](https://platform.network)
 
 ## License
 
-Apache 2.0
+This repository is released under the [Apache License 2.0](https://www.apache.org/licenses/LICENSE-2.0).